Control Systems

 Piano Mover's Problem: Problem Solving Algorithms for Autonomous Robotics 

The Piano Mover's problem involves planning, understood as a class of algorithms in the field of AI and control theory. In addition to sensing, servo control, etc., planning is about translating high-level task specifications to a low-level sub-task solution through a chain of actions. Many other considerations make autonomy hard, e.g., differential constraints, uncertainty, and optimality, to name only a few. Trajectory planning usually refers to the problem of obtaining the solution from a robot motion planning algorithm and determining how to move along the path in a way that respects the mechanical limitations of the robot and the cargo. This excellent paper discusses cylindrical algebraic decomposition (CAD). This particular planning version focuses on autonomous solving the path planning problems for thin, elongated objects moving through narrow corridors. We can think of so many applications in the off-world situation. Currently, I'm adding this implementation based on the cylindrical decomposition to the library of solutions. The architecture is similar to typical machine learning enterprise solutions, where a "tournament" of algorithms is applied to solve a problem.

Similarly, in path planning, competing algorithms (or parameter setup) work best in specific circumstances. However, the configuration space of the meta-level algorithm for unknown conditions on other planets is a bit of a challenge. Therefore, an ensemble or "tournament" approach is probably the way to go until a significant training and test data set can be obtained.   

https://arxiv.org/pdf/1309.1588.pdf #controlsystems #controltheory # #machinelearning #ai #robotics #planning #cylindricalalgebraicdecomposision #algorithms