The study of multi-agent capture and manipulation of an object has been an area of active interest for many researchers. This paper presents a novel approach using Genetic Algorithm to determine the optimal contact points and the total number of agents (mobile robots) required to capture a stationary generic 2D polygonal object. After the goal points are determined the agents then reach their respective goals using a decentralized projective path planning algorithm. Form closure of the object is obtained using the concept of accessibility angle. The object boundary is first expanded and the robots reach the expanded object goal points and then converge on the actual object. This ensures that the agents reach the actual goal points at the same time and have the correct orientation. Frictionless point contact between the object and robots is assumed. The shape of the robot is considered a circle such that it can only apply force in outward radial direction from its center and along the normal to the object boundary at the contact point. Simulations results are presented that prove the effectiveness of the proposed method.
Computation of fingertip positions for a form-closure grasp
