Planning of multiple autonomous vehicles using RRT

Abstract

Criteria such as driving safety and overall travel efficiency have led to increasing attempts towards autonomy of vehicles wherein different vehicles can plan their journey, maneuver as per scenario, and communicate to each other to create an error free travel plan. In this paper we present the use of Rapidly Exploring Random Trees (RRT) for the planning of multiple vehicles in traffic scenarios. The planner for each vehicle uses RRT to generate a travel plan. Spline curves are used for smoothing of the path generated by the RRT, which follows non-holonomic constraints. Priority is used as a coordination mechanism wherein a higher priority vehicle attempts to avoid all lower priority vehicles. The planner attempts to find the maximum speed at which the vehicle may travel and the corresponding path. Experimental results show that by using the approach, multiple vehicles may be planned to travel in a fairly complex obstacle grid. Further, the vehicles exhibited behaviors including vehicle following and overtaking which are commonly seen in everyday driving.