Soft-Computing Techniques for the Trajectory Planning of Multi-Robot Manipulator Systems

摘要

This chapter has presented a discussion on the application of soft-computing techniques for the trajectory planning of multi-robot manipulator systems. The application of such techniques derived on an approach based on a combination of fuzzy logic and genetic algorithms to solve in parallel the trajectory planning of two manipulators sharing a common workspace where both manipulators have to adjust their trajectories in the event of a possible collision. The planner combines a simple-GABTP with fuzzy correction units. The simple-GABTP solves the trajectory planning problem as an initial estimation without considering the influence of any obstacles. Once this stage is reached, fuzzy units assigned to each articulation verify that the manipulator is not facing a possible collision based on the magnitude of the APF exerted by the manipulators when considered as obstacles. If the initial output by the GABTP takes a link or links of the manipulator near the influence of that corresponds the modelled APF, the fuzzy units evaluate a correction value for the to the link and articulation involved in the possible collision and modifies not only the for that articulation, but also modifies the 's for the more anterior articulations in case where a distal articulation is involved. The approach presented here has shown a robust and stable performance throughout the different simulated cases. It produced suitable trajectories that are easily obtained since the direct output from the FuGABTP algorithm is a set of 's after each iteration. When a time dimension is added, these provide not only positional information but also the angular velocities and accelerations necessary to describe the trajectory profiles of each manipulator and allow the calculation of the necessary torques to be applied at each joint to produce the obtained trajectory. Since the planning is done in parallel for both manipulators, no further planning or scheduling of their motions are necessary as is required with decoupled approaches since the trajectories are free from any collision. The implementation of the simple-GABTP in the algorithm solves the problem of over compensation associated with the fuzzy units by maintaining the direction of the manipulators towards the desired goal at all times, keeping the links of the manipulator from over-reacting when approaching an obstacle. Finally, the proposed algorithm was applied to solve the trajectory planning problem of two manipulators sharing a common workspace with individual goals. In this case each manipulator considers the other as a mobile obstacle of unknown trajectory. The considered cases were successfully solved, obtaining suitable trajectories for both manipulators, obtaining an average of 0.20 seconds per iteration for the 7dof systems in the dynamic cases. In conclusion, the approach discussed in this chapter could be applicable for real-time application since the compilation of the algorithms that conform the suggested approach into executable files could reduce even more the iteration time.