Planning for In-Space Robotic Assembly of Modular CubeSats

In-space manufacturing has the potential to revolutionize the construction of structures in space, bypassing launch constraints and significantly accelerating deployment response times. This paper presents an assembly planner for modular CubeSats as part of the “Orbital Locker” project at the Mas-sachusetts Institute of Technology. The concept of operations involves a free-flying satellite that acts as a storage “locker”, carrying modular CubeSat components inside it and having the ability to assemble and deploy them. Orbital Locker is an initial small-scale demonstration that is intended to be scaled up. The CubeSat modules are dimensioned such that three modules stack to form a IV CubeSat. During launch, the modules are secured in storage stacks within the Locker. They are then assembled on demand into multi-module CubeSats and deployed from the Locker on orbit by a Cartesian gantry robot. The 50 cm x 25 cm x 25 cm prototype Locker is capable of assembling up to three IU CubeSats. This paper presents the design of the assembly planner which determines the assembly sequence used by the Locker to as-semble a requested CubeSat. The assembly planner consists of a global planner for high level symbolic planning and a local planner for low level motion and path planning. The local planner ensures that the gantry robot moves the satellite modules between the storage stacks and the assembly stack efficiently and without collisions, while the global planner determines the optimal sequence for moving different modules between stacks in order to correctly assemble the desired satellite. The global planner creates a graph representation of possible assembly configurations, with each graph edge between states corresponding to the movement of one module between stacks while satisfying the physical constraints of the Orbital Locker layout. It searches for the fastest sequence of module movements to assemble the desired multi-module satellite using a version of Dijkstra's algorithm. After discussing the planner, this paper provides examples of optimal assembly sequences generated by the global planner for the prototype Locker, and discusses the performance of the global planner for future scaled-up systems containing more stacks and modules. Autonomous assembly as demonstrated by the Orbital Locker's assembly planner lays the groundwork for a system that can deploy custom-configured satellites on orbit in minutes, rather than weeks or months.