A Method of Cooperative Pushing Task using Estimation of Unknown Environment by Measuring Overweight

In this research we have shown a method of estimating contacting status between an object and unknown obstacles by force sensors while multiple cooperative robots are pushing a box. This method guarantees the robots to complete pushing task by local avoidance, which is generated when one of robots senses an excessive load. Generally, the failure is caused by change in working environment, accumulation of motion error of the robots, and so on. Estimation of working environment in a quantitative analysis should be utilized for controlling robots according to some kinds of changes. Now we consider to moving a box-typed object to a destination point by pushing task as shown in Figure 1. Each robot can measure the load to the robot itself by a force sensor. The system can estimate the gravitational/external force and applied point by force sensing data and position/posture data recognized by the robots. The robot system keeps on a pushing task while the load to each robot is within a load limit. Encountering the unknown obstacles is recognized by one robot's sensing excessive load. Contacting status can be given as the by quantitative data (Figure 2). An avoiding path is created as the shortest path that avoids colliding a region where an estimated obstacle exists. The robot system we used in our experiment is shown in Figure 3. An end effector with a force sensor is equipped with each robot. We have shown the experiment of estimating and avoiding an unknown obstacle by two robots (Figure 4).

Keywords: Cooperative Sensing, Mobile Robots, and Estimation


  1. J.Sasaki, J.Ota, Y.Aiyama, T.Arai: "Optimal Arrangement for Handling Unknown Objects by Cooperative Mobile Robots", Proc. of the 1996 IEEE/RSJ International Conference on Intelligent Robot and Systems, pp. 112-117, 1996
Fig. 1 Problem Settlement     Fig. 2 Estimating Contact Point
Fig. 3 Experimental System                                  Fig. 4 Result of Avoidance