Abstract. An optimal control problem to find the fastest collision-free trajectory of a robot is presented. The dynamics of the robot is governed by ordinary differential equations. The collision avoidance criterion is a consequence of Farkas's lemma and is included in the model as state constraints. Finally an active set strategy based on backface culling is added to the sequential quadratic programming which solves the optimal control problem.Keywords: Optimal control, collision avoidance, backface culling, active set strategy.
Collision AvoidanceIn automotive industry robots have to work simultaneously on the same workpiece. The challenge is to find for each robot the fastest trajectory that avoids any collision with the surrounding obstacles and the other robots. We start with the establishment of the collision avoidance criterion.For simplicity, we suppose that only one obstacle is present in the workspace. As in [7,8] a collision detection can be obtained when the robot is approximated by a union of convex polyhedra. This union is called P and it given bywhere n P is the number of polyhedra in P . If p i denotes the number of faces inSimilarly, the obstacle is approximated by the following union of convex poly-where n Q is the number of polyhedra in Q. If q j is the number of faces in Q (j) , then C (j) ∈ R qj ×3 and d (j) ∈ R qj for j = 1, . . . , n Q . In the following, n P , A, b D.