Adaptive motion planning over environmental uncertainties

Abstract

Robots in the ideal environment and in the real world behave differently even though they received the same instructions from users. Since there exists unexpected environmental disturbance in the real world, robots are possible to move in imperfect manners. Therefore, users should consider how to deal with the disturbance when they design robot control, and it is very cumbersome. Our goal is to provide abstraction to users that they can control robots only with high-level robot control information. Users do not have to know robot control details such as how to translate a robot velocity to a motor control power value and how to deal with unpredictable environmental changes. In this paper, we propose a user-centric robot control framework which provides users with abstractions by hiding all the robot control details to users. In order to provide users with full abstraction (i.e., to make actual robot behave as similar as robots in the ideal environment), we propose feedback control methods. We implemented a PID controller which is one of the most widely used feedback controllers and we can reduce gaps between actual robots and robots in the ideal environment via the PID controller. However, we found out that the PID controller was insufficient to guarantee the deadline property, so we applied adaptive control method. One challenge of implementing adaptive control method is to develop a mechanism to dynamically change controller parameters to support the deadline property better and we developed a heuristic of adjusting controller parameters.