Fuzzy collision avoidance for industrial robots

An advanced concept to solve the collision avoidance problem in multi-robot workcells is presented which holds realtime conditions for industrial robot controllers. The approach is based on a fuzzy set description of critical configurations between robots in a common workspace. The authors' strategy assumes a careful offline path planning to unburden the algorithm from online path planning but to handle extraordinary events, which can not be treated offline. Though the authors' collision avoidance strategy is local, in general it switches towards the preplanned trajectory with decreasing degree of conflict and finally reaches planned goal positions. To handle the collision avoidance problem in realtime, the complex kinematic chain of any industrial robot with six degrees of freedom is replaced by a drastically reduced actual robot (RAR). Fuzzification of the actual RAR position in a workspace provides a fuzzy membership vector to sectors, which subdivide the robots' cylindrical workspace. A practical fuzzy-rulebase is given to evaluate the degree of conflict between the actual fuzzy membership vectors of two RARs and to propose secure fuzzy-membership vectors. Defuzzification gives crisp locations of a collision-free trajectory which are restricted by the preplanned trajectory locations if the conflict is small. The efficiency of this strategy is illustrated by an example with two PUMA 562 robots in decoupled motion.