LACART keeps small autonomous race cars clear of obstacles with lookahead barrier functions

Pedro Lopes, Amit Shivam and Luis Tiago Paiva have described LACART in an arXiv paper, a method that adjusts a vehicle's commands to avoid obstacles and follow a reference path, combining lookahead control barrier functions, a relaxed control Lyapunov function and quadratic programming. It handles circular obstacles and track edges in the same optimisation step, with autonomous racing as the test case.

In simulation it produced paths similar to an earlier method with lower median computation times, 0.31 against 0.33 ms per step. The work, based on a master's dissertation, was also shown on a 1:43-scale car with camera-based estimation at a 25 Hz control rate.