Optimal motion generation for autonomous vehicle in maze-like environment

In car-like autonomous vehicle systems, it is an essential task of generating the motion commands according to a given path/strategy. Quite a few theories and techniques have been proposed for the generation of the motion commands in autonomous vehicles, such as pure pursuit, Stanley's nonlinear feedback, chained-form control of kinematic model and the linearized optimal control of dynamic model. Here a non-linear optimization algorithm based on the vehicle's kinematic model and the actuators' model is proposed, which combines the control system dynamic behaviors and gives out the control sequences directly. It starts with modeling the local kinematic behavior and actuators' dynamics. Then online optimization algorithm is applied to the objective function of minimizing the energy cost, execution time and tracking error with some trade-off weights among them. The experiments showed that it worked well for vehicles running in maze-like environment.