Performance Analysis of Motion Planning for Outdoor Autonomous Delivery Robot

Motion planning is a process that integrates path planning and kinematic constraints to facilitate lateral and longitudinal control. For the context of autonomous delivery in an outdoor environment, our study focused on a designated delivery area starting from Dong Seoul University’s Building 2 as the starting point to Building 3 as the destination. To chart potential paths between these two points, we employed algorithms such as the A*, Rapidly-exploring Random Tree (RRT) and RRT*. These algorithms were evaluated based on criteria like generated path, delivery time, and stability to derive and analyze optimal routes for various scenarios. The actual navigation along these plotted routes taken into account kinematic constraints. We employed a two-axle rear-wheel-drive model based on the bicycle model for this analysis. Moreover, to set the Look Ahead Distance (LAD), we defined speed and curvature rooted in driving stability and determined the appropriate LAD for route navigation based on these parameters. While A* produced consistent path planning results, it generally took longer in computation time compared to RRT and RRT*. In other hand, RRT and RRT* were faster but displayed greater variability in outcomes. Thus, the selection of an appropriate path planning method should be important upon the operational environment and specific requirements.

I. Introduction

Autonomous driving technology has been steadily advancing and is being applied in various industrial sectors. Especially in delivery services, due to the recent pandemic situation and the need for social distancing and minimizing contact, the utilization of autonomous robots has been increasing [1] [2] [3]. In this context, the importance of research on stable and efficient robot movement, path planning, and path tracking is becoming even more significant.