Evaluating Multi-Robot Teamwork in Parameterised Environments

Abstract

Abstract. The work presented here investigates the impact of certain environmental parameters on the performance of a multi-robot team conducting exploration tasks. Experiments were conducted with physical robots and simulated robots, and a diverse set of metrics was computed. The experiments were structured to highlight several factors: (a) single-robot versus multi-robot tasks; (b) independent versus dependent (or "constrained") tasks; and (c) static versus dynamic task allocation modes. Four different task allocation mechanisms were compared, in two different exploration scenarios, with two different starting configurations for the robot team. The results highlight the distinctions between parameterised environments (characterised by the factors above, the robots' starting positions and the exploration scenario) and the effectiveness of each task allocation mechanism, illustrating that some mechanisms perform statistically better in particular environment parameterisations.

Mechanism Selection for Multi-Robot Task Allocation

“…In the fixed start, random task locations environments (Table 2), ssi was somewhat less dominant. These results show that, while ssi performs better overall, psi allocations do sometimes result in performance that is competitive with ssi allocations, as we have seen in previous work [23,24].…”

“…We are working to improve its performance for more general cases of a known environment, in particular for randomised robot start locations. We are also working to generalise the approach to consider more mechanisms and richer environments-combining it with our previous work on multi-robot, precedenceordered, and dynamically appearing tasks [23,24]-and on other maps. …”

“…Our work makes use of MRTeAm, a software framework for conducting multirobot coordination experiments [23,24]. It is written on top of ROS [18] and its main components are a central auctioneer agent and multiple robot controller agents.…”

“…In some environments, a simple, greedy mechanism which might not be expected to perform well in the general case may, in fact, perform competitively with more sophisticated mechanisms, with the advantage of scaling better. Prior work has shown evidence that this is the case in both simulated and physical experiments [22][23][24].…”

“…In particular, in our earlier work, we showed [23,24] that while the sequential single-item auction [12] performs better than the parallel single-item auction [12] when the allocation is carried out with robots clustered together geographically, this advantage diminishes as robots are distributed over space and tasks are distributed over space and time. Based on this observation, in this paper we propose a portfolio-based approach to the MRTA problem.…”