wangruoyao/pure_pursuit

ROS Pure-Pursuit Package for the F1/10 Racecar

F1/10 ROS Pure-Pursuit Racing Algorithm

Pure-Pursuit path tracking algorithm has proven itself to be highly capable and robust for high speed autonomous racing, and is our algorithm of choice for this simulator and the F1/10 physical platform. More information about the pure-pursuit algorithm can be found in this paper. The algorithm requires the odometry information of the racecar and a global racing line and works by generating a relative heading to a waypoint within the raceline which is a distance l_d away from the racecar on the raceline. This distance is the single tunable property of the pure-pursuit algorithm called the lookahead distance. The maximum achievalbe velocity of the racecar is proportional to the lookahead distance as a general rule of thumb. The paper linked above talks in general about the shortcomings of the pure-pursuit controller, the most significant of which is the tendency to cut corners (and potentially collide with the race track) if the lookahead distance is not properly tuned.

To overcome the limitations of the pure-pursuit controller as described in the paper, and to make it a true autonomous racing algorithm, we have introduced some changes to the algorithm by using a supervisory node to adapt the lookahead distance based on the properties of the track at each waypoint along the raceline. We call this modification the Adaptive-Lookahead Pure-Pursuit. For this tutorial, we will use one of the modified racelines which contain information about lookahead distance at each waypoint to demonstrate the performance of the pure-pursuit controller and why we prefer this controller for autonomous racing. Kill all existing ROS nodes and the simulator (if running) and open a new terminal to enter the following command:

user@ros-computer: roslaunch <master_ROS_package_name> purepursuit_one_car.launch

The rviz window created uses a pre-configured setup for the pure-pursuit controller. The elements to note within the window are the raceline which is composed of different segments denoted using different colors which correspond to unique lookahead distances assigned to that segment of the raceline. The racecar will achieve the highest velocity at green segments and lowest at the red segments. It is interesting to note that the racecar in this tutorial is travelling twice as fast in its slowest setting compared to the highest speed achieved using the T.E.B. planner and ROS navigation described in the previous tutorial section.