Adaptive Parameter Range Updating for Low-Precision Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Adaptive systems with limited ranges for system parameters and intermediary parameters face performance compromises due to range limitations, particularly in applications involving artificial neural networks (ANNs) and proportional-integral-derivative (PID) controllers, where fixed-point or low-precision computing units are used, leading to inferior results.
Innovation Solution
The adaptive system adjusts parameter ranges based on statistical probability distributions derived from historical values or system conditions, allowing for dynamic updating of finite ranges to minimize errors and improve performance, specifically by assigning probability distributions to parameters and intermediary variables, and setting ranges associated with confidence levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If fixed-point or low-precision computing units are used to implement adaptive systems, then device complexity and power consumption are reduced, but measurement precision and manufacturing precision deteriorate
Solution Approach 1:
The patent implements dynamic range adjustment where the finite range of fixed-point parameters is not static but adapts during the adaptation process. The system monitors the actual ranges required by parameters and intermediary parameters, then dynamically adjusts the finite ranges to accommodate them, resolving the contradiction between using fixed-point arithmetic and maintaining precision.
Solution Approach 2:
The system changes the parameter of finite range dynamically during operation. Instead of using a fixed finite range for fixed-point parameters, the system adjusts the range limits based on the actual needs of the adaptation process, allowing fixed-point computing to achieve precision comparable to floating-point systems while maintaining the benefits of reduced complexity.
2Ease of manufacture
If fixed finite ranges are imposed on system parameters and intermediary parameters, then device implementation becomes simpler, but system performance deteriorates due to clipping and low precision
Solution Approach 1:
The patent transforms static finite ranges into dynamic adjustable ranges. The system continuously monitors the ranges required by parameters and intermediary parameters during adaptation, and adjusts the finite ranges accordingly. This dynamic approach maintains simple fixed-point implementation while preventing clipping and precision loss that would occur with fixed ranges.
Solution Approach 2:
The system performs self-adjustment of parameter ranges during the adaptation process. By monitoring the actual ranges needed and automatically adjusting the finite ranges, the system eliminates the need for conservative pre-setting of ranges, thereby improving performance without adding external control complexity.
3Ease of manufacture
If conservative finite ranges are pre-set for parameters, then implementation becomes easier, but errors caused by clipping and low precision increase
Solution Approach 1:
The patent replaces pre-set conservative finite ranges with dynamically adjusted ranges that adapt during the adaptation process. The system monitors the actual ranges required by parameters and intermediary parameters, then adjusts the finite ranges to match these requirements, eliminating clipping errors and precision loss while maintaining implementation simplicity.
Solution Approach 2:
The system performs preliminary monitoring of parameter ranges during the adaptation process, then adjusts the finite ranges based on this information. This preliminary observation allows the system to set appropriate ranges without being overly conservative, thereby improving precision while maintaining ease of implementation.
Data Source
AI summary
An adaptive system and a method for adjusting a parameter in the adaptive system includes operating an adaptive system with an output signal produced from an input signal applied to an input, in which a parameter with a finite range is determined based on a difference between the output signal and a target output signal. In one example the parameter with the finite range is a fixed-point parameter or an analog parameter. The parameter is accessed from the adaptive system. A probability distribution of the parameter is assigned. The finite range for the parameter is updated based on the probability distribution which has been assigned. The probability distribution function may be updated alone with the finite range of the parameter. The probability distribution may be derived from one or more historical values of the parameter, and a plurality of system parameters belonging to an identical category of data as the parameter.