High accuracy is usually difficult to obtain with a robotic arm installed on a mobile base, since the errors of the base are transferred to the manipulator. This paper proposes a method to address this problem through integration of a self-calibration algorithm and low-cost sensors. The self-calibration algorithm might be repeated several times during execution of a mission by the robot and is only based on the internal sensors of the robot, meaning that external observers or reference point transceivers (e.g., ultrasonic transceivers) are not used. The proposed self-calibration system was implemented on a pole climbing robot and effectively improved the positioning accuracy of the climbing arm.