Comparing Online Robot Joint Space Trajectory Optimization for Task Space Applications

Online trajectory planning allows robotic manipulators to adapt to changing environments and dynamic tasks, which becomes increasingly relevant through human-robot interaction. Optimizing motion objectives is a common way of planning but is usually conducted in joint space, whereas many planning objectives are rather defined in task space. Joint Space Planning (JSP) first transforms target poses into joint space and then performs conventional optimization. Implicit Task Space Planning (ITSP) uses a special objective function to directly consider task quantities and is similar to economic model predictive control. Remarkably, comparative studies of both approaches with respect to their performance are scarce. This contribution aims at filling this gap by providing a systematic analysis of both methods using a 6-DoF collaborative robot in practical experiments. Results show that the first method is less sensitive to local minima and profits from active reconfiguration abilities while it generates less straight-lined motions and is not able to account for target manifolds in the task space. The second method can easily consider target manifolds and performs more straight-lined motions while it is more sensitive to local minima and provides active reconfiguration only partially. Although the cost function of the second method is significantly more complex, the computational effort relativizes during planning.

Automated summary

This study compared the performance of Joint Space Planning and Implicit Task Space Planning in robotic trajectory planning, finding that each method has its own advantages and disadvantages, with the former being less sensitive to local minima and the latter more easily considering target manifolds.