Safe and Efficient Mobile Robot Path Planning in Open World Environments

Abstract

An open world environment in mobile robot path planning generally refers to a working space with unpredefined/unknown obstacles. It is a challenging task for path planning in open world environments, especially those having multiple dynamic obstacles with time-varying moving speed. In this context, safety become the key performance indicator, resulting in most of global path planning algorithms (e.g., A*, RRT*, etc) no longer in force due to their inadequate adaptability to dynamic environments. Therefore, we propose in this paper a novel safety-aware RRT* algorithm for the path planning of mobile robots in open world environments with dynamic obstacles. We leverage the safety-field model (firstly developed for self-driving cars with constrained path, i.e. vehicle lane) to mobile robot path planning (with unconstrained moving routes in the working space in general). In particular, a fine-grained energy map is developed as a safety indicator illustrating the real-time risk to the working space caused by dynamic obstacles in terms of distance away from the robot (which presents the potential energy), moving direction and moving speed (which present the kinetic energy). This is in strong contrast to the artificial potential field model that is widely utilized in the obstacle avoidance in dynamic environments, which only considers the potential energy and provides a coarse risk indication to robots. In order to improve the efficiency of the proposed safety-aware RRT* algorithm, we further propose a pruning strategy for the sampling process of RRT* based on the fine-grained energy map to guide the rapidly-exploring random tree growing in a safe direction. The parent node re-selection and rewiring process of the proposed algorithm also incorporates the safety-field based energy map. Experimental results show that the proposed safety-aware RRT* algorithm can provide excellent safety and efficiency in dynamic environments.