An empirical study of task bundling for sequential stochastic tasks in multi-robot task allocation

This paper studies multi-robot task allocation in a setting where tasks are revealed sequentially and where it is possible to execute bundles of tasks. Particularly, we are interested in tasks that have synergies so that the greater the number of tasks executed together, the larger the potential performance gain. We consider tasks that are fed to robots for an infinite or indefinite time horizon. Robots may bundle multiple tasks to minimize some system cost (e.g., fuel), but doing so incurs an additional waiting time for bundling tasks. If the robots reduce their bundle size to minimize waiting time, task executions fail to make the most of possible synergies. Thus the system cost may increases, and the queue of waiting tasks may even overflow if task completions too slow. This paper is an analysis of bundling, giving an understanding of the important bundle size parameter. Based on qualitative properties of any multi-robot system that bundles sequential stochastic tasks, we propose multiple simple bundling policies. Experiments show how these polices perform in the multi-robot routing domain, showing that they are efficient compared to a baseline system where robots do not bundle tasks but iterate instantaneous assignments and executions of tasks.