Path planning for autonomous vehicles using clothoid based smoothing of A' generated paths and optimal control

Sammanfattning: Autonomous vehicles is a rapidly expanding field and the need for robust and efficient path planners is high. We approach the global path- planning problem for an autonomous load carrier for quarry environments, developed at Volvo Construction Equipment, using two methods: a two- step path planning and path smoothing approach, and a method based on an optimal control formulation of the path planning problem. The two-step method is based on smoothing an initial path found by A', an efficient grid search algorithm, by fitting a curve consisting of as few clothoid segments as possible to the A' path. The smoothing is done by rewriting the non-linear optimization problem to a convex form by a linearization of the deviation constraints around the curvature of the A' path. An iterative method is then used to relax the l0-norm, which measures the number of non-zero elements in a vector, with a weighted l1-norm which, in turn, is then solved efficiently using CVX in Matlab. The optimal control based path planning method solves the nonlinear optimization problem using IPOPT. It was found that the completeness of the A' algorithm, coupled with the guaranteed solution of a convex problem, resulted in a very robust method that was able to find paths through mazes and difficult situations. The optimal control approach produced better paths, but had a tendency to sometimes show inconsistent behavior.