Publications
Publications
CSL Publications
Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magn. Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magn
2026
Patras, Alexandros; Lalis, Spyros; Antonopoulos, Christos D.; Bellas, Nikolaos
DPUConfig: Optimizing ML Inference in FPGAs Using Reinforcement Learning Miscellaneous
2026, (arXiv:2602.12847 [cs]).
@misc{patrasDPUConfigOptimizingML2026,
title = {DPUConfig: Optimizing ML Inference in FPGAs Using Reinforcement Learning},
author = {Alexandros Patras and Spyros Lalis and Christos D. Antonopoulos and Nikolaos Bellas},
url = {http://arxiv.org/abs/2602.12847},
doi = {10.48550/arXiv.2602.12847},
year = {2026},
date = {2026-02-01},
urldate = {2026-02-18},
publisher = {arXiv},
abstract = {Heterogeneous embedded systems, with diverse computing elements and accelerators such as FPGAs, offer a promising platform for fast and flexible ML inference, which is crucial for services such as autonomous driving and augmented reality, where delays can be costly. However, efficiently allocating computational resources for deep learning applications in FPGA-based systems is a challenging task. A Deep Learning Processor Unit (DPU) is a parameterizable FPGA-based accelerator module optimized for ML inference. It supports a wide range of ML models and can be instantiated multiple times within a single FPGA to enable concurrent execution. This paper introduces DPUConfig, a novel runtime management framework, based on a custom Reinforcement Learning (RL) agent, that dynamically selects optimal DPU configurations by leveraging real-time telemetry data monitoring, system utilization, power consumption, and application performance to inform its configuration selection decisions. The experimental evaluation demonstrates that the RL agent achieves energy efficiency 95% (on average) of the optimal attainable energy efficiency for several CNN models on the Xilinx Zynq UltraScale+ MPSoC ZCU102.},
note = {arXiv:2602.12847 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Heterogeneous embedded systems, with diverse computing elements and accelerators such as FPGAs, offer a promising platform for fast and flexible ML inference, which is crucial for services such as autonomous driving and augmented reality, where delays can be costly. However, efficiently allocating computational resources for deep learning applications in FPGA-based systems is a challenging task. A Deep Learning Processor Unit (DPU) is a parameterizable FPGA-based accelerator module optimized for ML inference. It supports a wide range of ML models and can be instantiated multiple times within a single FPGA to enable concurrent execution. This paper introduces DPUConfig, a novel runtime management framework, based on a custom Reinforcement Learning (RL) agent, that dynamically selects optimal DPU configurations by leveraging real-time telemetry data monitoring, system utilization, power consumption, and application performance to inform its configuration selection decisions. The experimental evaluation demonstrates that the RL agent achieves energy efficiency 95% (on average) of the optimal attainable energy efficiency for several CNN models on the Xilinx Zynq UltraScale+ MPSoC ZCU102.
Patras, Alexandros; Lalis, Spyros; Antonopoulos, Christos D.; Bellas, Nikolaos
DPUConfig: Optimizing ML Inference in FPGAs Using Reinforcement Learning Miscellaneous
2026, (arXiv:2602.12847 [cs]).
@misc{patras_dpuconfig_2026,
title = {DPUConfig: Optimizing ML Inference in FPGAs Using Reinforcement Learning},
author = {Alexandros Patras and Spyros Lalis and Christos D. Antonopoulos and Nikolaos Bellas},
url = {http://arxiv.org/abs/2602.12847},
doi = {10.48550/arXiv.2602.12847},
year = {2026},
date = {2026-02-01},
urldate = {2026-02-18},
publisher = {arXiv},
abstract = {Heterogeneous embedded systems, with diverse computing elements and accelerators such as FPGAs, offer a promising platform for fast and flexible ML inference, which is crucial for services such as autonomous driving and augmented reality, where delays can be costly. However, efficiently allocating computational resources for deep learning applications in FPGA-based systems is a challenging task. A Deep Learning Processor Unit (DPU) is a parameterizable FPGA-based accelerator module optimized for ML inference. It supports a wide range of ML models and can be instantiated multiple times within a single FPGA to enable concurrent execution. This paper introduces DPUConfig, a novel runtime management framework, based on a custom Reinforcement Learning (RL) agent, that dynamically selects optimal DPU configurations by leveraging real-time telemetry data monitoring, system utilization, power consumption, and application performance to inform its configuration selection decisions. The experimental evaluation demonstrates that the RL agent achieves energy efficiency 95% (on average) of the optimal attainable energy efficiency for several CNN models on the Xilinx Zynq UltraScale+ MPSoC ZCU102.},
note = {arXiv:2602.12847 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Heterogeneous embedded systems, with diverse computing elements and accelerators such as FPGAs, offer a promising platform for fast and flexible ML inference, which is crucial for services such as autonomous driving and augmented reality, where delays can be costly. However, efficiently allocating computational resources for deep learning applications in FPGA-based systems is a challenging task. A Deep Learning Processor Unit (DPU) is a parameterizable FPGA-based accelerator module optimized for ML inference. It supports a wide range of ML models and can be instantiated multiple times within a single FPGA to enable concurrent execution. This paper introduces DPUConfig, a novel runtime management framework, based on a custom Reinforcement Learning (RL) agent, that dynamically selects optimal DPU configurations by leveraging real-time telemetry data monitoring, system utilization, power consumption, and application performance to inform its configuration selection decisions. The experimental evaluation demonstrates that the RL agent achieves energy efficiency 95% (on average) of the optimal attainable energy efficiency for several CNN models on the Xilinx Zynq UltraScale+ MPSoC ZCU102.
Tirana, Joana; Lalis, Spyros; Chatzopoulos, Dimitris
Implementation and Evaluation of Multi-Hop Parallel Split Learning Journal Article
In: IEEE Access, vol. 14, pp. 9419–9434, 2026, ISSN: 2169-3536.
@article{tiranaImplementationEvaluationMultiHop2026,
title = {Implementation and Evaluation of Multi-Hop Parallel Split Learning},
author = {Joana Tirana and Spyros Lalis and Dimitris Chatzopoulos},
url = {https://ieeexplore.ieee.org/document/11348123/},
doi = {10.1109/ACCESS.2026.3653864},
issn = {2169-3536},
year = {2026},
date = {2026-01-01},
urldate = {2026-02-18},
journal = {IEEE Access},
volume = {14},
pages = {9419–9434},
abstract = {Offloading techniques like Split Federated Learning (SplitFed) or Parallel Split Learning (Parallel SL) enable multiple resource-constrained data owner devices to participate in collaborative training processes with the help of resourceful compute nodes. Recent findings indicate that the number of compute nodes (hops) significantly impacts the training delay. Yet, determining the ideal number of hops is not an easy task. Therefore, in this work, we propose a mathematical model that estimates the training delay of single- and multi-hop Parallel SL. This tool not only helps in determining the optimal number of hops before deployment, but also serves as an evaluation tool in future research works. Further, we construct a lightweight optimization problem that targets maximizing the pipeline parallelism at a theoretical level. Also, we present the SplitPipe framework, which allows the support of pipeline parallelism at the system level as well. Finally, we conduct a thorough numerical evaluation, which first validates the accuracy of the proposed estimation model and then presents a detailed analysis of multi- and single-hop Parallel SL.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Offloading techniques like Split Federated Learning (SplitFed) or Parallel Split Learning (Parallel SL) enable multiple resource-constrained data owner devices to participate in collaborative training processes with the help of resourceful compute nodes. Recent findings indicate that the number of compute nodes (hops) significantly impacts the training delay. Yet, determining the ideal number of hops is not an easy task. Therefore, in this work, we propose a mathematical model that estimates the training delay of single- and multi-hop Parallel SL. This tool not only helps in determining the optimal number of hops before deployment, but also serves as an evaluation tool in future research works. Further, we construct a lightweight optimization problem that targets maximizing the pipeline parallelism at a theoretical level. Also, we present the SplitPipe framework, which allows the support of pipeline parallelism at the system level as well. Finally, we conduct a thorough numerical evaluation, which first validates the accuracy of the proposed estimation model and then presents a detailed analysis of multi- and single-hop Parallel SL.
Tirana, Joana; Lalis, Spyros; Chatzopoulos, Dimitris
Implementation and Evaluation of Multi-Hop Parallel Split Learning Journal Article
In: IEEE Access, vol. 14, pp. 9419–9434, 2026, ISSN: 2169-3536.
@article{tirana_implementation_2026,
title = {Implementation and Evaluation of Multi-Hop Parallel Split Learning},
author = {Joana Tirana and Spyros Lalis and Dimitris Chatzopoulos},
url = {https://ieeexplore.ieee.org/document/11348123/},
doi = {10.1109/ACCESS.2026.3653864},
issn = {2169-3536},
year = {2026},
date = {2026-01-01},
urldate = {2026-02-18},
journal = {IEEE Access},
volume = {14},
pages = {9419–9434},
abstract = {Offloading techniques like Split Federated Learning (SplitFed) or Parallel Split Learning (Parallel SL) enable multiple resource-constrained data owner devices to participate in collaborative training processes with the help of resourceful compute nodes. Recent findings indicate that the number of compute nodes (hops) significantly impacts the training delay. Yet, determining the ideal number of hops is not an easy task. Therefore, in this work, we propose a mathematical model that estimates the training delay of single- and multi-hop Parallel SL. This tool not only helps in determining the optimal number of hops before deployment, but also serves as an evaluation tool in future research works. Further, we construct a lightweight optimization problem that targets maximizing the pipeline parallelism at a theoretical level. Also, we present the SplitPipe framework, which allows the support of pipeline parallelism at the system level as well. Finally, we conduct a thorough numerical evaluation, which first validates the accuracy of the proposed estimation model and then presents a detailed analysis of multi- and single-hop Parallel SL.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Offloading techniques like Split Federated Learning (SplitFed) or Parallel Split Learning (Parallel SL) enable multiple resource-constrained data owner devices to participate in collaborative training processes with the help of resourceful compute nodes. Recent findings indicate that the number of compute nodes (hops) significantly impacts the training delay. Yet, determining the ideal number of hops is not an easy task. Therefore, in this work, we propose a mathematical model that estimates the training delay of single- and multi-hop Parallel SL. This tool not only helps in determining the optimal number of hops before deployment, but also serves as an evaluation tool in future research works. Further, we construct a lightweight optimization problem that targets maximizing the pipeline parallelism at a theoretical level. Also, we present the SplitPipe framework, which allows the support of pipeline parallelism at the system level as well. Finally, we conduct a thorough numerical evaluation, which first validates the accuracy of the proposed estimation model and then presents a detailed analysis of multi- and single-hop Parallel SL.
2025
Polychronis, Giorgos; Pournaropoulos, Foivos; Antonopoulos, Christos D.; Lalis, Spyros
Using Machine Learning to Take Stay-or-Go Decisions in Data-driven Drone Missions Miscellaneous
2025, (arXiv:2512.04773 [cs]).
@misc{polychronisUsingMachineLearning2025,
title = {Using Machine Learning to Take Stay-or-Go Decisions in Data-driven Drone Missions},
author = {Giorgos Polychronis and Foivos Pournaropoulos and Christos D. Antonopoulos and Spyros Lalis},
url = {http://arxiv.org/abs/2512.04773},
doi = {10.48550/arXiv.2512.04773},
year = {2025},
date = {2025-12-01},
urldate = {2026-02-18},
publisher = {arXiv},
abstract = {Drones are becoming indispensable in many application domains. In data-driven missions, besides sensing, the drone must process the collected data at runtime to decide whether additional action must be taken on the spot, before moving to the next point of interest. If processing does not reveal an event or situation that requires such an action, the drone has waited in vain instead of moving to the next point. If, however, the drone starts moving to the next point and it turns out that a follow-up action is needed at the previous point, it must spend time to fly-back. To take this decision, we propose different machine-learning methods based on branch prediction and reinforcement learning. We evaluate these methods for a wide range of scenarios where the probability of event occurrence changes with time. Our results show that the proposed methods consistently outperform the regression-based method proposed in the literature and can significantly improve the worst-case mission time by up to 4.1x. Also, the achieved median mission time is very close, merely up to 2.7% higher, to that of a method with perfect knowledge of the current underlying event probability at each point of interest.},
note = {arXiv:2512.04773 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Drones are becoming indispensable in many application domains. In data-driven missions, besides sensing, the drone must process the collected data at runtime to decide whether additional action must be taken on the spot, before moving to the next point of interest. If processing does not reveal an event or situation that requires such an action, the drone has waited in vain instead of moving to the next point. If, however, the drone starts moving to the next point and it turns out that a follow-up action is needed at the previous point, it must spend time to fly-back. To take this decision, we propose different machine-learning methods based on branch prediction and reinforcement learning. We evaluate these methods for a wide range of scenarios where the probability of event occurrence changes with time. Our results show that the proposed methods consistently outperform the regression-based method proposed in the literature and can significantly improve the worst-case mission time by up to 4.1x. Also, the achieved median mission time is very close, merely up to 2.7% higher, to that of a method with perfect knowledge of the current underlying event probability at each point of interest.
Polychronis, Giorgos; Pournaropoulos, Foivos; Antonopoulos, Christos D.; Lalis, Spyros
Using Machine Learning to Take Stay-or-Go Decisions in Data-driven Drone Missions Miscellaneous
2025, (arXiv:2512.04773 [cs]).
@misc{polychronis_using_2025,
title = {Using Machine Learning to Take Stay-or-Go Decisions in Data-driven Drone Missions},
author = {Giorgos Polychronis and Foivos Pournaropoulos and Christos D. Antonopoulos and Spyros Lalis},
url = {http://arxiv.org/abs/2512.04773},
doi = {10.48550/arXiv.2512.04773},
year = {2025},
date = {2025-12-01},
urldate = {2026-02-18},
publisher = {arXiv},
abstract = {Drones are becoming indispensable in many application domains. In data-driven missions, besides sensing, the drone must process the collected data at runtime to decide whether additional action must be taken on the spot, before moving to the next point of interest. If processing does not reveal an event or situation that requires such an action, the drone has waited in vain instead of moving to the next point. If, however, the drone starts moving to the next point and it turns out that a follow-up action is needed at the previous point, it must spend time to fly-back. To take this decision, we propose different machine-learning methods based on branch prediction and reinforcement learning. We evaluate these methods for a wide range of scenarios where the probability of event occurrence changes with time. Our results show that the proposed methods consistently outperform the regression-based method proposed in the literature and can significantly improve the worst-case mission time by up to 4.1x. Also, the achieved median mission time is very close, merely up to 2.7% higher, to that of a method with perfect knowledge of the current underlying event probability at each point of interest.},
note = {arXiv:2512.04773 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Drones are becoming indispensable in many application domains. In data-driven missions, besides sensing, the drone must process the collected data at runtime to decide whether additional action must be taken on the spot, before moving to the next point of interest. If processing does not reveal an event or situation that requires such an action, the drone has waited in vain instead of moving to the next point. If, however, the drone starts moving to the next point and it turns out that a follow-up action is needed at the previous point, it must spend time to fly-back. To take this decision, we propose different machine-learning methods based on branch prediction and reinforcement learning. We evaluate these methods for a wide range of scenarios where the probability of event occurrence changes with time. Our results show that the proposed methods consistently outperform the regression-based method proposed in the literature and can significantly improve the worst-case mission time by up to 4.1x. Also, the achieved median mission time is very close, merely up to 2.7% higher, to that of a method with perfect knowledge of the current underlying event probability at each point of interest.
Guan, Jiexiong; Hu, Zhenqing; Antonopoulos, Christos D.; Bellas, Nikolaos; Lalis, Spyros; Smirni, Evgenia; Zhou, Gang; Agrawal, Gagan; Ren, Bin
TMModel: Modeling Texture Memory and Mobile GPU Performance to Accelerate DNN Computations Proceedings Article
In: Proceedings of the 39th ACM International Conference on Supercomputing, pp. 205–220, ACM, Salt Lake City USA, 2025, ISBN: 979-8-4007-1537-2.
@inproceedings{guanTMModelModelingTexture2025,
title = {TMModel: Modeling Texture Memory and Mobile GPU Performance to Accelerate DNN Computations},
author = {Jiexiong Guan and Zhenqing Hu and Christos D. Antonopoulos and Nikolaos Bellas and Spyros Lalis and Evgenia Smirni and Gang Zhou and Gagan Agrawal and Bin Ren},
url = {https://dl.acm.org/doi/10.1145/3721145.3725774},
doi = {10.1145/3721145.3725774},
isbn = {979-8-4007-1537-2},
year = {2025},
date = {2025-06-01},
urldate = {2026-02-18},
booktitle = {Proceedings of the 39th ACM International Conference on Supercomputing},
pages = {205–220},
publisher = {ACM},
address = {Salt Lake City USA},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Guan, Jiexiong; Hu, Zhenqing; Antonopoulos, Christos D.; Bellas, Nikolaos; Lalis, Spyros; Smirni, Evgenia; Zhou, Gang; Agrawal, Gagan; Ren, Bin
TMModel: Modeling Texture Memory and Mobile GPU Performance to Accelerate DNN Computations Proceedings Article
In: Proceedings of the 39th ACM International Conference on Supercomputing, pp. 205–220, ACM, Salt Lake City USA, 2025, ISBN: 979-8-4007-1537-2.
@inproceedings{guan_tmmodel_2025,
title = {TMModel: Modeling Texture Memory and Mobile GPU Performance to Accelerate DNN Computations},
author = {Jiexiong Guan and Zhenqing Hu and Christos D. Antonopoulos and Nikolaos Bellas and Spyros Lalis and Evgenia Smirni and Gang Zhou and Gagan Agrawal and Bin Ren},
url = {https://dl.acm.org/doi/10.1145/3721145.3725774},
doi = {10.1145/3721145.3725774},
isbn = {979-8-4007-1537-2},
year = {2025},
date = {2025-06-01},
urldate = {2026-02-18},
booktitle = {Proceedings of the 39th ACM International Conference on Supercomputing},
pages = {205–220},
publisher = {ACM},
address = {Salt Lake City USA},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Gkeka, Maria Rafaela; Sun, Bowen; Smirni, Evgenia; Antonopoulos, Christos D.; Lalis, Spyros; Bellas, Nikolaos
Black-box Adversarial Attacks on CNN-based SLAM Algorithms Miscellaneous
2025, (arXiv:2505.24654 [cs]).
@misc{gkekaBlackboxAdversarialAttacks2025,
title = {Black-box Adversarial Attacks on CNN-based SLAM Algorithms},
author = {Maria Rafaela Gkeka and Bowen Sun and Evgenia Smirni and Christos D. Antonopoulos and Spyros Lalis and Nikolaos Bellas},
url = {http://arxiv.org/abs/2505.24654},
doi = {10.48550/arXiv.2505.24654},
year = {2025},
date = {2025-05-01},
urldate = {2026-02-18},
publisher = {arXiv},
abstract = {Continuous advancements in deep learning have led to significant progress in feature detection, resulting in enhanced accuracy in tasks like Simultaneous Localization and Mapping (SLAM). Nevertheless, the vulnerability of deep neural networks to adversarial attacks remains a challenge for their reliable deployment in applications, such as navigation of autonomous agents. Even though CNN-based SLAM algorithms are a growing area of research there is a notable absence of a comprehensive presentation and examination of adversarial attacks targeting CNN-based feature detectors, as part of a SLAM system. Our work introduces black-box adversarial perturbations applied to the RGB images fed into the GCN-SLAM algorithm. Our findings on the TUM dataset [30] reveal that even attacks of moderate scale can lead to tracking failure in as many as 76% of the frames. Moreover, our experiments highlight the catastrophic impact of attacking depth instead of RGB input images on the SLAM system.},
note = {arXiv:2505.24654 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Continuous advancements in deep learning have led to significant progress in feature detection, resulting in enhanced accuracy in tasks like Simultaneous Localization and Mapping (SLAM). Nevertheless, the vulnerability of deep neural networks to adversarial attacks remains a challenge for their reliable deployment in applications, such as navigation of autonomous agents. Even though CNN-based SLAM algorithms are a growing area of research there is a notable absence of a comprehensive presentation and examination of adversarial attacks targeting CNN-based feature detectors, as part of a SLAM system. Our work introduces black-box adversarial perturbations applied to the RGB images fed into the GCN-SLAM algorithm. Our findings on the TUM dataset [30] reveal that even attacks of moderate scale can lead to tracking failure in as many as 76% of the frames. Moreover, our experiments highlight the catastrophic impact of attacking depth instead of RGB input images on the SLAM system.
Gkeka, Maria Rafaela; Sun, Bowen; Smirni, Evgenia; Antonopoulos, Christos D.; Lalis, Spyros; Bellas, Nikolaos
Black-box Adversarial Attacks on CNN-based SLAM Algorithms Miscellaneous
2025, (arXiv:2505.24654 [cs]).
@misc{gkeka_black-box_2025,
title = {Black-box Adversarial Attacks on CNN-based SLAM Algorithms},
author = {Maria Rafaela Gkeka and Bowen Sun and Evgenia Smirni and Christos D. Antonopoulos and Spyros Lalis and Nikolaos Bellas},
url = {http://arxiv.org/abs/2505.24654},
doi = {10.48550/arXiv.2505.24654},
year = {2025},
date = {2025-05-01},
urldate = {2026-02-18},
publisher = {arXiv},
abstract = {Continuous advancements in deep learning have led to significant progress in feature detection, resulting in enhanced accuracy in tasks like Simultaneous Localization and Mapping (SLAM). Nevertheless, the vulnerability of deep neural networks to adversarial attacks remains a challenge for their reliable deployment in applications, such as navigation of autonomous agents. Even though CNN-based SLAM algorithms are a growing area of research there is a notable absence of a comprehensive presentation and examination of adversarial attacks targeting CNN-based feature detectors, as part of a SLAM system. Our work introduces black-box adversarial perturbations applied to the RGB images fed into the GCN-SLAM algorithm. Our findings on the TUM dataset [30] reveal that even attacks of moderate scale can lead to tracking failure in as many as 76% of the frames. Moreover, our experiments highlight the catastrophic impact of attacking depth instead of RGB input images on the SLAM system.},
note = {arXiv:2505.24654 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Continuous advancements in deep learning have led to significant progress in feature detection, resulting in enhanced accuracy in tasks like Simultaneous Localization and Mapping (SLAM). Nevertheless, the vulnerability of deep neural networks to adversarial attacks remains a challenge for their reliable deployment in applications, such as navigation of autonomous agents. Even though CNN-based SLAM algorithms are a growing area of research there is a notable absence of a comprehensive presentation and examination of adversarial attacks targeting CNN-based feature detectors, as part of a SLAM system. Our work introduces black-box adversarial perturbations applied to the RGB images fed into the GCN-SLAM algorithm. Our findings on the TUM dataset [30] reveal that even attacks of moderate scale can lead to tracking failure in as many as 76% of the frames. Moreover, our experiments highlight the catastrophic impact of attacking depth instead of RGB input images on the SLAM system.
Aslanidis, Theodoros; Kosta, Sokol; Lalis, Spyros; Chatzopoulos, Dimitris
Cross-Domain DRL Agents for Efficient Job Placement in the Cloud-Edge Continuum Proceedings Article
In: Proceedings of the 5th Workshop on Machine Learning and Systems, pp. 276–285, Association for Computing Machinery, New York, NY, USA, 2025, ISBN: 979-8-4007-1538-9.
@inproceedings{aslanidisCrossDomainDRLAgents2025,
title = {Cross-Domain DRL Agents for Efficient Job Placement in the Cloud-Edge Continuum},
author = {Theodoros Aslanidis and Sokol Kosta and Spyros Lalis and Dimitris Chatzopoulos},
url = {https://dl.acm.org/doi/10.1145/3721146.3721934},
doi = {10.1145/3721146.3721934},
isbn = {979-8-4007-1538-9},
year = {2025},
date = {2025-04-01},
urldate = {2026-02-18},
booktitle = {Proceedings of the 5th Workshop on Machine Learning and Systems},
pages = {276–285},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
series = {EuroMLSys '25},
abstract = {The growing computational demands of modern applications call for resource management strategies that effectively utilize the strengths of both cloud and edge computing. Deep Reinforcement Learning (DRL) has shown great promise in addressing these challenges, offering advanced decision-making capabilities that optimize resource allocation and system performance. However, deploying DRL agents in cloud-edge continuum infrastructures remains a significant challenge due to their dependence on infrastructure-specific state-action representations. This paper presents a novel architectural framework for DRL agents that incorporates feature extraction and adaptation mechanisms to enable their seamless operation across diverse environments. By transforming state features into an infrastructure-agnostic representation, our approach reduces the need for extensive retraining when system configurations change. Experimental results show that our method outperforms both a heuristic method and a DRL baseline algorithm while achieving faster convergence when infrastructure and workloads change. This work is an important step forward in developing transferable and adaptable DRL solutions for real-world cloud-edge resource management challenges.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The growing computational demands of modern applications call for resource management strategies that effectively utilize the strengths of both cloud and edge computing. Deep Reinforcement Learning (DRL) has shown great promise in addressing these challenges, offering advanced decision-making capabilities that optimize resource allocation and system performance. However, deploying DRL agents in cloud-edge continuum infrastructures remains a significant challenge due to their dependence on infrastructure-specific state-action representations. This paper presents a novel architectural framework for DRL agents that incorporates feature extraction and adaptation mechanisms to enable their seamless operation across diverse environments. By transforming state features into an infrastructure-agnostic representation, our approach reduces the need for extensive retraining when system configurations change. Experimental results show that our method outperforms both a heuristic method and a DRL baseline algorithm while achieving faster convergence when infrastructure and workloads change. This work is an important step forward in developing transferable and adaptable DRL solutions for real-world cloud-edge resource management challenges.
Aslanidis, Theodoros; Kosta, Sokol; Lalis, Spyros; Chatzopoulos, Dimitris
Cross-Domain DRL Agents for Efficient Job Placement in the Cloud-Edge Continuum Proceedings Article
In: Proceedings of the 5th Workshop on Machine Learning and Systems, pp. 276–285, Association for Computing Machinery, New York, NY, USA, 2025, ISBN: 979-8-4007-1538-9.
@inproceedings{aslanidis_cross-domain_2025,
title = {Cross-Domain DRL Agents for Efficient Job Placement in the Cloud-Edge Continuum},
author = {Theodoros Aslanidis and Sokol Kosta and Spyros Lalis and Dimitris Chatzopoulos},
url = {https://dl.acm.org/doi/10.1145/3721146.3721934},
doi = {10.1145/3721146.3721934},
isbn = {979-8-4007-1538-9},
year = {2025},
date = {2025-04-01},
urldate = {2026-02-18},
booktitle = {Proceedings of the 5th Workshop on Machine Learning and Systems},
pages = {276–285},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
series = {EuroMLSys '25},
abstract = {The growing computational demands of modern applications call for resource management strategies that effectively utilize the strengths of both cloud and edge computing. Deep Reinforcement Learning (DRL) has shown great promise in addressing these challenges, offering advanced decision-making capabilities that optimize resource allocation and system performance. However, deploying DRL agents in cloud-edge continuum infrastructures remains a significant challenge due to their dependence on infrastructure-specific state-action representations. This paper presents a novel architectural framework for DRL agents that incorporates feature extraction and adaptation mechanisms to enable their seamless operation across diverse environments. By transforming state features into an infrastructure-agnostic representation, our approach reduces the need for extensive retraining when system configurations change. Experimental results show that our method outperforms both a heuristic method and a DRL baseline algorithm while achieving faster convergence when infrastructure and workloads change. This work is an important step forward in developing transferable and adaptable DRL solutions for real-world cloud-edge resource management challenges.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The growing computational demands of modern applications call for resource management strategies that effectively utilize the strengths of both cloud and edge computing. Deep Reinforcement Learning (DRL) has shown great promise in addressing these challenges, offering advanced decision-making capabilities that optimize resource allocation and system performance. However, deploying DRL agents in cloud-edge continuum infrastructures remains a significant challenge due to their dependence on infrastructure-specific state-action representations. This paper presents a novel architectural framework for DRL agents that incorporates feature extraction and adaptation mechanisms to enable their seamless operation across diverse environments. By transforming state features into an infrastructure-agnostic representation, our approach reduces the need for extensive retraining when system configurations change. Experimental results show that our method outperforms both a heuristic method and a DRL baseline algorithm while achieving faster convergence when infrastructure and workloads change. This work is an important step forward in developing transferable and adaptable DRL solutions for real-world cloud-edge resource management challenges.
Tirana, Joana; Lalis, Spyros; Chatzopoulos, Dimitris
Estimating the Training Time in Single- and Multi-Hop Split Federated Learning Proceedings Article
In: Proceedings of the 8th International Workshop on Edge Systems, Analytics and Networking, pp. 37–42, Association for Computing Machinery, New York, NY, USA, 2025, ISBN: 979-8-4007-1559-4.
@inproceedings{tiranaEstimatingTrainingTime2025,
title = {Estimating the Training Time in Single- and Multi-Hop Split Federated Learning},
author = {Joana Tirana and Spyros Lalis and Dimitris Chatzopoulos},
url = {https://dl.acm.org/doi/10.1145/3721888.3722096},
doi = {10.1145/3721888.3722096},
isbn = {979-8-4007-1559-4},
year = {2025},
date = {2025-03-01},
urldate = {2026-02-18},
booktitle = {Proceedings of the 8th International Workshop on Edge Systems, Analytics and Networking},
pages = {37–42},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
series = {EdgeSys '25},
abstract = {Split Federated Learning (SFL) is an upcoming and promising approach that balances the two main goals of distributed training, i.e., (i) the data remains at the data owners, and (ii) even devices with resource limitations can participate in the training. This is achieved by splitting the model into multiple parts and offloading them to designated compute nodes. Recent findings show that the number of compute nodes (hops) plays a significant role in the training delay. However, determining the ideal number of hops is not an easy task. Therefore, in this work, we propose a mathematical model that estimates the training delay of single- and multi-hop SFL. This tool not only helps in searching the optimal number of hops before the real deployment happens but also can be used as a lightweight evaluation tool in future research works in SFL. Our numerical evaluations show that the model can make correct estimations with an error smaller than 3.86%. Finally, we have constructed a lightweight optimization problem that finds the optimal cut layers (split points) and model part assignment to minimize training delay.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Split Federated Learning (SFL) is an upcoming and promising approach that balances the two main goals of distributed training, i.e., (i) the data remains at the data owners, and (ii) even devices with resource limitations can participate in the training. This is achieved by splitting the model into multiple parts and offloading them to designated compute nodes. Recent findings show that the number of compute nodes (hops) plays a significant role in the training delay. However, determining the ideal number of hops is not an easy task. Therefore, in this work, we propose a mathematical model that estimates the training delay of single- and multi-hop SFL. This tool not only helps in searching the optimal number of hops before the real deployment happens but also can be used as a lightweight evaluation tool in future research works in SFL. Our numerical evaluations show that the model can make correct estimations with an error smaller than 3.86%. Finally, we have constructed a lightweight optimization problem that finds the optimal cut layers (split points) and model part assignment to minimize training delay.
Tirana, Joana; Lalis, Spyros; Chatzopoulos, Dimitris
Estimating the Training Time in Single- and Multi-Hop Split Federated Learning Proceedings Article
In: Proceedings of the 8th International Workshop on Edge Systems, Analytics and Networking, pp. 37–42, Association for Computing Machinery, New York, NY, USA, 2025, ISBN: 979-8-4007-1559-4.
@inproceedings{tirana_estimating_2025,
title = {Estimating the Training Time in Single- and Multi-Hop Split Federated Learning},
author = {Joana Tirana and Spyros Lalis and Dimitris Chatzopoulos},
url = {https://dl.acm.org/doi/10.1145/3721888.3722096},
doi = {10.1145/3721888.3722096},
isbn = {979-8-4007-1559-4},
year = {2025},
date = {2025-03-01},
urldate = {2026-02-18},
booktitle = {Proceedings of the 8th International Workshop on Edge Systems, Analytics and Networking},
pages = {37–42},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
series = {EdgeSys '25},
abstract = {Split Federated Learning (SFL) is an upcoming and promising approach that balances the two main goals of distributed training, i.e., (i) the data remains at the data owners, and (ii) even devices with resource limitations can participate in the training. This is achieved by splitting the model into multiple parts and offloading them to designated compute nodes. Recent findings show that the number of compute nodes (hops) plays a significant role in the training delay. However, determining the ideal number of hops is not an easy task. Therefore, in this work, we propose a mathematical model that estimates the training delay of single- and multi-hop SFL. This tool not only helps in searching the optimal number of hops before the real deployment happens but also can be used as a lightweight evaluation tool in future research works in SFL. Our numerical evaluations show that the model can make correct estimations with an error smaller than 3.86%. Finally, we have constructed a lightweight optimization problem that finds the optimal cut layers (split points) and model part assignment to minimize training delay.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Split Federated Learning (SFL) is an upcoming and promising approach that balances the two main goals of distributed training, i.e., (i) the data remains at the data owners, and (ii) even devices with resource limitations can participate in the training. This is achieved by splitting the model into multiple parts and offloading them to designated compute nodes. Recent findings show that the number of compute nodes (hops) plays a significant role in the training delay. However, determining the ideal number of hops is not an easy task. Therefore, in this work, we propose a mathematical model that estimates the training delay of single- and multi-hop SFL. This tool not only helps in searching the optimal number of hops before the real deployment happens but also can be used as a lightweight evaluation tool in future research works in SFL. Our numerical evaluations show that the model can make correct estimations with an error smaller than 3.86%. Finally, we have constructed a lightweight optimization problem that finds the optimal cut layers (split points) and model part assignment to minimize training delay.
2024
Pournaropoulos, Foivos; Patras, Alexandros; Antonopoulos, Christos D.; Bellas, Nikolaos; Lalis, Spyros
In: Future Generation Computer Systems, vol. 157, pp. 210–225, 2024.
@article{pournaropoulosFluidityProvidingFlexible2024,
title = {Fluidity: Providing flexible deployment and adaptation policy experimentation for serverless and distributed applications spanning cloud–edge–mobile environments},
author = {Foivos Pournaropoulos and Alexandros Patras and Christos D. Antonopoulos and Nikolaos Bellas and Spyros Lalis},
doi = {10.1016/j.future.2024.03.031},
year = {2024},
date = {2024-08-01},
journal = {Future Generation Computer Systems},
volume = {157},
pages = {210–225},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Pournaropoulos, Foivos; Patras, Alexandros; Antonopoulos, Christos D.; Bellas, Nikolaos; Lalis, Spyros
In: Future Generation Computer Systems, vol. 157, pp. 210–225, 2024.
@article{pournaropoulos_fluidity_2024,
title = {Fluidity: Providing flexible deployment and adaptation policy experimentation for serverless and distributed applications spanning cloud–edge–mobile environments},
author = {Foivos Pournaropoulos and Alexandros Patras and Christos D. Antonopoulos and Nikolaos Bellas and Spyros Lalis},
doi = {10.1016/j.future.2024.03.031},
year = {2024},
date = {2024-08-01},
journal = {Future Generation Computer Systems},
volume = {157},
pages = {210–225},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Polychronis, Giorgos
Flexible and Efficient Deployment of Data Processing Pipelines on Wireless IoT Systems Journal Article
In: 2024 IEEE Sensors Applications Symposium (SAS), vol. 1, pp. 1–6, 2024.
@article{polychronisFlexibleEfficientDeployment2024,
title = {Flexible and Efficient Deployment of Data Processing Pipelines on Wireless IoT Systems},
author = {Giorgos Polychronis},
url = {https://doi.org/10.1109/sas60918.2024.10636514},
doi = {10.1109/sas60918.2024.10636514},
year = {2024},
date = {2024-07-01},
journal = {2024 IEEE Sensors Applications Symposium (SAS)},
volume = {1},
pages = {1–6},
publisher = {IEEE},
abstract = {There is an increasing number of IoT devices that do not merely act as sensor nodes but are also sufficiently powerful to perform some local data processing before data is sent upstream to more powerful servers or the cloud. In this paper, we present a framework for the flexible deployment and execution of sensing and data processing application components on embedded, industrial-strength nodes that can be connected to a wide variety of sensors. This is done using textual, declarative descriptions instead of actual code, in conjunction with a pub/sub approach for data exchange between such components. As a result, it becomes possible to build and deploy in-network data processing pipelines in efficient way, even on top of lowbandwidth wireless links. Our evaluation shows that the proposed approach can reduce the size of application logic by 95.3% and the deployment time by up to 91.1% compared to a more conventional deployment of binaries. Also, the ability to process data near or even directly on the nodes that are connected to the sensors proving the raw measurements, can reduce the number of messages by 97.7% and improve message latency by up to 44.3% vs sending all data and processing it on a server},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
There is an increasing number of IoT devices that do not merely act as sensor nodes but are also sufficiently powerful to perform some local data processing before data is sent upstream to more powerful servers or the cloud. In this paper, we present a framework for the flexible deployment and execution of sensing and data processing application components on embedded, industrial-strength nodes that can be connected to a wide variety of sensors. This is done using textual, declarative descriptions instead of actual code, in conjunction with a pub/sub approach for data exchange between such components. As a result, it becomes possible to build and deploy in-network data processing pipelines in efficient way, even on top of lowbandwidth wireless links. Our evaluation shows that the proposed approach can reduce the size of application logic by 95.3% and the deployment time by up to 91.1% compared to a more conventional deployment of binaries. Also, the ability to process data near or even directly on the nodes that are connected to the sensors proving the raw measurements, can reduce the number of messages by 97.7% and improve message latency by up to 44.3% vs sending all data and processing it on a server
Polychronis, Giorgos
Flexible and Efficient Deployment of Data Processing Pipelines on Wireless IoT Systems Journal Article
In: 2024 IEEE Sensors Applications Symposium (SAS), vol. 1, pp. 1–6, 2024.
@article{polychronis_flexible_2024,
title = {Flexible and Efficient Deployment of Data Processing Pipelines on Wireless IoT Systems},
author = {Giorgos Polychronis},
url = {https://doi.org/10.1109/sas60918.2024.10636514},
doi = {10.1109/sas60918.2024.10636514},
year = {2024},
date = {2024-07-01},
journal = {2024 IEEE Sensors Applications Symposium (SAS)},
volume = {1},
pages = {1–6},
publisher = {IEEE},
abstract = {There is an increasing number of IoT devices that do not merely act as sensor nodes but are also sufficiently powerful to perform some local data processing before data is sent upstream to more powerful servers or the cloud. In this paper, we present a framework for the flexible deployment and execution of sensing and data processing application components on embedded, industrial-strength nodes that can be connected to a wide variety of sensors. This is done using textual, declarative descriptions instead of actual code, in conjunction with a pub/sub approach for data exchange between such components. As a result, it becomes possible to build and deploy in-network data processing pipelines in efficient way, even on top of lowbandwidth wireless links. Our evaluation shows that the proposed approach can reduce the size of application logic by 95.3% and the deployment time by up to 91.1% compared to a more conventional deployment of binaries. Also, the ability to process data near or even directly on the nodes that are connected to the sensors proving the raw measurements, can reduce the number of messages by 97.7% and improve message latency by up to 44.3% vs sending all data and processing it on a server},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
There is an increasing number of IoT devices that do not merely act as sensor nodes but are also sufficiently powerful to perform some local data processing before data is sent upstream to more powerful servers or the cloud. In this paper, we present a framework for the flexible deployment and execution of sensing and data processing application components on embedded, industrial-strength nodes that can be connected to a wide variety of sensors. This is done using textual, declarative descriptions instead of actual code, in conjunction with a pub/sub approach for data exchange between such components. As a result, it becomes possible to build and deploy in-network data processing pipelines in efficient way, even on top of lowbandwidth wireless links. Our evaluation shows that the proposed approach can reduce the size of application logic by 95.3% and the deployment time by up to 91.1% compared to a more conventional deployment of binaries. Also, the ability to process data near or even directly on the nodes that are connected to the sensors proving the raw measurements, can reduce the number of messages by 97.7% and improve message latency by up to 44.3% vs sending all data and processing it on a server
Polychronis, Giorgos; Koutsoubelias, Manos; Lalis, Spyros
Should I Stay or Should I Go: A Learning Approach for Drone-based Sensing Applications Proceedings Article
In: 2024 20th International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS-IoT), pp. 339–346, IEEE, Abu Dhabi, United Arab Emirates, 2024, ISBN: 979-8-3503-6944-1.
@inproceedings{polychronisShouldStayShould2024,
title = {Should I Stay or Should I Go: A Learning Approach for Drone-based Sensing Applications},
author = {Giorgos Polychronis and Manos Koutsoubelias and Spyros Lalis},
url = {https://ieeexplore.ieee.org/document/10621483/},
doi = {10.1109/DCOSS-IoT61029.2024.00058},
isbn = {979-8-3503-6944-1},
year = {2024},
date = {2024-04-01},
urldate = {2026-02-18},
booktitle = {2024 20th International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS-IoT)},
pages = {339–346},
publisher = {IEEE},
address = {Abu Dhabi, United Arab Emirates},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Panagou, Ioanna-Maria; Bellas, Nikolaos; Moneta, Lorenzo; Sengupta, Sanjiban
Accelerating Machine Learning Inference on GPUs with SYCL Journal Article
In: Proceedings of the 12th International Workshop on OpenCL and SYCL, pp. 1–2, 2024.
@article{panagou_accelerating_2024,
title = {Accelerating Machine Learning Inference on GPUs with SYCL},
author = {Ioanna-Maria Panagou and Nikolaos Bellas and Lorenzo Moneta and Sanjiban Sengupta},
url = {https://doi.org/10.1145/3648115.3648123},
doi = {10.1145/3648115.3648123},
year = {2024},
date = {2024-04-01},
journal = {Proceedings of the 12th International Workshop on OpenCL and SYCL},
pages = {1–2},
publisher = {ACM},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Polychronis, Giorgos; Koutsoubelias, Manos; Lalis, Spyros
Should I Stay or Should I Go: A Learning Approach for Drone-based Sensing Applications Proceedings Article
In: 2024 20th International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS-IoT), pp. 339–346, IEEE, Abu Dhabi, United Arab Emirates, 2024, ISBN: 979-8-3503-6944-1.
@inproceedings{polychronis_should_2024,
title = {Should I Stay or Should I Go: A Learning Approach for Drone-based Sensing Applications},
author = {Giorgos Polychronis and Manos Koutsoubelias and Spyros Lalis},
url = {https://ieeexplore.ieee.org/document/10621483/},
doi = {10.1109/DCOSS-IoT61029.2024.00058},
isbn = {979-8-3503-6944-1},
year = {2024},
date = {2024-04-01},
urldate = {2026-02-18},
booktitle = {2024 20th International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS-IoT)},
pages = {339–346},
publisher = {IEEE},
address = {Abu Dhabi, United Arab Emirates},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
2023
Patras, Alexandros; Pournaropoulos, Foivos; Bellas, Nikolaos; Antonopoulos, Christos D.; Lalis, Spyros; Goutha, Maria; Nanos, Anastasios
A Minimal Testbed for Experimenting with Flexible Resource and Application Management in Heterogeneous Edge-Cloud Systems Proceedings Article
In: 1st International Workshop on Machine Learning for Autonomic System Operation in the Device-Edge-Cloud Continuum (MLSysOps 2023), 2023.
@inproceedings{patrasMinimalTestbedExperimenting2023,
title = {A Minimal Testbed for Experimenting with Flexible Resource and Application Management in Heterogeneous Edge-Cloud Systems},
author = {Alexandros Patras and Foivos Pournaropoulos and Nikolaos Bellas and Christos D. Antonopoulos and Spyros Lalis and Maria Goutha and Anastasios Nanos},
doi = {10.5555/3639940.3639992},
year = {2023},
date = {2023-09-01},
booktitle = {1st International Workshop on Machine Learning for Autonomic System Operation in the Device-Edge-Cloud Continuum (MLSysOps 2023)},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Pournaropoulos, Foivos; Antonopoulos, Christos D.; Lalis, Spyros
Supporting the Adaptive Deployment of Modular Applications in Cloud-Edge-Mobile Systems Proceedings Article
In: Elsevier BV, Rende, Italy, 2023.
@inproceedings{pournaropoulosSupportingAdaptiveDeployment2023,
title = {Supporting the Adaptive Deployment of Modular Applications in Cloud-Edge-Mobile Systems},
author = {Foivos Pournaropoulos and Christos D. Antonopoulos and Spyros Lalis},
doi = {10.5555/3639940.3639966},
year = {2023},
date = {2023-09-01},
publisher = {Elsevier BV},
address = {Rende, Italy},
abstract = {We introduce Fluidity, a framework enabling the flexible and adaptive deployment of modular applications in systems comprising cloud, edge, and mobile IoT nodes. Based on a declarative description of application requirements, Fluidity plans and executes an initial deployment of application components in the cloud-edge-mobile continuum. At runtime, Fluidity monitors resource availability and the position of mobile nodes, and adapts the deployment of the application accordingly, without any intervention from the application owner or system administrator. Notably, Fluidity allows applications to provide their own deployment and adaptation policies and to switch between different policies at runtime, while the application is running. We discuss the design and implementation of Fluidity in detail and provide an evaluation using a lab testbed, where the mobile node is a simulated drone. Our results show that the core mechanisms of Fluidity can adapt the application at reasonable overhead.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We introduce Fluidity, a framework enabling the flexible and adaptive deployment of modular applications in systems comprising cloud, edge, and mobile IoT nodes. Based on a declarative description of application requirements, Fluidity plans and executes an initial deployment of application components in the cloud-edge-mobile continuum. At runtime, Fluidity monitors resource availability and the position of mobile nodes, and adapts the deployment of the application accordingly, without any intervention from the application owner or system administrator. Notably, Fluidity allows applications to provide their own deployment and adaptation policies and to switch between different policies at runtime, while the application is running. We discuss the design and implementation of Fluidity in detail and provide an evaluation using a lab testbed, where the mobile node is a simulated drone. Our results show that the core mechanisms of Fluidity can adapt the application at reasonable overhead.
Pournaropoulos, Foivos; Antonopoulos, Christos D.; Lalis, Spyros
Supporting the Adaptive Deployment of Modular Applications in Cloud-Edge-Mobile Systems Proceedings Article
In: Elsevier BV, Rende, Italy, 2023.
@inproceedings{pournaropoulos_supporting_2023,
title = {Supporting the Adaptive Deployment of Modular Applications in Cloud-Edge-Mobile Systems},
author = {Foivos Pournaropoulos and Christos D. Antonopoulos and Spyros Lalis},
doi = {10.5555/3639940.3639966},
year = {2023},
date = {2023-09-01},
publisher = {Elsevier BV},
address = {Rende, Italy},
abstract = {We introduce Fluidity, a framework enabling the flexible and adaptive deployment of modular applications in systems comprising cloud, edge, and mobile IoT nodes. Based on a declarative description of application requirements, Fluidity plans and executes an initial deployment of application components in the cloud-edge-mobile continuum. At runtime, Fluidity monitors resource availability and the position of mobile nodes, and adapts the deployment of the application accordingly, without any intervention from the application owner or system administrator. Notably, Fluidity allows applications to provide their own deployment and adaptation policies and to switch between different policies at runtime, while the application is running. We discuss the design and implementation of Fluidity in detail and provide an evaluation using a lab testbed, where the mobile node is a simulated drone. Our results show that the core mechanisms of Fluidity can adapt the application at reasonable overhead.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We introduce Fluidity, a framework enabling the flexible and adaptive deployment of modular applications in systems comprising cloud, edge, and mobile IoT nodes. Based on a declarative description of application requirements, Fluidity plans and executes an initial deployment of application components in the cloud-edge-mobile continuum. At runtime, Fluidity monitors resource availability and the position of mobile nodes, and adapts the deployment of the application accordingly, without any intervention from the application owner or system administrator. Notably, Fluidity allows applications to provide their own deployment and adaptation policies and to switch between different policies at runtime, while the application is running. We discuss the design and implementation of Fluidity in detail and provide an evaluation using a lab testbed, where the mobile node is a simulated drone. Our results show that the core mechanisms of Fluidity can adapt the application at reasonable overhead.
Patras, Alexandros; Pournaropoulos, Foivos; Bellas, Nikolaos; Antonopoulos, Christos D.; Lalis, Spyros; Goutha, Maria; Nanos, Anastasios
A Minimal Testbed for Experimenting with Flexible Resource and Application Management in Heterogeneous Edge-Cloud Systems Proceedings Article
In: 1st International Workshop on Machine Learning for Autonomic System Operation in the Device-Edge-Cloud Continuum (MLSysOps 2023), 2023.
@inproceedings{patras_minimal_2023,
title = {A Minimal Testbed for Experimenting with Flexible Resource and Application Management in Heterogeneous Edge-Cloud Systems},
author = {Alexandros Patras and Foivos Pournaropoulos and Nikolaos Bellas and Christos D. Antonopoulos and Spyros Lalis and Maria Goutha and Anastasios Nanos},
doi = {10.5555/3639940.3639992},
year = {2023},
date = {2023-09-01},
booktitle = {1st International Workshop on Machine Learning for Autonomic System Operation in the Device-Edge-Cloud Continuum (MLSysOps 2023)},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Polychronis, Giorgos; Lalis, Spyros
Flexible Computation Offloading at the Edge for Autonomous Drones with Uncertain Flight Times Book
IEEE, 2023.
@book{polychronis_flexible_2023,
title = {Flexible Computation Offloading at the Edge for Autonomous Drones with Uncertain Flight Times},
author = {Giorgos Polychronis and Spyros Lalis},
url = {https://doi.org/10.1109/dcoss-iot58021.2023.00043},
doi = {10.1109/dcoss-iot58021.2023.00043},
year = {2023},
date = {2023-06-01},
volume = {2014},
publisher = {IEEE},
keywords = {},
pubstate = {published},
tppubtype = {book}
}
Gkeka, Maria Rafaela; Patras, Alexandros; Tavoularis, Nikolaos; Piperakis, Stylianos; Hourdakis, Emmanouil; Trahanias, Panos; Antonopoulos, Christos D.; Lalis, Spyros; Bellas, Nikolaos
Reconfigurable System-on-Chip Architectures for Robust Visual SLAM on Humanoid Robots Proceedings Article
In: ACM Transactions on Embedded Computing Systems, pp. 1–29, 2023.
@inproceedings{gkekaReconfigurableSystemonChipArchitectures2023,
title = {Reconfigurable System-on-Chip Architectures for Robust Visual SLAM on Humanoid Robots},
author = {Maria Rafaela Gkeka and Alexandros Patras and Nikolaos Tavoularis and Stylianos Piperakis and Emmanouil Hourdakis and Panos Trahanias and Christos D. Antonopoulos and Spyros Lalis and Nikolaos Bellas},
doi = {10.1145/3570210},
year = {2023},
date = {2023-01-01},
booktitle = {ACM Transactions on Embedded Computing Systems},
pages = {1–29},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Gkeka, Maria Rafaela; Patras, Alexandros; Tavoularis, Nikolaos; Piperakis, Stylianos; Hourdakis, Emmanouil; Trahanias, Panos; Antonopoulos, Christos D.; Lalis, Spyros; Bellas, Nikolaos
Reconfigurable System-on-Chip Architectures for Robust Visual SLAM on Humanoid Robots Proceedings Article
In: ACM Transactions on Embedded Computing Systems, pp. 1–29, 2023.
@inproceedings{gkeka_reconfigurable_2023,
title = {Reconfigurable System-on-Chip Architectures for Robust Visual SLAM on Humanoid Robots},
author = {Maria Rafaela Gkeka and Alexandros Patras and Nikolaos Tavoularis and Stylianos Piperakis and Emmanouil Hourdakis and Panos Trahanias and Christos D. Antonopoulos and Spyros Lalis and Nikolaos Bellas},
doi = {10.1145/3570210},
year = {2023},
date = {2023-01-01},
booktitle = {ACM Transactions on Embedded Computing Systems},
pages = {1–29},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
2022
Polychronis, Giorgos; Lalis, Spyros
Joint Edge Resource Allocation and Path Planning for Drones with Energy Constraints Book Section
In: pp. 378–399, 2022.
@incollection{polychronisJointEdgeResource2022,
title = {Joint Edge Resource Allocation and Path Planning for Drones with Energy Constraints},
author = {Giorgos Polychronis and Spyros Lalis},
url = {https://doi.org/10.1007/978-3-031-34776-4_20},
doi = {10.1007/978-3-031-34776-4_20},
year = {2022},
date = {2022-11-01},
pages = {378–399},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Polychronis, Giorgos; Lalis, Spyros
Joint Edge Resource Allocation and Path Planning for Drones with Energy Constraints Book Section
In: pp. 378–399, 2022.
@incollection{polychronis_joint_2022,
title = {Joint Edge Resource Allocation and Path Planning for Drones with Energy Constraints},
author = {Giorgos Polychronis and Spyros Lalis},
url = {https://doi.org/10.1007/978-3-031-34776-4_20},
doi = {10.1007/978-3-031-34776-4_20},
year = {2022},
date = {2022-11-01},
pages = {378–399},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Panagou, Ioanna-Maria; Gkeka, Maria Rafaela; Patras, Alexandros; Lalis, Spyros; Antonopoulos, Christos D.; Bellas, Nikolaos
FPGA Roofline modeling and its Application to Visual SLAM Journal Article
In: 2022 32nd International Conference on Field-Programmable Logic and Applications (FPL), pp. 130–135, 2022.
@article{panagouFPGARooflineModeling2022,
title = {FPGA Roofline modeling and its Application to Visual SLAM},
author = {Ioanna-Maria Panagou and Maria Rafaela Gkeka and Alexandros Patras and Spyros Lalis and Christos D. Antonopoulos and Nikolaos Bellas},
url = {https://doi.org/10.1109/fpl57034.2022.00030},
doi = {10.1109/fpl57034.2022.00030},
year = {2022},
date = {2022-08-01},
journal = {2022 32nd International Conference on Field-Programmable Logic and Applications (FPL)},
pages = {130–135},
address = {Belfast, United Kingdom},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Panagou, Ioanna-Maria; Gkeka, Maria Rafaela; Patras, Alexandros; Lalis, Spyros; Antonopoulos, Christos D.; Bellas, Nikolaos
FPGA Roofline modeling and its Application to Visual SLAM Journal Article
In: 2022 32nd International Conference on Field-Programmable Logic and Applications (FPL), pp. 130–135, 2022.
@article{panagou_fpga_2022,
title = {FPGA Roofline modeling and its Application to Visual SLAM},
author = {Ioanna-Maria Panagou and Maria Rafaela Gkeka and Alexandros Patras and Spyros Lalis and Christos D. Antonopoulos and Nikolaos Bellas},
url = {https://doi.org/10.1109/fpl57034.2022.00030},
doi = {10.1109/fpl57034.2022.00030},
year = {2022},
date = {2022-08-01},
journal = {2022 32nd International Conference on Field-Programmable Logic and Applications (FPL)},
pages = {130–135},
address = {Belfast, United Kingdom},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Polychronis, Giorgos; Lalis, Spyros
Planning Computation Offloading on Shared Edge Infrastructure for Multiple Drones Journal Article
In: 2022 IEEE 42nd International Conference on Distributed Computing Systems Workshops (ICDCSW), vol. 2014, pp. 302–307, 2022.
@article{polychronisPlanningComputationOffloading2022,
title = {Planning Computation Offloading on Shared Edge Infrastructure for Multiple Drones},
author = {Giorgos Polychronis and Spyros Lalis},
url = {https://doi.org/10.1109/icdcsw56584.2022.00063},
doi = {10.1109/icdcsw56584.2022.00063},
year = {2022},
date = {2022-07-01},
journal = {2022 IEEE 42nd International Conference on Distributed Computing Systems Workshops (ICDCSW)},
volume = {2014},
pages = {302–307},
address = {Bologna, Italy},
abstract = {This work presents an offline heuristic for the fair allocation of shared edge server resources to support computation offloading for drones so as to evenly reduce the respective mission times.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This work presents an offline heuristic for the fair allocation of shared edge server resources to support computation offloading for drones so as to evenly reduce the respective mission times.
Polychronis, Giorgos; Lalis, Spyros
Planning Computation Offloading on Shared Edge Infrastructure for Multiple Drones Journal Article
In: 2022 IEEE 42nd International Conference on Distributed Computing Systems Workshops (ICDCSW), vol. 2014, pp. 302–307, 2022.
@article{polychronis_planning_2022,
title = {Planning Computation Offloading on Shared Edge Infrastructure for Multiple Drones},
author = {Giorgos Polychronis and Spyros Lalis},
url = {https://doi.org/10.1109/icdcsw56584.2022.00063},
doi = {10.1109/icdcsw56584.2022.00063},
year = {2022},
date = {2022-07-01},
journal = {2022 IEEE 42nd International Conference on Distributed Computing Systems Workshops (ICDCSW)},
volume = {2014},
pages = {302–307},
address = {Bologna, Italy},
abstract = {This work presents an offline heuristic for the fair allocation of shared edge server resources to support computation offloading for drones so as to evenly reduce the respective mission times.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This work presents an offline heuristic for the fair allocation of shared edge server resources to support computation offloading for drones so as to evenly reduce the respective mission times.
Kalogirou, Christos; Antonopoulos, Christos D.; Lalis, Spyros; Bellas, Nikolaos
Dynamic Management of CPU Resources Towards Energy Efficient and Profitable Datacentre Operation Book Section
In: pp. 131–151, Virtual Conference, 2022.
@incollection{kalogirouDynamicManagementCPU2022,
title = {Dynamic Management of CPU Resources Towards Energy Efficient and Profitable Datacentre Operation},
author = {Christos Kalogirou and Christos D. Antonopoulos and Spyros Lalis and Nikolaos Bellas},
url = {https://doi.org/10.1007/978-3-031-22698-4_7},
doi = {10.1007/978-3-031-22698-4_7},
year = {2022},
date = {2022-06-01},
pages = {131–151},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Grigoropoulos, Nasos; Lalis, Spyros
Fractus: Orchestration of Distributed Applications in the Drone-Edge-Cloud Continuum Book
Virtual Conference, 2022.
@book{grigoropoulos_fractus_2022,
title = {Fractus: Orchestration of Distributed Applications in the Drone-Edge-Cloud Continuum},
author = {Nasos Grigoropoulos and Spyros Lalis},
year = {2022},
date = {2022-06-01},
address = {Virtual Conference},
abstract = {This work presents an orchestration framework for the automated deployment of component-based applications in the drone-edge-cloud continuum, which provides users with abstractions for describing the application’s placement and communication requirements, allocates resources in a mission-aware fashion by considering the drone operation area, establishes and maintains connectivity between components by transparently leveraging different networking capabilities, and tackles safety and privacy issues via policy-based access to mobility and sensor resources.},
keywords = {},
pubstate = {published},
tppubtype = {book}
}
This work presents an orchestration framework for the automated deployment of component-based applications in the drone-edge-cloud continuum, which provides users with abstractions for describing the application’s placement and communication requirements, allocates resources in a mission-aware fashion by considering the drone operation area, establishes and maintains connectivity between components by transparently leveraging different networking capabilities, and tackles safety and privacy issues via policy-based access to mobility and sensor resources.
Kalogirou, Christos; Antonopoulos, Christos D.; Lalis, Spyros; Bellas, Nikolaos
Dynamic Management of CPU Resources Towards Energy Efficient and Profitable Datacentre Operation Book Section
In: pp. 131–151, Virtual Conference, 2022.
@incollection{kalogirou_dynamic_2022,
title = {Dynamic Management of CPU Resources Towards Energy Efficient and Profitable Datacentre Operation},
author = {Christos Kalogirou and Christos D. Antonopoulos and Spyros Lalis and Nikolaos Bellas},
url = {https://doi.org/10.1007/978-3-031-22698-4_7},
doi = {10.1007/978-3-031-22698-4_7},
year = {2022},
date = {2022-06-01},
pages = {131–151},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Gkeka, Maria Rafaela; Patras, Alexandros; Tavoularis, Nikolaos; Piperakis, Stylianos; Hourdakis, Emmanouil; Trahanias, Panos; Antonopoulos, Christos D.; Lalis, Spyros; Bellas, Nikolaos
FPGA Accelerators for Robust Visual SLAM on Humanoid Robots Journal Article
In: Proceedings of the 2022 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays, pp. 51–51, 2022.
@article{gkekaFPGAAcceleratorsRobust2022,
title = {FPGA Accelerators for Robust Visual SLAM on Humanoid Robots},
author = {Maria Rafaela Gkeka and Alexandros Patras and Nikolaos Tavoularis and Stylianos Piperakis and Emmanouil Hourdakis and Panos Trahanias and Christos D. Antonopoulos and Spyros Lalis and Nikolaos Bellas},
url = {https://doi.org/10.1145/3490422.3502331},
doi = {10.1145/3490422.3502331},
year = {2022},
date = {2022-02-01},
journal = {Proceedings of the 2022 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays},
pages = {51–51},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Gkeka, Maria Rafaela; Patras, Alexandros; Tavoularis, Nikolaos; Piperakis, Stylianos; Hourdakis, Emmanouil; Trahanias, Panos; Antonopoulos, Christos D.; Lalis, Spyros; Bellas, Nikolaos
FPGA Accelerators for Robust Visual SLAM on Humanoid Robots Journal Article
In: Proceedings of the 2022 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays, pp. 51–51, 2022.
@article{gkeka_fpga_2022,
title = {FPGA Accelerators for Robust Visual SLAM on Humanoid Robots},
author = {Maria Rafaela Gkeka and Alexandros Patras and Nikolaos Tavoularis and Stylianos Piperakis and Emmanouil Hourdakis and Panos Trahanias and Christos D. Antonopoulos and Spyros Lalis and Nikolaos Bellas},
url = {https://doi.org/10.1145/3490422.3502331},
doi = {10.1145/3490422.3502331},
year = {2022},
date = {2022-02-01},
journal = {Proceedings of the 2022 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays},
pages = {51–51},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
2021
Koutsoubelias, Manos; Grigoropoulos, Nasos; Polychronis, Giorgos; Badakis, Giannis; Lalis, Spyros
System Architecture for Autonomous Drone-based Remote Sensing Book Section
In: pp. 220–242, Virtual Conference, 2021.
@incollection{koutsoubeliasSystemArchitectureAutonomous2021,
title = {System Architecture for Autonomous Drone-based Remote Sensing},
author = {Manos Koutsoubelias and Nasos Grigoropoulos and Giorgos Polychronis and Giannis Badakis and Spyros Lalis},
url = {https://doi.org/10.1007/978-3-030-94822-1_13},
doi = {10.1007/978-3-030-94822-1_13},
year = {2021},
date = {2021-11-01},
pages = {220–242},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Koutsoubelias, Manos; Grigoropoulos, Nasos; Polychronis, Giorgos; Badakis, Giannis; Lalis, Spyros
System Architecture for Autonomous Drone-based Remote Sensing Book Section
In: pp. 220–242, Virtual Conference, 2021.
@incollection{koutsoubelias_system_2021,
title = {System Architecture for Autonomous Drone-based Remote Sensing},
author = {Manos Koutsoubelias and Nasos Grigoropoulos and Giorgos Polychronis and Giannis Badakis and Spyros Lalis},
url = {https://doi.org/10.1007/978-3-030-94822-1_13},
doi = {10.1007/978-3-030-94822-1_13},
year = {2021},
date = {2021-11-01},
pages = {220–242},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Bellas, Nikolaos; Antonopoulos, Christos D.; Lalis, Spyros; Gkeka, Maria Rafaela; Patras, Alexandros; Keramidas, Georgios; Stamoulis, Iakovos; Tavoularis, Nikolaos; Piperakis, Stylianos; Hourdakis, Emmanouil; Trahanias, Panos; Zikas, Paul; Papagiannakis, George; Kartsonaki, Ioanna
Architectures for SLAM and Augmented Reality Computing Journal Article
In: 2021 31st International Conference on Field-Programmable Logic and Applications (FPL), pp. 378–379, 2021.
@article{bellasArchitecturesSLAMAugmented2021,
title = {Architectures for SLAM and Augmented Reality Computing},
author = {Nikolaos Bellas and Christos D. Antonopoulos and Spyros Lalis and Maria Rafaela Gkeka and Alexandros Patras and Georgios Keramidas and Iakovos Stamoulis and Nikolaos Tavoularis and Stylianos Piperakis and Emmanouil Hourdakis and Panos Trahanias and Paul Zikas and George Papagiannakis and Ioanna Kartsonaki},
url = {https://doi.org/10.1109/fpl53798.2021.00073},
doi = {10.1109/fpl53798.2021.00073},
year = {2021},
date = {2021-09-01},
journal = {2021 31st International Conference on Field-Programmable Logic and Applications (FPL)},
pages = {378–379},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Polychronis, Giorgos; Lalis, Spyros
Safe Optimistic Path Planning for Autonomous Drones under Dynamic Energy Costs Journal Article
In: 2021 IEEE International Intelligent Transportation Systems Conference (ITSC), pp. 1927–1933, 2021.
@article{polychronis_safe_2021,
title = {Safe Optimistic Path Planning for Autonomous Drones under Dynamic Energy Costs},
author = {Giorgos Polychronis and Spyros Lalis},
url = {https://doi.org/10.1109/itsc48978.2021.9564911},
doi = {10.1109/itsc48978.2021.9564911},
year = {2021},
date = {2021-09-01},
journal = {2021 IEEE International Intelligent Transportation Systems Conference (ITSC)},
pages = {1927–1933},
address = {Indianapolis, United States},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Bellas, Nikolaos; Antonopoulos, Christos D.; Lalis, Spyros; Gkeka, Maria Rafaela; Patras, Alexandros; Keramidas, Georgios; Stamoulis, Iakovos; Tavoularis, Nikolaos; Piperakis, Stylianos; Hourdakis, Emmanouil; Trahanias, Panos; Zikas, Paul; Papagiannakis, George; Kartsonaki, Ioanna
Architectures for SLAM and Augmented Reality Computing Journal Article
In: 2021 31st International Conference on Field-Programmable Logic and Applications (FPL), pp. 378–379, 2021.
@article{bellas_architectures_2021,
title = {Architectures for SLAM and Augmented Reality Computing},
author = {Nikolaos Bellas and Christos D. Antonopoulos and Spyros Lalis and Maria Rafaela Gkeka and Alexandros Patras and Georgios Keramidas and Iakovos Stamoulis and Nikolaos Tavoularis and Stylianos Piperakis and Emmanouil Hourdakis and Panos Trahanias and Paul Zikas and George Papagiannakis and Ioanna Kartsonaki},
url = {https://doi.org/10.1109/fpl53798.2021.00073},
doi = {10.1109/fpl53798.2021.00073},
year = {2021},
date = {2021-09-01},
journal = {2021 31st International Conference on Field-Programmable Logic and Applications (FPL)},
pages = {378–379},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Badakis, Giannis; Koutsoubelias, Manos; Lalis, Spyros
Robust Precision Landing for Autonomous Drones Combining Vision-based and Infrared Sensors Journal Article
In: 2021 IEEE Sensors Applications Symposium (SAS), pp. 1–6, 2021.
@article{badakisRobustPrecisionLanding2021,
title = {Robust Precision Landing for Autonomous Drones Combining Vision-based and Infrared Sensors},
author = {Giannis Badakis and Manos Koutsoubelias and Spyros Lalis},
url = {https://doi.org/10.1109/sas51076.2021.9530091},
doi = {10.1109/sas51076.2021.9530091},
year = {2021},
date = {2021-08-01},
journal = {2021 IEEE Sensors Applications Symposium (SAS)},
pages = {1–6},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Badakis, Giannis; Koutsoubelias, Manos; Lalis, Spyros
Robust Precision Landing for Autonomous Drones Combining Vision-based and Infrared Sensors Journal Article
In: 2021 IEEE Sensors Applications Symposium (SAS), pp. 1–6, 2021.
@article{badakis_robust_2021,
title = {Robust Precision Landing for Autonomous Drones Combining Vision-based and Infrared Sensors},
author = {Giannis Badakis and Manos Koutsoubelias and Spyros Lalis},
url = {https://doi.org/10.1109/sas51076.2021.9530091},
doi = {10.1109/sas51076.2021.9530091},
year = {2021},
date = {2021-08-01},
journal = {2021 IEEE Sensors Applications Symposium (SAS)},
pages = {1–6},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Grigoropoulos, Nasos; Lalis, Spyros
Fractus: Orchestration of Distributed Applications in the Drone-Edge-Cloud Continuum Journal Article
In: 2022 IEEE 46th Annual Computers, Software, and Applications Conference (COMPSAC), pp. 838–848, 2021.
@article{grigoropoulosFractusOrchestrationDistributed2021,
title = {Fractus: Orchestration of Distributed Applications in the Drone-Edge-Cloud Continuum},
author = {Nasos Grigoropoulos and Spyros Lalis},
url = {https://doi.org/10.1109/compsac54236.2022.00134},
doi = {10.1109/compsac54236.2022.00134},
year = {2021},
date = {2021-06-01},
journal = {2022 IEEE 46th Annual Computers, Software, and Applications Conference (COMPSAC)},
pages = {838–848},
address = {Virtual Conference},
abstract = {This work presents an orchestration framework for the automated deployment of component-based applications in the drone-edge-cloud continuum, which provides users with abstractions for describing the application’s placement and communication requirements, allocates resources in a mission-aware fashion by considering the drone operation area, establishes and maintains connectivity between components by transparently leveraging different networking capabilities, and tackles safety and privacy issues via policy-based access to mobility and sensor resources.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This work presents an orchestration framework for the automated deployment of component-based applications in the drone-edge-cloud continuum, which provides users with abstractions for describing the application’s placement and communication requirements, allocates resources in a mission-aware fashion by considering the drone operation area, establishes and maintains connectivity between components by transparently leveraging different networking capabilities, and tackles safety and privacy issues via policy-based access to mobility and sensor resources.
Grigoropoulos, Nasos; Lalis, Spyros
Fractus: Orchestration of Distributed Applications in the Drone-Edge-Cloud Continuum Journal Article
In: 2022 IEEE 46th Annual Computers, Software, and Applications Conference (COMPSAC), pp. 838–848, 2021.
@article{grigoropoulos_fractus_2021,
title = {Fractus: Orchestration of Distributed Applications in the Drone-Edge-Cloud Continuum},
author = {Nasos Grigoropoulos and Spyros Lalis},
url = {https://doi.org/10.1109/compsac54236.2022.00134},
doi = {10.1109/compsac54236.2022.00134},
year = {2021},
date = {2021-06-01},
journal = {2022 IEEE 46th Annual Computers, Software, and Applications Conference (COMPSAC)},
pages = {838–848},
address = {Virtual Conference},
abstract = {This work presents an orchestration framework for the automated deployment of component-based applications in the drone-edge-cloud continuum, which provides users with abstractions for describing the application’s placement and communication requirements, allocates resources in a mission-aware fashion by considering the drone operation area, establishes and maintains connectivity between components by transparently leveraging different networking capabilities, and tackles safety and privacy issues via policy-based access to mobility and sensor resources.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This work presents an orchestration framework for the automated deployment of component-based applications in the drone-edge-cloud continuum, which provides users with abstractions for describing the application’s placement and communication requirements, allocates resources in a mission-aware fashion by considering the drone operation area, establishes and maintains connectivity between components by transparently leveraging different networking capabilities, and tackles safety and privacy issues via policy-based access to mobility and sensor resources.
Maroudas, Emmanouil; Antonopoulos, Christos D.; Bellas, Nikolaos; Lalis, Spyros
Exploring the Potential of Context-Aware Dynamic CPU Undervolting Journal Article
In: Proceedings of the 18th ACM International Conference on Computing Frontiers, pp. 73–82, 2021.
@article{maroudasExploringPotentialContextAware2021,
title = {Exploring the Potential of Context-Aware Dynamic CPU Undervolting},
author = {Emmanouil Maroudas and Christos D. Antonopoulos and Nikolaos Bellas and Spyros Lalis},
url = {https://doi.org/10.1145/3457388.3458658},
doi = {10.1145/3457388.3458658},
year = {2021},
date = {2021-05-01},
journal = {Proceedings of the 18th ACM International Conference on Computing Frontiers},
pages = {73–82},
address = {Virtual Conference},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Kasidakis, Theodoros; Polychronis, Giorgos; Koutsoubelias, Manos; Lalis, Spyros
Reducing the Mission Time of Drone Applications through Location-Aware Edge Computing Journal Article
In: 2021 IEEE 5th International Conference on Fog and Edge Computing (ICFEC), pp. 45–52, 2021.
@article{kasidakis_reducing_2021,
title = {Reducing the Mission Time of Drone Applications through Location-Aware Edge Computing},
author = {Theodoros Kasidakis and Giorgos Polychronis and Manos Koutsoubelias and Spyros Lalis},
url = {https://doi.org/10.1109/icfec51620.2021.00014},
doi = {10.1109/icfec51620.2021.00014},
year = {2021},
date = {2021-05-01},
journal = {2021 IEEE 5th International Conference on Fog and Edge Computing (ICFEC)},
pages = {45–52},
address = {Virtual Conference},
abstract = {In data-driven applications, which go beyond simple data collection, drones may need to process sensor measurements at certain locations, during the mission. However, the onboard computing platforms typically have strong resource limitations, which may lead to significant delays and long mission times. To address this problem, we explore the potential of offloading heavyweight computations from the drone to nearby edge computing infrastructure. We discuss a concrete implementation for a service-oriented application software stack, which takes offloading decisions based on the expected service invocation time and the locations of the servers expected to be available in the mission area. We evaluate our implementation using an experimental setup that combines a hardware-in-the-loop and software-in-the-loop configuration. Our results show that the proposed approach can reduce the total mission time significantly, by up to 48% vs local-only processing, and by 10% vs more naive opportunistic offloading, depending on the mission scenario.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In data-driven applications, which go beyond simple data collection, drones may need to process sensor measurements at certain locations, during the mission. However, the onboard computing platforms typically have strong resource limitations, which may lead to significant delays and long mission times. To address this problem, we explore the potential of offloading heavyweight computations from the drone to nearby edge computing infrastructure. We discuss a concrete implementation for a service-oriented application software stack, which takes offloading decisions based on the expected service invocation time and the locations of the servers expected to be available in the mission area. We evaluate our implementation using an experimental setup that combines a hardware-in-the-loop and software-in-the-loop configuration. Our results show that the proposed approach can reduce the total mission time significantly, by up to 48% vs local-only processing, and by 10% vs more naive opportunistic offloading, depending on the mission scenario.