Industrial robot workflow consists of a set of tasks that has to be repeated multiple times. The efficiency with which the robot performs its tasks is important for majority of the production domains. In fact, the more tasks a robot can perform in a time interval the more advantages it provides to the company. The majority of tasks have a certain freedom of execution. For example, closed-contour tasks can be started and finished at any point of the curve or a set of possible starting end-effector orientations exist. We propose a method that is able to improve the given sequence of robotic tasks by adding certain freedom to (i) the starting position point along the curve, (ii) orientation of the end-effector and (iii) a robot configuration. The proposed approach does not depend on the production domain or the algorithm for constructing initial task sequence.
We propose to apply hierarchical optimization. Its goal is to perform a fast local search on every nested stage instead of applying slow global optimization techniques to the whole problem. We specify three nested stages. Each outer stage depends on the results of optimization on the inner stage.
The algorithm showed significant improvement of the production time on the test instances from the cutting-deburring domain, i.e., it improved the time by 38% for the instances with 15 holes in 68 seconds.The evaluation results are available online in MS Excel 2007 format here. Further details can be found in the following publication:
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2014.