Defend the practicality of single-integrator models in multi-robot coordination control

Abstract

Single-integrator models have been widely used to model robot kinematics in multi-robot coordination control problems. However, it is also widely believed that this model is too simple to lead to practically useful control laws. In this paper, we prove that if a gradient-descent distributed control law designed for single integrators has been proved to be convergent for a given coordination task, then the control law can be readily modified to adapt for various motion constraints including velocity saturation, obstacle avoidance, and nonholonomic models. This result is valid for a wide range of coordination tasks. It defends the practical usefulness of many existing coordination control laws designed based on single-integrator models and suggests a new methodology to design coordination control laws subject motion constraints.