iRobotLab

研究/各研究の話/2013年度

Summary†

• If depending on the action a robot takes, the future state will vary infinitely.
  • If we take this point into consideration, we need to simultaneously predict the internal state and the action that the robot adopt.
• Focusing on the method that compensate the current action as the appropriate action using next time and future actions that robots will take.
• For achievement of this problem, we state advance prediction using Online SVR (support vector regression) and LQR (linear-quadratic regulator).
  • Proposed system is composed two parts.
    • State Predictor: To predict next state of the robot. In this study we apply Online SVR.
    • Action Predictor: To predict next action of the robot. In this study we apply two method. First, Online SVR. And second, LQR.

Outline figure†