Interacting multiple model-based human motion prediction for motion planning of companion robots

Abstract

Motion planning of human-companion robots is a challenging problem and its solution has numerous applications. This paper proposes an autonomous motion planning framework for human-companion robots to accompany humans in a socially desirable manner, which takes into account the safety and comfort requirements. An Interacting Multiple Model-Unscented Kalman Filter (IMM-UKF) estimation and prediction approach is developed to estimate human motion states from sensor data and predict human position and speed for a finite horizon. Based on the predicted human states, the robot motion planning is formulated as a model predictive control (MPC) problem. Simulations have demonstrated the superior performance of the IMM-UKF approach and the effectiveness of the MPC planner in facilitating the socially desirable companion behavior.