Multi-Axis Neural Control With Precision-Weighted Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing control systems for multiple drive axes often fail to meet precision requirements for all axes simultaneously, leading to excessive specifications and increased calculation costs due to the need to design systems based on the strictest axis, resulting in inefficient performance.
Innovation Solution
A management apparatus utilizing a neural network with parameters decided through reinforcement learning, which evaluates rewards for each drive axis based on their specific precision requirements, allowing for relative adjustment of rewards to efficiently meet precision demands across all axes at a limited calculation cost.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If feedback control is performed by designing control systems in conformity with the drive axis having the strictest precision requirement, then all drive axes can satisfy their precision requirements, but excessive specifications are generated for drive axes with less strict requirements, resulting in increased calculation costs
Solution Approach 1:
The patent applies local quality by assigning different weight coefficients to different drive axes in the reinforcement learning reward function. Each drive axis receives a weight coefficient corresponding to its specific precision requirements, allowing the control system to optimize each axis independently according to its local precision needs rather than applying uniform strict specifications to all axes. This resolves the contradiction by enabling precise control where needed while reducing calculation burden where less precision is required.
2Reliability
If uniform control specifications are applied to all drive axes based on the strictest requirement, then consistent control quality is achieved across all axes, but calculation efficiency decreases due to excessive computation for axes with lower precision demands
Solution Approach 1:
The patent employs parameter changes by dynamically adjusting the weight coefficients in the reward function based on the specific precision requirements of each drive axis. Instead of using uniform control specifications, the system modifies the reward parameters to reflect the varying precision needs of different axes, thereby maintaining reliable control quality where required while improving calculation efficiency by reducing unnecessary computational constraints on axes with lower precision demands.
Data Source
AI summary
Provided is a management apparatus for managing a processing apparatus including a driver configured to drive a target object in regard to a plurality of drive axes, and a controller configured to control the driver using a neural network for which a parameter for outputting a manipulated variable to the target object is decided by reinforcement learning. The management apparatus includes a learning unit configured to decide the parameter of the neural network by reinforcement learning. The learning unit performs the reinforcement learning by evaluating a reward obtained from a control result of the target object by the controller, and relatively adjusts rewards regarding the respective drive axes in accordance with required precisions for the respective drive axes.