Multi-agent control approach for autonomous mobile manipulators: Determination of the best footsteps combination

This paper presents our ongoing efforts toward the development of a distributed multi-agent framework for autonomous control of mobile manipulators. The proposed scheme assigns a reactive agent (Joint agent) to control each articulation of the manipulator, a hybrid agent (Mobile base agent) for the control of the mobile base, and a Supervisory agent to coordinate and synchronize the work of all the precedent agents. Each Control agent implements a Simulation-verification technique, in order to optimize, locally and independently from the other agents, the value of a predefined Objective function (fObj). The paper illustrates, also, the methodology we have followed to determine the best combination among possible footsteps for the Control agents of the system. This combination will be the input, among others, of the procedure seeking the optimal solution bringing the position of the end-effector of the mobile manipulator as close as possible to the imposed Target position. For this aim, different simulation scenarios are described and carried out, with and without considering breakdowns of some articulations of the manipulator or the mobile base. For the evaluation of the obtained solutions and the selection of the best footsteps combination, we have considered two criteria (i) fObj values and (ii) the number of iterations.