A Generalised Multi-Attribute Task Sequencing Approach for Robotics Optical Inspection Systems

Robot programming usually consists of four steps: (1) task planning; (2) task sequencing; (3) path planning and (4) motion planning. Task (2) and (3-4) are strongly coupled. For example, the optimal robot path, which is function of the robot kinematics, relies on the pre-defined schedule of tasks, whose sequencing is computed based on the assumption that the travelling “cost” from one task to the next is only driven by the Euclidean distance in Cartesian space. Current methods tends to decouple the problem and sequentially compute the task sequencing in the T-space, and then compute the robot path by solving the inverse kinematics in the C-space. However, those approaches suffer the capability to reach a global optimum. This paper aims at developing a novel approach which integrates some of the key computational requirements of the path planning in the early stage of the task sequencing. Multi- attribute objectives are introduced to take into account: robot pose and reachability, data quality, obstacles avoidance, overall cycle time. The paper introduces a novel multi-attribute approach to find the optimized task sequencing via candidate poses solving inverse kinematics in the T-space. This is based on the core idea to combine T-space and C-space. The proposed solution has been tested on a vision-based inspection robot system with application to automotive body assembly systems. Results could however impact a wider area, from navigation systems, game and graph theory, to autonomous driving systems.