Closed-Loop Task Allocation Using Inter-Robot Encounters

Collaborative teams of robots can accomplish multiple tasks at the same time by dividing themselves among the tasks. For example, swarming robots operating in disaster-prone environments might be divided among different tasks such as surveillance, data relay, and rescue. The problem of allocating tasks to individual robots becomes especially challenging when there is no central coordinator and when the robots have limited communication as well as sensing capabilities.

In naturally occurring swarms, e.g., insect colonies, simple and local interactions among individuals contribute to the collective operations of the swarm. For example, ant colonies have demonstrated remarkably complex behaviors made possible via simple interactions which occur when individual ants simply pass by each other as they move around.

Inspired by such observations, this project develops a mechanism to allocate tasks to individual robots in a decentralized manner, without the need for any communication among the robots. We allow robots to switch between tasks with a given probability when they encounter other robots in the domain. By appropriately designing the transition probabilities and allowing the robots to measure the task allocated to the encountering robot, we illustrate that the swarm can achieve any desired task distribution. Furthermore, once the desired task distribution has been achieved, the robots can “turn-off” the task transitions in a closed-loop manner, using information obtained only from local encounters. The following video illustrates the deployment of the decentralized task allocation algorithm on the Robotarium:

 

Investigators:

  • Siddharth Mayya
  • Sean Wilson
  • Magnus Egerstedt

Related Publications:

Siddharth Mayya, Sean Wilson, Magnus Egerstedt, “Closed-Loop Task Allocation in Robot Swarms Using Inter-Robot Encounters”, Under Review, 2019.

Comments are closed.