Ubiquitous Distributed Deep Reinforcement Learning at the Edge: Analyzing Byzantine Agents in Discrete Action Spaces
Open accessAbstract
The integration of edge computing in next-generation mobile networks is bringing low-latency and high-bandwidth ubiquitous connectivity to a myriad of cyber-physical systems. This will further boost the increasing intelligence that is being embedded at the edge in various types of autonomous systems, where collaborative machine learning has the potential to play a significant role. This paper discusses some of the challenges in multi-agent distributed deep reinforcement learning that can occur in the presence of byzantine or malfunctioning agents. As the simulation-to-reality gap gets bridged, the probability of malfunctions or errors must be taken into account. We show how wrong discrete actions can significantly affect the collaborative learning effort. In particular, we analyze the effect of having a fraction of agents that might perform the wrong action with a given probability. We study the ability of the system to converge towards a common working policy through the collaborative learning process based on the number of experiences from each of the agents to be aggregated for each policy update, together with the fraction of wrong actions from agents experiencing malfunctions. Our experiments are carried out in a simulation environment using the Atari testbed for the discrete action spaces, and advantage actor-critic (A2C) for the distributed multi-agent training.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Environmental perturbations affect different agents learning in parallel differently.
“Also, in some of our recent works [8], [73], we have considered environmental perturbations that affect differently different agents that are learning in parallel.”
From Sim-to-Real Transfer in Deep Reinforcement Learning for Robotics: a Survey (2020) · cites this paper as [73]
The Simulations and Results section describes 16 agents learning in parallel and tests how wrong actions by one or more agents affect collaborative learning. But the Related Works section explicitly distinguishes this paper’s focus on agents’ actions from perturbations of the environment; it does not show environmental perturbations affecting different agents differently. Page numbers are unavailable in our copy. (LLM, conf=0.84)
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Asynchronous Methods for Deep Reinforcement Learning
Mnih, V., Badia, A. P., Mirza, M., Graves, A., Lillicrap, T. P., Harley, T., Silver, D., & Kavukcuoglu, K. (2016) · arXiv · doi:10.48550/arxiv.1602.01783 · Cited by 7,132 · see citing works
Deep reinforcement learning for robotic manipulation with asynchronous off-policy updates
Gu, S., Holly, E., Lillicrap, T., & Levine, S. (2017) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2017.7989385 · Cited by 1,485 · see citing works
Nguyen, T. T., Nguyen, N. D., & Nahavandi, S. (2020) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2020.2977374 · Cited by 1,306 · see citing works
Towards Optimally Decentralized Multi-Robot Collision Avoidance via Deep Reinforcement Learning
Long, P., Fanl, T., Liao, X., Liu, W., Zhang, H., & Pan, J. (2018) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2018.8461113 · Cited by 558 · see citing works
Smart Resource Allocation for Mobile Edge Computing: A Deep Reinforcement Learning Approach
Wang, J., Zhao, L., Liu, J., & Kato, N. (2019) · IEEE Transactions on Emerging Topics in Computing · doi:10.1109/tetc.2019.2902661 · Cited by 450 · see citing works
Deep Reinforcement Learning for Vehicular Edge Computing
Ning, Z., Dong, P., Wang, X., Rodrigues, J. J. P. C., & Xia, F. (2019) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3317572 · Cited by 296 · see citing works
Vulnerability of Deep Reinforcement Learning to Policy Induction Attacks
Behzadan, V., & Munir, A. (2017) · Lecture notes in computer science · doi:10.1007/978-3-319-62416-7_19 · Cited by 247 · see citing works
Reinforcement Learning with Perturbed Rewards
Wang, J., Liu, Y., & Li, B. (2020) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v34i04.6086 · Cited by 111 · see citing works
Meta reinforcement learning for sim-to-real domain adaptation
Arndt, K., Hazara, M., Ghadirzadeh, A., & Kyrki, V. (2020) · 2020 IEEE International Conference on Robotics and Automation (ICRA) · doi:10.1109/icra40945.2020.9196540 · Cited by 97 · see citing works
Enhancing Autonomy with Blockchain and Multi-Access Edge Computing in Distributed Robotic Systems
Queralta, J. P., Li, Q., Zou, Z., & Westerlund, T. (2020) · International Conference on Fog and Mobile Edge Computing (FMEC) · doi:10.1109/fmec49853.2020.9144809 · Cited by 46 · see citing works
Self-citation
Flagged first, then by citations
Research works printed but not linked: 7
- J. Matas et al. Sim-to-real reinforcement learning for deformable object manipulation. arXiv, 2018.
- R. Raileanu et al. Modeling others using oneself in multi-agent reinforcement learning. arXiv preprint arXiv:1802.09640, 2018. · arXiv
- R. Conde et al. Time-varying formation controllers for unmanned aerial vehicles using deep reinforcement learning. arXiv, 2017.
- A. Gleave et al. Adversarial policies: Attacking deep reinforcement learning. arXiv preprint arXiv:1905.10615, 2019. · arXiv
- B. Balaji et al. Deepracer: Educational autonomous racing platform for experimentation with sim2real reinforcement learning. arXiv:1911.01562, 2019. · arXiv
- Y. Yu et al. Multi-agent deep reinforcement learning multiple access for heterogeneous wireless networks with imperfect channels. arXiv preprint arXiv:2003.11210, 2020. · arXiv
- I. Ilahi et al. Challenges and countermeasures for adversarial attacks on deep reinforcement learning. arXiv preprint arXiv:2001.09684, 2020. · arXiv
Cited by
Sorted by citations
Sim-to-Real Transfer in Deep Reinforcement Learning for Robotics: a Survey
Zhao, W., Queralta, J. P., & Westerlund, T. (2020) · 2020 IEEE Symposium Series on Computational Intelligence (SSCI) · doi:10.1109/ssci47803.2020.9308468 · Cited by 941 · see citing works
Self-citationFederated Learning in Robotic and Autonomous Systems
Xianjia, Y., Queralta, J. P., Heikkonen, J., & Westerlund, T. (2021) · Procedia Computer Science · doi:10.1016/j.procs.2021.07.041 · Cited by 79 · see citing works
Self-citation
- Authors
- Zhao, W., Queralta, J. P., Qingqing, L., & Westerlund, T.
- Affiliations
- University of Turku (FI)
- Corresponding author
- Zhao, Wenshuai
- Venue
- Procedia Computer Science
- Publisher
- Elsevier BV
- Published
- 2020
- Volume
- 177
- Pages
- 324-329
- ISSN
- 1877-0509 (ISSN-L)
- OpenAlex
- W3098752252
- Cited by
- 9
- References
- 10
- Citations by year
- 2026: 1 · 2025: 1 · 2024: 1 · 2022: 2 · 2021: 2 · 2020: 1 · 2012: 1
- FWCI
- 0.48
- Type
- conference paper
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- Locations
- Procedia Computer Science (publishedVersion, cc-by-nc-nd)
- arXiv (Cornell University) (submittedVersion)
- UTUPub (University of Turku) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
- Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
- Keywords
- distributed deep reinforcement learning, multi-agent reinforcement learning, Byzantine agents, discrete action space, edge computing, cooperative learning, policy convergence, Advantage Actor-Critic, Atari, sim-to-real gap
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
