Model Predictive Control-Based Probabilistic Search Method for Autonomous Ground Robot in a Dynamic Environment

Abstract

Target search using autonomous robots is an important application for both civil and military scenarios. In this paper, a model predictive control (MPC)-based probabilistic search method is presented for a ground robot to localize a stationary target in a dynamic environment. The robot is equipped with a binary sensor for target detection, of which the uncertainties of binary observation are modeled as a Gaussian function. Under the model predictive control framework, the probability map of the target is updated via the recursive Bayesian estimation and the collision avoidance with obstacles is enforced using barrier functions. By approximating the updated probability map using a Gaussian Mixture Model, an analytical form of the objective function in the prediction horizon is derived, which is promising to reduce the computation complexity compared to numerical integration methods. The effectiveness of the proposed method is demonstrated by performing simulations in dynamic scenarios with both static and moving obstacles.