Two-level structure swarm formation system with self-organized topology network

In this work, a two-level mobile robot swarm system with self-organized formation network is proposed. Initially, based on the position information of the robots, a relation-invariable persistent formation (RIPF) algorithm can automatically organize the swarm network and construct an optimal persistent formation. At the upper formation planning level, the collision-free reference paths of the swarm can be planned for guiding the robots to reach and maintain a desired distance with their neighbors. Then, at the lower formation tracking control level, a neural-dynamic combined model predictive control (MPC) method is applied to drive the swarm moving on the reference paths. The MPC can reformulate the system into a convex minimization problem, which can further be transformed into a constrained quadratic programming (QP) problem such that an efficient QP solver, called primal-dual neural network (PDNN), is implemented to obtain the optimal control inputs online for the robots. In the end, simulation results show the effectiveness of the proposed formation system.