RRT∗ and Goal-Driven Variable Admittance Control for Obstacle Avoidance in Manual Guidance Applications

In manual guidance robotic applications, like the handling of large and heavy objects in a cluttered environment, it is important to guarantee that the operator accurately reaches the goal position without collisions with working isles or other obstacles in the surrounding environment. When the transported object is bulky, the operator's view is obstructed and the situation becomes more critical. In this work, a novel variable admittance control provides the operator with a directional haptic feedback about the best motion direction towards the goal. This feedback allows the user to accurately reach the target position in a cluttered environment, also in case his/her view is partially or totally obstructed. To select the best motion direction in a cluttered workspace, a tree-based structure rooted in the goal is optimally built offline to fully explore the environment free-space based on the workspace layout and regardless of the initial position. Then, at each time instant, the optimal motion direction is determined based on the current position with respect to the exploring structure and on the user motion intention. In this work, to build the tree structure, we adapt RRT∗ algorithm to the manual guidance context and we define a tailored cost function. The performance is evaluated in many scenarios with a variable number of obstacles of different shapes involving several subjects and a Comau Smart Six robot.