An Approach to Cooperative Robot Behavior for Object Transportation

Abstract

This dissertation presents two cooperative object transportation techniques according to the characteristics of objects: passive and active. The passive object is a typical object, which cannot communicate with and detect other robots. The active object, however, has abilities to communicate with robots and can measure the distance from other robots using proximity sensors. Typical areas of research in cooperative object transportation include grasping, pushing, and caging techniques, but these require precise grasping behaviors, iterative motion correction according to the object pose, and the real-time acquisition of the object shape, respectively. For solving these problems, we propose two new object transportation techniques by considering the properties of objects. First, this dissertation presents a multi-agent behavior to cooperatively transport an active object using a sound signal and interactive communication. We first developed a sound localization method, which estimates the sound source from an active object by using three microphone sensors. Next, since the active object cannot be recalled by only a single robot, the robots organized a heterogeneous team by themselves with a pusher, a puller, and a supervisor. This self-organized team succeeded in moving the active object to a goal using the cooperation of its neighboring robots and interactive communication between the object and robots. Second, this dissertation presents a new cooperative passive object transportation technique using cyclic shift motion. The proposed technique does not need to consider the shape or the pose of objects, and equipped tools are also unnecessary for object transportation. Multiple robots create a parallel row formation using a virtual electric dipole field and then push multiple objects into the formation. This parallel row is extended to the goal using cyclic motion by the robots. The above processes are decentralized and activated based on the finite state machine of each robot. Simulations and practical experiments are presented to verify the proposed techniques.