Real-time constraint-based planning and control of robotic manipulators for safe human–robot collaboration

Recent advances in Rapidly-exploring random tree: A review

“…In Ref. [ 241 ], Kelly et al design a human-assisted RRT path planner and validate on a Franco-Emika Panda robotic arm, which overcomes the constraint of actuator saturation and limited joint ranges and avoids the property of RRT that it tends to fall into the local optimal solution, and the authors summarize the conclusions as follows: (1) The planner takes 30–130 ms to plan paths with varying levels of obstacle complexity, so the human-robot collaborative system is able to perform path-planning tasks in complex scenarios when the rate of change of dynamic obstacles in the scene is less than 1 Hz. (2) The trajectory-based explicit reference governor, as a closed feedback control scheme, has a maximum average computation time of 1 ms and can therefore be neglected with respect to the path planner.…”

“…Under some accuracy-sensitive path planning requirements, such as ore transportation [ 222 ], the smoothed paths may no longer be asymptotically optimal solutions under the RRT framework, and may even generate unreasonable local paths, which makes manual intervention extremely important. Certain experimental results show that better performance can be achieved with human-assisted path planning [ 241 ], which is one of the conclusions that human-robot collaboration is the future trends. Balancing unbiased and biased sampling based on human experience is also one of the important research directions for RRT improvement [ 182 ].…”