Modeling and analysis of dynamic multi-agent planar manipulation

A dynamic model for multi-agent manipulation is investigated under a nonlinear control framework. The motion of the rigid body on a plane under two-finger pushing is modeled as a nonlinear system. The local controllability is derived for different cooperation patterns. The motion of the rigid body under pushing is composed of two stages: during the first stage, the fingers maintain contact with the object, and through adjusting the pushing position, orientation and force on the object, the configuration of the object can reach some subsets in configuration space. Due to local contact constraints, the fingers may lose contact with the object where the second stage of motion starts. Here, the motion of the object is governed by the friction force only. Thus manipulation planning consists of two parts. First, an initial state response problem needs to be solved to find the subset of configuration space which can reach the final configuration under the friction-governed sliding. The remaining problem can be categorized as optimal control problem, which solves the cooperation between agents which force the object moving into the subset found by the initial state response problem. The possibility of cooperation is illustrated through an example.