Eckhard Freund

Real-time pathfinding in multirobot systems including obstacle avoidance

The paper describes an approach to the solution of the find path problem, including obstacle avoidance in multirobot systems. The design of multirobot systems requires an overall approach where not only parts of the structure, including the hierarchy are considered, but a complete concept, including the dynamics of the robots, has to be developed. In this paper the structure of these systems is based on the nonlinear con trol approach. The method for real-time pathfinding itself uses a systematic design procedure for multirobot systems, which includes the hierarchical coordinator. This hierarchical coordinator is designed for real-time collision avoidance, where the collision avoidance strategy is based on an analyti cally described avoidance trajectory that serves for collision detection as well as avoidance. The efficiency of the new approach for real-time pathfinding is demonstrated by several cases of practical interest, such as collision avoidance between three robots, interaction of three stationary robots and a stationary obstacle, as well as interaction of mobile robots and moving obstacles, including obstacles of variable size and different shapes.

Collision avoidance for industrial robots with arbitrary motion

For effective application of industrial robots in flexible manufacturing systems and in automatic assembly, guided automatic collision avoidance is an important and—in spite of many applications—so far unsolved problem. The basis is unconstrained path control of each robot controlled, for example, by optical sensors in connection with different priority levels assigned to the robots. Iterative, time-consuming, hierarchical control methods are not applicable due to the complexity and the real-time constraints of this task. As a new and successful approach the application of various hierarchical decision strategies in connection with extremely time-efficient evaluation of decision tables is presented. This concept is based on a suitable description of the actual possible collision spaces.