Decentralized prioritized planning in large multirobot teams

Abstract

Abstract-In this paper, we address the problem of distributed motion planning for large teams of hundreds of robots in constrained environments. We introduce two distributed prioritized planning algorithms: an efficient, complete method which is shown to converge to the centralized prioritized planner solution, and a sparse method in which robots discover collisions probabilistically. Planning is divided into a number of iterations, during which every robot simultaneously and independently computes a planning solution based on other robots' path information from the previous iteration. Paths are exchanged in ways that exploit the cooperative nature of the team and a statistical phenomenon known as the "birthday paradox". Performance is measured in simulated 2D environments with teams of up to 240 robots. We find that in moderately constrained environments, these methods generate solutions of similar quality to a centralized prioritized planner, but display interesting communication and planning time characteristics.I. INTRODUCTION Much previous research has been done on the problem of cooperative path planning. Unfortunately many proposed approaches do not scale to teams of hundreds of robots in constrained environments. One approach which has been shown to be effective for reasonably large teams is prioritized planning [1]. In this approach, robots sequentially plan paths according to a prioritization function. However, this means robots must plan paths in order, resulting in a linear increase in overall planning time with the number of robots. Prioritized planning is also centralized, creating a potential computational and communication bottleneck, as well as a single point of failure.However, in many domains, the strict ordering of sequential planning is likely to be unnecessarily expensive because not all robots need to avoid all other robots. Online prioritized approaches, such as [2] and [3], take advantage of this property by determining which sets of robots need to be planned sequentially by detecting interactions via local observations. In this paper, we describe an approach that distributes prioritized planning, allowing each robot to plan at the same time, then look for collisions between paths and require lower priority robots to replan. The intuition behind this is that if robot paths are not dependent on all other paths, the number of replanning iterations might be quite low and overall planning time will be reduced because no single node needs to plan for each robot in sequence. We prove that distributing the planning in this way still converges to the same result as the centralized planner. Experimental results support the intuition, showing a dramatic reduction in the number of planning iterations. However, because the length

Perception, Planning, Control, and Coordination for Autonomous Vehicles

“…The inherent high complexity hinders these methods from being applicable in real-time multi-vehicle motion planning. Decoupled planning is more efficient, which can be further divided into prioritized planning [315][316][317] and path-velocity planning [318,319]. Since the path of the vehicle typically follows the lane, the centralized multi-vehicle motion planning to explore all the possible paths may be over-kill.…”

Probabilistic Multi Robot Path Planning in Dynamic Environments: A Comparison between A* and DFS

“…Planning for each robot individually make it less complexity than the centralized approaches. The decoupled approaches also categorized into prioritized [13,14,15,16] planning and path coordination. The prioritized path planning was first proposed by [17], where priorities are assigned to each robot either from motion constraints or randomly.…”

“…Decoupled approaches plan the path independently for each robot, then modify these path to prevent the collision that may occur between the robots. Coupled approaches deal with the robots as one composite robot that combines the sum of all the degree of freedom DOF [7]. The decoupled approaches can be either prioritized or coordinated.…”

Towards an Understanding of the Value of Cooperation in Uncertain World

“…Coordination among cooperative agents is a major challenge in MAS (Multi-Agent System). Much work focuses on defining theoretical foundations to guide agent cooperation and coordination, [5], [6]. It is not realized until recently that the cooperation may not increase team performance, or even worsen it under certain circumstances.…”