Abstract: A real-time re-planning method for satellite formation reconfiguration in deep space is proposed in this paper. First, the Legendre pseudospectral method (LPM) is employed to transform the re-planning problem into an optimization nonlinear programming (NLP) problem from current configuration to the final configuration. Next, according to the feature of satellite formation reconfiguration, a co-evolutionary particle swarm optimization (CPSO) method is used during every re-planning. In this way, all the constraints could be satisfied during the entire optimization, and computational complexity for calculating the gradient information with traditional optimization methods could be avoided. Furthermore, the direction of optimization could be provided in advance, and makes the computation real-time. Then, the re-planning strategy is presented, which could make the reconfiguration work well. Simulation results demonstrate that compared with traditional control methods, this re-planning method could optimize the reconfiguration problem better than the given trajectories, decrease the energy consumption and guarantee collision avoidance during the entire reconfiguration.

Key words: Formation reconfiguration, Collision avoidance, Pseudospectral method, Co-evolutionary particle swarm optimization, Real-time re-planning