Improved Obstacle Avoidance for Autonomous Robots with ORCA-FLC

TL;DR

This paper introduces ORCA-FL, an enhanced obstacle avoidance method for autonomous robots that uses fuzzy logic controllers to better handle uncertainty, outperforming traditional ORCA in multi-agent scenarios.

Contribution

The paper proposes ORCA-FL, integrating fuzzy logic controllers with ORCA, and introduces a fuzzy Q reinforcement learning algorithm for tuning FLCs, improving obstacle avoidance performance.

Findings

ORCA-FL reduces collisions compared to ORCA in multi-agent tests.

Fuzzy Q reinforcement learning effectively optimizes FLC parameters.

ORCA-FL performs better when robot velocity exceeds a threshold.

Abstract

Obstacle avoidance enables autonomous agents and robots to operate safely and efficiently in dynamic and complex environments, reducing the risk of collisions and damage. For a robot or autonomous system to successfully navigate through obstacles, it must be able to detect such obstacles. While numerous collision avoidance algorithms like the dynamic window approach (DWA), timed elastic bands (TEB), and reciprocal velocity obstacles (RVO) have been proposed, they may lead to suboptimal paths due to fixed weights, be computationally expensive, or have limited adaptability to dynamic obstacles in multi-agent environments. Optimal reciprocal collision avoidance (ORCA), which improves on RVO, provides smoother trajectories and stronger collision avoidance guarantees. We propose ORCA-FL to improve on ORCA by using fuzzy logic controllers (FLCs) to better handle uncertainty and imprecision for obstacle avoidance in path planning. Numerous multi-agent experiments are conducted and it is shown that ORCA-FL can outperform ORCA in reducing the number of collision if the agent has a velocity that exceeds a certain threshold. In addition, a proposed algorithm for improving ORCA-FL using fuzzy Q reinforcement learning (FQL) is detailed for optimizing and tuning FLCs.