Classic robotics and the almost recent robots still rely on fixed behavior basd control. So, the recent paradigm of robots and robotics is increasing the robot"s ability that can deal with uncertainties in the real world. A general approach is Reinfor...
Classic robotics and the almost recent robots still rely on fixed behavior basd control. So, the recent paradigm of robots and robotics is increasing the robot"s ability that can deal with uncertainties in the real world. A general approach is Reinforcement Learning(RL). Neural Q-Learning was introduced in 2000 and it is based on Q-Learning for Linear Quadratic Regulation(LQR). We replace it with online neural Q-Learning and apply it to a simulation of a straightforward nonlinear mobile robot, Two-Wheeled Inverted Pendulum(TWIP) and we show that TWIP is successfully balancing.