Accurate sensorless lead-through programming for lightweight robots in structured environments

Nowadays, programming an industrial manipulator is a complex and time-consuming activity, and this prevents industrial robots from being massively used in companies characterized by high production flexibility and rapidly changing products. The introduction of sensor-based lead-through programming approaches (where the operator manually guides the robot to teach new positions), instead, allows to increase the speed and reduce the complexity of the programming phase, yielding an effective solution to enhance flexibility. Nevertheless, some drawbacks arise, like for instance lack of accuracy, need to ensure the human operator safety, and need for force/torque sensors (the standard devices adopted for lead-through programming) that are expensive, fragile and difficult to integrate in the robot controller. This paper presents a novel approach to lead-through robot programming. The proposed strategy does not rely on dedicated hardware since torques due to operator's forces are estimated using a model-based observer fed with joint position, joint velocity and motor current measures. On the basis of this information, the external forces applied to the manipulator are reconstructed. A voting system identifies the largest Cartesian component of the force/torque applied to the manipulator in order to obtain accurate lead-through programming via admittance control. Finally an optimization stage is introduced in order to track the joint position displacements computed by the admittance filter as much as possible, while enforcing obstacle avoidance constraints, actuation bounds and Tool Center Point (TCP) operational space velocity limits. The proposed approach has been implemented and experimentally tested on an ABB dual-arm concept robot FRIDA