Development of a Motion Planner based on Nonlinear Optimization for Free-Floating Robots

Abstract

While the launch of spacecraft and satellites is increasing rapidly, the accumulation of debris in space creates an obstacle for future space missions. Free-floating robots are used to tackle the issue of grasping the tumbling, non-cooperative space debris and stabilizing them. Free-floating robots are manipulator arms attached to the base of an unactuated satellite. The core problem is calculating the robot's trajectory for grasping as the debris has unknown parameters (momentum, pose, et cetera) within a specific time window. The use of an optimization-based motion planner is considered in this research as it provides the criterion for robustness and stability. Trajectory planning is formulated as an Optimal Control Problem with various motion constraints and a cost function that has to be minimized. The drawback of this technique is the time consumption. Parallelization techniques, using multi-core CPU and GPU, will be analyzed to reduce the time. Comparison between different techniques for solution of the equation of motion for the non-holonomic system, optimization methodologies will also be carried out. The methodology undertaken is as follows: The Optimal Control problem is analyzed, and parts that could parallelize within the Optimal Control Framework will be determined. OpenMP-based parallel functions for CPU processing and CUDA-based parallel functions for GPU processing are to be developed. The speedup gain, if any, for the developed parallel functions will be examined. Different optimization methodologies will be implemented and compared. The trajectory optimization problem will be drawn, with new task space. The parallel routines developed, which provide a significant improvement over the state-of-the-art techniques, will be implemented for the motion planning problem. The statistical analysis of the results from the motion planner will be carried out.