Multi-robot Scheduling and Path-Planning for Non-overlapping Operator Attention

Abstract

There is a growing need for robots to perform complex tasks autonomously. However, there remain certain tasks that cannot -or should not -be completely automated. While these tasks may require one or several operators, we can oftentimes schedule when an operator should assist. We build on our previous work to present a methodology for allocating operator attention across multiple robots while attempting to minimize the execution time of the robots involved. In this paper, we: 1) Analyze of the complexity of this problem, 2) Provide a scalable methodology for designing robot policies so that few operators can oversee many robots, 3) Describe a methodology for designing both policies and robot trajectories to permit operators to assist many robots, and 4) Present simulation and hardware experiments demonstrating our methodologies.

Scheduling and Path-Planning for Operator Oversight of Multiple Robots

“…We build upon our recent work [10,11], to perform multi-robot planning [40,41]. We also find complementary goals in [42] where a robot attempts to move from one location in its environment to another by calculating which obstacles can be minimally displaced to generate a feasible trajectory.…”

“…In our previous work [11], we described the operator scheduling problem and presented a geometric approach for its solution. There were several issues with the proposed methodology related to the computational complexity of creating the entire set of obstacles with the coordination space.…”

“…Our preliminary solution [11] required generating the entire set of obstacles within the coordination space. Here, we instead use a lazy approach which only checks sampled locations.…”

“…Thus, our objective is to develop a planning strategy for the remote task involving multiple robots such that the operator can pay sufficient attention to each robot during critical operations. This work's contributions are: extending our preliminary ideas from [10,11] in the following directions: First, we analyze the complexity of this problem. Secondly, we present a sampling-based approach that allows us to design policies for many teleoperators instead of a complete algorithm that only works for a small set of operators.…”

“…This procedure is performed on the robots with the shortest current plans so that extensions to their plans due to re-planning should have a minimal effect on the mission's overall length. Then generate alternative trajectories for the robots, provided operator-denied times o t den , and create an alternative goal location x altgoal to account for any shifts in the ending times of the robot plans (lines [8][9][10][11].…”