Advances in Reinforcement Learning for Autonomous Systems

The field of reinforcement learning (RL) is rapidly advancing, with a focus on developing more efficient and effective methods for autonomous systems. Recent developments have highlighted the importance of carefully designing reward functions to guide agents towards optimal decision-making. Additionally, there is a growing interest in using probabilistic and non-parametric methods to model complex tasks and environments. Notable papers in this area include:

MAER-Nav, which introduces a novel framework for bidirectional motion learning in robot navigation, enabling robust navigation in complex environments.
Reinforcement Learning for Safe Autonomous Two Device Navigation of Cerebral Vessels in Mechanical Thrombectomy, which proposes a safe dual-device RL algorithm for navigating cerebral vessels, achieving a 96% success rate in simulation.
On learning racing policies with reinforcement learning, which leverages domain randomization and actuator dynamics modeling to learn a racing policy that outperforms expert human drivers in RC racing.

Sources

Reward Design for Reinforcement Learning Agents

Pretrained Bayesian Non-parametric Knowledge Prior in Robotic Long-Horizon Reinforcement Learning

CRLLK: Constrained Reinforcement Learning for Lane Keeping in Autonomous Driving

Evaluation of Remote Driver Performance in Urban Environment Operational Design Domains

A Survey of Reinforcement Learning-Based Motion Planning for Autonomous Driving: Lessons Learned from a Driving Task Perspective

MAER-Nav: Bidirectional Motion Learning Through Mirror-Augmented Experience Replay for Robot Navigation

Reinforcement Learning for Safe Autonomous Two Device Navigation of Cerebral Vessels in Mechanical Thrombectomy

Control Center Framework for Teleoperation Support of Automated Vehicles on Public Roads

Learning from Disengagements: An Analysis of Safety Driver Interventions during Remote Driving

Probabilistic Curriculum Learning for Goal-Based Reinforcement Learning

Reinforcement Learning for Solving the Pricing Problem in Column Generation: Applications to Vehicle Routing

On learning racing policies with reinforcement learning