Parameter Optimization Using Jitter-Aware Indicator Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing parameter optimization algorithms for autonomous driving face challenges due to unstable indicators and jitter, leading to degraded performance and potential deviation in optimization direction, resulting in inefficiencies and convergence failures.
Innovation Solution
A method that compares two groups of performance indicator values before and after parameter adjustment to distinguish between changes caused by jitter and actual parameter adjustments, optimizing the parameter only when the change is not due to jitter, using hypothesis tests to ensure reliable convergence and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If parameter optimization algorithm is used to adjust parameters based on performance indicator, then parameter optimization function can be implemented, but indicator jitter causes optimization direction to deviate and performance to degrade to random search
Solution Approach 1:
The patent introduces an intermediary mechanism (variance analysis and hypothesis testing) between the performance indicator and the parameter optimization algorithm. This intermediary filters out the jitter noise by comparing whether the indicator change exceeds the expected variance range, thereby preventing false optimization directions while preserving genuine performance changes.
Solution Approach 2:
The patent implements a feedback mechanism where the optimization algorithm continuously monitors the performance indicator, compares it against the pre-calculated variance range, and adjusts parameters only when the change is statistically significant. This feedback loop ensures that optimization decisions are based on reliable signals rather than noise.
2Productivity
If parameter adjustment is made based on unstable indicator, then optimization process can proceed, but optimization effect is submerged in noise and convergence fails
Solution Approach 1:
The patent performs preliminary actions by calculating the variance range of the performance indicator before starting the optimization process. This pre-computed variance information is then used throughout the optimization to filter out noise, enabling the algorithm to proceed efficiently without being misled by indicator jitter.
3Adaptability or versatility
If multiple services share system resources, then system utilization is improved, but mutual impact between services makes parameter optimization complex and coupled
Solution Approach 1:
The patent extracts the variance characteristics of the performance indicator separately for each service, allowing the optimization algorithm to consider the unique noise profile of each service. This extraction approach simplifies the complex coupled system by treating each service's optimization independently while accounting for resource sharing effects through individual variance analysis.
Data Source
Figure 1~2
Figure 3~5
AI summary
A parameter optimization method and apparatus, and an electronic device are provided. The method includes: obtaining a first group of values of a first performance indicator, where the first group of values of the first performance indicator is generated when a first service runs based on a first value of a to-be-adjusted parameter in a first time period, the first performance indicator is associated with performance of the first service, and the first service runs on a first device; determining, based on the first group of values of the first performance indicator and a second group of values of the first performance indicator, that a value change of the first performance indicator is caused by a change of the to-be-adjusted parameter from a second value to the first value, where the second group of values of the first performance indicator is generated when the first service runs based on the second value of the to-be-adjusted parameter in a second time period; and optimizing the to-be-adjusted parameter based on the first value. This solution may be applied to end-to-end performance optimization of an autonomous driving service of an intelligent vehicle, and helps improve parameter optimization efficiency.