A scalable multi-robot task allocation algorithm

In modern warehouses, robots are employed along with humans to perform those jobs that are less productive and unsafe for humans. This paper focuses on the tasks where robots fetch outgoing objects from their respective storage racks and bring them to the packaging dock. This requires a careful task allocation along with route planning such that the total path traveled (cost) by all the robots to complete all the tasks is minimized. The number of tasks that can be performed in a single run (part of the same route) depends on the maximum capacity of the robot and the combined weight of the objects on that route. This task allocation problem can be mapped to the capacity-constrained vehicle routing problem (CVRP), which is an NP-hard problem. Though there exist a number of algorithms that provide a near-optimal solution to a CVRP instance, they do not scale well with the number of nodes (tasks). Thus, we develop a fast heuristic algorithm, called nearest-neighbor based Clustering And Routing (nCAR) that provides a solution in a fraction of time compared to the stateof-the-art solution, and also reduces the cost of the solution when there are a large number of nodes. We compare the performance of nCAR with the Google OR-Tools and obtain a solution at 1/6-th of time for 2000 tasks. Though OR-Tools provides a low-cost solution for small number of nodes, it takes 1.5 times more execution time and routes to complete all the tasks, compared to nCAR.