A Penalty Method Based Approach for Autonomous Navigation using Nonlinear Model Predictive Control (1805.02524v2)

Abstract: This paper presents a novel model predictive control strategy for controlling autonomous motion systems moving through an environment with obstacles of general shape. In order to solve such a generic non-convex optimization problem and find a feasible trajectory that reaches the destination, the approach employs a quadratic penalty method to enforce the obstacle avoidance constraints, and several heuristics to bypass local minima behind an obstacle. The quadratic penalty method itself aids in avoiding such local minima by gradually finding a path around the obstacle as the penalty factors are successively increased. The inner optimization problems are solved in real time using the proximal averaged Newton-type method for optimal control (PANOC), a first-order method which exhibits low runtime and is suited for embedded applications. The method is validated by extensive numerical simulations and shown to outperform state-of-the-art solvers in runtime and robustness.