Small humanoid robot localization by landmarks

This paper presents an academic prototype small humanoid robot with vision-system to achieving the posture estimation. With the on-board personal digital assistant (PDA) to in charge of the image processing and then recognize its surroundings. According to the information of the image data from the PDA-based real-time image sensor, the Nios-based motor control chip is employed to achieve the desired motions, including forward walking, backward walking and turning. To navigate the robot, a triangulation method with a CCD camera with three artificial color landmarks is used to determine the robotpsilas posture using the on-board PDA. An extended Kalman filter (EKF) algorithm is employed to improve estimation accuracy of the humanoid robot. Computer simulation and experiment results are conducted to verify the efficacy of the proposed method.