Operator Intent Prediction with Subgoal Transition Probability Learning for Shared Control Applications

Abstract

In this paper, we propose a novel intent prediction method for shared control applications where the task is modeled via subgoals. The proposed method takes into consideration the human operator’s real-time action and updates the prediction model based on observed subgoal transitions. We describe the transition probabilities update law, its convergence property, and its effectiveness in reflecting observed subgoal transitions with constructed probabilities. Experiments were conducted on a physical platform using a hydraulic excavator for a trenching-and-loading task with human-machine shared control. Results corroborate the proposed method and indicate that it can effectively update the prediction model and better reflect subgoal transition probabilities based on observations.