A distributed T-HTN planning method for multiple spacecraft based on time information heuristics

The improvement of rocket carrying capacity and spacecraft miniaturization technology have promoted the development of multiple spacecraft collaborative detection. For example, observation of multiple constellations, sampling of multiple lunar rovers, and joint exploration of small celestial bodies by multiple instruments. Compared to detecting with a single spacecraft, multiple spacecraft detection can improve the benefit of a single mission and increase the success rate of the mission. However, the current ground-based control mode is difficult to effectively respond because of the sharp increasement in the number of spacecrafts. Therefore, multiple spacecraft need an autonomous planning method, which is the core of achieving autonomous operation of spacecraft, ensuring a long-term safe operation in orbit and reducing spacecraft operations from the ground. A spacecraft contains multiple subsystems with different internal characteristics, and there are constraints between subsystems. Moreover, some coupled constraints such as inter subsystem constraints, inter device constraints, and external environmental constraints, leading to a low efficiency in multi spacecraft collaborative planning. Considering each spacecraft and the subsystems inside as intelligent agents to obtain a hierarchical multi-agent system. Firstly, we define the constraint in planning through the time information of activities, and propose a distributed Timeline Hierarchical Task Network (T-HTN) planning method for multiple spacecraft based on time information heuristics. Secondly, during collaborative solving, priorities should be given to activities with constraints between spacecrafts. A multi-agent asynchronous interaction strategy is designed to reduce the impact of shared activities inside the agent. A hierarchical backtracking conflict feedback mechanism has been further studied, with the time information propagation of lower-level primitive-activities, avoiding the repeated decomposition of upper-level composite-activities and reducing the levels of backtracking required when conflicts occur. Thirdly, a state cost transition graph is constructed. Invalid states and their branches are removed in the graph based on time heuristic values to narrow the searching range of planning and solving, thereby improving the efficiency of multi spacecraft collaborative task planning. Finally, the scene of multi-node lander landing on the surface of asteroid is designed and experiments are conducted, with results showing that our method can obtain planning results more quickly.