Autonomous Bus Fleet Control Using Multiagent Reinforcement Learning

Wang, Sung Jung; Chang, S. K. Jason

doi:10.1155/2021/6654254

Cited by 5 publications

(5 citation statements)

References 24 publications

(24 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These algorithms learn the optimal policy by iteratively updating their estimates of the value function or the policy. The value function represents the expected cumulative reward under a given policy, while the policy represents the mapping from states to actions [13,14]. The Q-learning algorithm starts with the Q-values set to zero.…”

Section: Introductionmentioning

confidence: 99%

Implementing Deep Reinforcement Learning in Autonomous Control Systems

Noureldin Ragheb,

Mervat M. A. Mahmoud

2024

ARASET

View full text Add to dashboard Cite

Developing a safe and reliable autonomous vehicle has been a significant focus in recent years. Supervised learning methods require large amounts of labelled data for training, making it expensive. The performance of these agents is limited to the data provided in training and the inability to generalize performance in different environments. In addition, some driving situations, such as near-accident scenarios, are difficult to cover in the training data. As a result, the autonomous driving agent may behave unexpectedly in safety-critical situations, making it unreliable for safe transportation. Reinforcement learning is a potential solution for these issues. This research paper explores the potential of applying deep reinforcement learning techniques to autonomous driving, with a spotlight on comparing two popular deep reinforcement learning algorithms: Deep Q-Network (DQN) and Deep Deterministic Policy Gradient (DDPG). The study uses the CARLA simulator, which provides a realistic environment and conditions for testing autonomous driving algorithms. The study finds that DDPG outperforms DQN regarding average reward, but DQN performs better regarding collision rate.

show abstract

Section: Introductionmentioning

confidence: 99%

Implementing Deep Reinforcement Learning in Autonomous Control Systems

Noureldin Ragheb,

Mervat M. A. Mahmoud

2024

ARASET

View full text Add to dashboard Cite

show abstract

“…Recent developments in single-agent Reinforcement Learning (RL) as well as the appearance of cutting-edge neural architectures have fueled fast progress in Multi-Agent Reinforcement Learning (MARL). MARL has achieved great success in multi-player and video games [1,2], and showed promises for a variety of practical applications, including motion planning for autonomous vehicles [3,4] and distributed multi-robot control [5,6]. However, achieving decentralized and scalable cooperation among agents remains a key challenge in MARL.…”

Section: Introductionmentioning

confidence: 99%

Full communication memory networks for team-level cooperation learning

Wang

Sartoretti

2023

Auton Agent Multi-Agent Syst

View full text Add to dashboard Cite

Communication in multi-agent systems is a key driver of team-level cooperation, for instance allowing individual agents to augment their knowledge about the world in partially-observable environments. In this paper, we propose two reinforcement learning-based multi-agent models, namely FCMNet and FCMTran. The two models both allow agents to simultaneously learn a differentiable communication mechanism that connects all agents as well as a common, cooperative policy conditioned upon received information. FCMNet utilizes multiple directional recurrent neural networks to sequentially transmit and encode the current observation-based messages sent by every other agent at each timestep. FCMTran further relies on the encoder of a modified transformer to simultaneously aggregate multiple self-generated messages sent by all agents at the previous timestep into a single message that is used in the current timestep. Results from evaluating our models on a challenging set of StarCraft II micromanagement tasks with shared rewards show that FCMNet and FCMTran both outperform recent communication-based methods and value decomposition methods in almost all tested StarCraft II micromanagement tasks. We further improve the performance of our models by combining them with value decomposition techniques; there, in particular, we show that FCMTran with value decomposition significantly pushes the state-of-the-art on one of the hardest benchmark tasks without any task-specific tuning. We also investigate the robustness of FCMNet under communication disturbances (i.e., binarized messages, random message loss, and random communication order) in an asymmetric collaborative pathfinding task with individual rewards, demonstrating FMCNet's potential applicability in real-world robotic tasks.

show abstract

Section: Introductionmentioning

confidence: 99%

Full Communication Memory Networks for Team-Level Cooperation Learning

Wang

Sartoretti

2023

Preprint

View full text Add to dashboard Cite

Communication in multi-agent systems is a key driver of team-level cooperation, for instance allowing individual agents to augment their knowledge about the world in partially-observable environments. In this paper, we propose two reinforcement learning-based multi-agent models, namely FCMNet and FCMTran. The two models both allow agents to simultaneously learn a differentiable communication mechanism that connects all agents as well as a common, cooperative policy conditioned upon received information. FCMNet utilizes multiple directional recurrent neural networks to sequentially transmit and encode the current observation-based messages sent by every other agent at each timestep. FCMTran further relies on the encoder of a modified transformer to simultaneously aggregate multiple self-generated messages sent by all agents at the previous timestep into a single message that is used in the current timestep. Results from evaluating our models on a challenging set of StarCraft II micromanagement tasks with shared rewards show that FCMNet and FCMTran both outperform recent communication-based methods and value decomposition methods in almost all tested StarCraft II micromanagement tasks. We further improve the performance of our models by combining them with value decomposition techniques; there, in particular, we show that FCMTran with value decomposition significantly pushes the state-of-the-art on one of the hardest benchmark tasks without any task-specific tuning. We also investigate the robustness of FCMNet under communication disturbances (i.e., binarized messages, random message loss, and random communication order) in an asymmetric collaborative pathfinding task with individual rewards, demonstrating FMCNet’s potential applicability in real-world robotic tasks.

show abstract

Autonomous Bus Fleet Control Using Multiagent Reinforcement Learning

Cited by 5 publications

References 24 publications

Implementing Deep Reinforcement Learning in Autonomous Control Systems

Implementing Deep Reinforcement Learning in Autonomous Control Systems

Full communication memory networks for team-level cooperation learning

Full Communication Memory Networks for Team-Level Cooperation Learning

Contact Info

Product

Resources

About