Experiments in Sensor-based Collis1on Avoidance

Abs t r ac t A control strategy for sensor-based collision avoidance and the supporting experimental results are reported in this paper. The real-time arm control software cc)ntinuously monitors the object clistance measured by the proximity sensors. When this distance is less than a preset threshold, the collision avoidance control acticm is initiated to inhibit motion towards the object and thus prevent collision. This is accomplished by employing an outer feedback loop to perturb the end-effecter reference motion trajectory in real-time based on the sensory data. The perturbation is generated by a proportionall)lus-integral (1’1) collision avoidance controller acting on the difference between the sensecl distance ancl tlte user-specified threshold. The proposed approach to collision avoidance is analogous to controlling the virtual force applied to the end-effecter by a hypothetical spring-plus-damper attached to the object’s surface. This approach is computationally very fast, requires minimal modification to the existing manipulator positioning system, and provides the manipulator with an on-line collision avoidance capability to react autonomously’ and intelligently. Experimental results are reported to demonstrate end-effecter collision avoidance with an approaching target and while reaching inside a truss opening,