Improving Kinodynamic Planners for Vehicular Navigation with Learned Goal-Reaching Controllers

TL;DR

This paper introduces a learning-enhanced kinodynamic planning method that uses offline-trained controllers to improve path quality and efficiency in vehicular navigation, adaptable across different environments.

Contribution

It presents a novel framework combining reinforcement learning with sampling-based kinodynamic planners to generate promising controls for faster, higher-quality path planning.

Findings

Higher quality paths with fewer iterations

Reduced computation time compared to traditional methods

Effective across different robotic systems

Abstract

This paper aims to improve the path quality and computational efficiency of sampling-based kinodynamic planners for vehicular navigation. It proposes a learning framework for identifying promising controls during the expansion process of sampling-based planners. Given a dynamics model, a reinforcement learning process is trained offline to return a low-cost control that reaches a local goal state (i.e., a waypoint) in the absence of obstacles. By focusing on the system's dynamics and not knowing the environment, this process is data-efficient and takes place once for a robotic system. In this way, it can be reused in different environments. The planner generates online local goal states for the learned controller in an informed manner to bias towards the goal and consecutively in an exploratory, random manner. For the informed expansion, local goal states are generated either via (a) medial axis information in environments with obstacles, or (b) wavefront information for setups with traversability costs. The learning process and the resulting planning framework are evaluated for a first and second-order differential drive system, as well as a physically simulated Segway robot. The results show that the proposed integration of learning and planning can produce higher quality paths than sampling-based kinodynamic planning with random controls in fewer iterations and computation time.