Research On Autonomous Collision Avoidance Control For Spacecrafts During Close Operations

With the increasing of complex close operations, the requirements of rapid response, autonomy and reliability for spacecraft are becoming higher and higher, which poses greater challenges to dynamics and control of the space operations. The presented thesis performed researches on collision avoidance control theory and applications for close operation spacecraft. In-depth study of the three key technologies, i.e. relative motion dynamics, collision risk assessment methods and nonlinear control method were carried out to establish a dynamic intelligent collision avoidance control method for spacecraft close operation, which was then used in typical rendezvous missions such as cluster spacecraft close operation and on-orbit servicing operations.The main researches in theoretical aspects are:Firstly, the close relative dynamic model was investigated. Taking the body coordinate system of the control target as the reference system, a coupled relative kinetic model with dual quaternion expression was established. Theoretical analysis proved its consistency with the traditional kinetic equations. Simulation analysis for the prediction accuracy of the dual quaternion expression and its traditional counterpart revealed the feasibility of the dual quaternion based relative motion model for the calculation of the relative status. On basis of this, relative motion error model was established and linearized, transforming the problem of close maneuvering to tracking control.Secondly, the collision risk assessment method was developed. The general process of spacecraft collision risk assessment was analyzed. The applicability of two commonly used risk assessment methods, namely the miss distance method and collision probability model, were discussed. With the definition of dynamic safety domain, the "3C" collision risk assessment model based on system performance parameters was developed. Spacecraft collision risks were estimated using hierarchical evaluating method. Simulation results indicated that "3C" model can effectively reduce the false alarm rate.Then, dynamic gain scheduling control method was studied. The explicit expression of the dynamic gain scheduling(DGS) method was derived and extended for the general multi-input nonlinear systems. Meanwhile, DGS was proved to be consistent with nonlinear dynamic inversion for full state feedback. For particular applications which are difficult to give explicit expression to, numerical calculation based multi-input DGS control law was derived. Two applications, i.e. F16 and the space manipulator validated the effectiveness of DGS for nonlinear systems.Finally, dynamic intelligent collision avoidance control method was proposed. The dynamic intelligent collision avoidance control(DICAC) algorithm, which consists of tracking control, collision risk assessment and collision avoidance control, was presented. Besides DGS, the BP neural network based PD control was developed, in which automatic differentiation method was introduced to address the inaccuracy in solving for gradient, and numerical simulation validated the effectiveness of the method. Based on the analysis of the factors affecting the collision avoidance control, dynamic wall-follow method was proposed, which avoided unreachable problem to the target. Once a local minimum point is reached, optimaztion technology would be adopted to ensure the feasibility of DICAC algorithm. Comparison between the DICAC algorithm and strengthened LQR/APF control algorithm in simulation showed effectiveness, stability and robustness of DICAC for both one module and two modules maneuver.Based on the basic research mentioned above, application work were also studied:Firstly, DICAC was applied to cluster spacecraft. Three typical operations, such as dispersion and assembling, the module insert/eject and rapid reconstruction, were described. Based on DGS, "3C" assessment methods and dynamic wall-following method, DICAC was performed to simulate the missions of concern, which demonstrated the effectiveness of the presented dynamic intelligent autonomous collision avoidance control algorithm.Secondly, the application of DICAC for on-orbit servicing is studied. On-orbit rendezvous tasks were described. The DICAC algorithm, which consists of automatic differentiation method based BP neural network PD control, "3C" safety assessment method, and dynamic wall-following method, was designed for rendezvous operation to validate the feasibility of the presented DICAC algorithm.Finally, experimental verification was performed on the ground experiment testbed for On-Orbit Servicing. The main parameters of the simulator, i.e. moment of inertia, and thrust, were identified. Four task scenarios, i.e. no obstacles, static obstacles, obstacles with unvaried motion in the same direction, and obstacles with unvaried motion in the different direction were studied using the DICAC algorithm. The results indicated that the DICAC is a valid autonomous collision avoidance control algorithm.In conclusion, the presented work formed a collision avoidance control algorithm. The research results would further improve the close operation reliability and safety for spacecraft close operations, and provide the autonomous safe operation and control system design of future spacecraft with an effective method.