IEEE TRANSACTIONS ON ROBOTICS AND AUTOMATION, VOL. 18, NO. 5, OCTOBER2002 769 Cooperative Transport by Multiple Mobile Robots in Unknown Static Environments Associated With Real-Time Task Assignment Natsuki Miyata, Member, IEEE, Jun Ota, Member, IEEE, Tamio Arai, Member, IEEE, and Hajime Asama, Member, IEEE Abstract—This paper deals with a task-assignment architecture for cooperative transport by multiple mobile robots in an unknown static environment. The architecture should satisfy three features: deal with a variety of tasks in time and space, deal with a large number of tasks compared with the number of robots, and de- cide behavior in real time. The authors propose the following ap- proach: We consider the unit of task (task instance) as the job that should be done in a short time by one robot. Based on envi- ronmental information, task instances are dynamically generated using task templates. The priority of task instances is evaluated dy- namically based on the number of robots and the configuration in the workspace. In addition, we avoid generating too many task in- stances by suppressing object motion. The main part of the archi- tecture consists of two real-time planners: a priority-based task-as- signment planner solved by using a linear programming method, and motion planners based on short-time estimation. The effective- ness of the proposed architecture is verified by a cooperative trans- port simulation in an unknown environment. Index Terms—Cooperative transport, dynamic task domain, multiple mobile robots, real time, task assignment. I. INTRODUCTION A SYSTEM for handling flexible materials is in high de- mand at places such as construction sites and distribution centers. Many researchers (including the authors) have tried to realize such a system by means of the cooperation of multiple mobile robots [1]–[15]. This is because the robots can be dis- tributed freely according to the various types of the objects. The focus of these studies has been on the control and manipula- tion processes. At places such as the above-mentioned ones, however, robots have to detect changes in the geometry of the surroundings on demand. However, the sensing process takes a considerable amount of time because of the limited ability Manuscript received March 12, 2001; revised March 28, 2002. This paper was recommended for publication by Associate Editor L. Parker and Editor S. Hutchinson upon evaluation of the reviewers’ comments. This paper was presented at the IEEE/RSJ International Conference on Intelligent Robots and Systems, Kyongju, Japan, October, 1999, and at the IEEE International Conference on Robotics and Automation, San Francisco, CA, April 2000. N. Miyata is with the Digital Human Laboratory, National Institute of Advanced Industrial Science and Technology (AIST), Tokyo 135–0064, Japan (e-mail: n.miyata@aist.go.jp). J. Ota and T. Arai are with the University of Tokyo, Tokyo 113–8656, Japan (e-mail: ota@prince.pe.u-tokyo.ac.jp; arai@prince.pe.u-tokyo.ac.jp). H. Asama is with the Chemical Engineering Laboratory, Institute of Physical and Chemical Research (RIKEN), Saitama 351–0198, Japan (e-mail: asama@cel.riken.go.jp). Digital Object Identifier 10.1109/TRA.2002.803464 Fig. 1. Cooperative transport in an open environment. of each robot’s sensor. Hence, to cope with unexpected situa- tions and achieve the final goal of the robot group (to trans- port a given object to a certain configuration), it is necessary to plan and execute various tasks including the sensing process as well as the manipulating process (Fig. 1). For example, in order to handle an object, robots must recognize their own posi- tions based on known landmarks. They must then search around their surroundings adequately in order to detect unexpected sit- uations as soon as possible. If a robot finds a movable object like a trashcan or a door in its way, it has to change the object’s path or remove the obstacles. When robots find a moving ob- stacle (e.g., a human), they have to approach it and instruct it to get out of their way. In this paper, these various required actions of robots are called “tasks.” When we focus on a task-assignment architec- ture, there are three parts to consider (Sections I-A.1, I-A.2 and I-A.3). First of all, robots have inadequate information about their working space at the beginning of the whole work. The task will be generated as a result of their action. From the view- point of the “characteristics of a task” in such work, the task assignment planner should have the ability to deal with the fol- lowing two aspects (Section I-A.1 and I-A.2). 1) Dynamic Change of Tasks: There are several kinds of requirements for the robots to achieve their main purpose of moving an object to a certain configuration, and not all of them are always needed. For example, a trashcan in the robots’ way 1042-296X/02$17.00 © 2002 IEEE