Robot Control System Using Segmented Modules for Real-Time Stability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing control systems for robots with multiple degrees of freedom face challenges in real-time control due to the combination of continuous and discrete state variables, leading to inefficient search algorithms and potential communication delays, which can result in suboptimal motion control and instability, especially in complex tasks like bipedal walking.
Innovation Solution
A control system comprising a high-frequency internal module and a low-frequency external module, where both modules search for action plans within defined search ranges based on measured state values, with the internal module prioritizing fast computation and the external module compensating for delays to ensure optimal behavior under specified objectives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a high-speed arithmetic processor is used to speed up optimization computation, then the derivation speed of solutions improves, but communication delay occurs when using external processors
Solution Approach 1:
The control system is segmented into two distinct modules: an internal high-speed processor for real-time control computations and an external arithmetic processor for complex optimization problems. This segmentation allows each module to operate independently at its optimal speed without being constrained by communication delays between them.
Solution Approach 2:
The internal high-speed processor performs preliminary control actions and maintains real-time control based on current state values. Complex optimization computations are performed in advance by the external processor when time permits, and results are stored for the internal processor to execute immediately when needed, eliminating the need for real-time communication during critical control moments.
2Reliability
If full search or random search algorithms are used to solve optimization problems with continuous and discrete variables, then solution completeness improves, but computation time increases significantly
Solution Approach 1:
The search space is segmented into continuous variable dimensions and discrete variable dimensions. Different search strategies are applied to each segment: efficient gradient-based methods for continuous variables and targeted discrete search for discrete variables, avoiding the need for exhaustive full search while maintaining solution completeness.
Solution Approach 2:
The optimization approach changes parameters by transforming the discrete-continuous optimization problem into a series of sub-problems. Continuous variables are optimized first to reduce the search space, then discrete variables are optimized based on the continuous variable results, significantly reducing computation time while maintaining solution quality.
3Manufacturing precision
If the control cycle is extended to allow more computation time, then solution quality improves, but robot stability deteriorates due to delayed response
Solution Approach 1:
The control architecture segments computations into real-time critical tasks handled by the internal processor and non-critical optimization tasks handled by the external processor. This ensures that stability-critical control actions are executed within strict time limits while allowing longer computation cycles for quality optimization of non-critical parameters.
Solution Approach 2:
A multi-level feedback mechanism is implemented where the internal processor provides immediate feedback for stability control, and the external processor provides periodic feedback for optimization. This layered feedback ensures that robot stability is maintained through fast internal loops while solution quality improves through slower external optimization loops.
Data Source
AI summary
Provided is a control system and the like capable of deriving at high speed a solution to the optimization problem of combinations of continuous state variable and discrete state variables. According to the control system, by setting a search range (first search range) of internal action candidates ai1 for an internal module mod1 smaller than a search range (second search range) of external action candidates ai2 for a low-frequency external module mod2, the arithmetic computing speed is accelerated accordingly. Thereby, when it is necessary for a robot R to cope with a disturbance emergently on the basis of measured state values of the robot R, the operation of robot R can be controlled according to the arithmetic computing result from the high-frequency internal module mod1 without waiting for the arithmetic computing result form the low-frequency external module mod2.