Sensor fusion of vision, force and acceleration for compliant robot motion control

When a robotic arm has to interact with unknown environments, the efficiency and the accuracy of integrating all the sensor measurements, as well as given information, contribute towards the success of the task performed. In fact, in such operations, sensors play a critical role, helping to overcome many of the difficulties of uncertain world model and unknown environments.In this paper, a new sensor fusion approach for robotic arms with constrained motions is presented. It is based on the integration of three different kind of sensors: force, acceleration and computer vision, with the goal of improving the task performance in terms of accuracy and speed. The proposed sensor fusion approach offers a better estimation of the contact surface, eliminating errors due to the camera pose calibration or measurement disturbances such as the inertial forces. The proposed observer has been validated experimentally on a Stäubli robotic arm, showing how it improves the response of the task controller.