ML Map Compression for Autonomous Vehicle Control Fidelity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data compression techniques indiscriminately compress data files, failing to preserve the performance of critical data portions necessary for predefined tasks, particularly in the context of autonomous vehicle map data.
Innovation Solution
A machine-learning compression model is trained through simulations of a virtual vehicle system to intelligently compress map data, preserving salient features necessary for autonomous vehicle operations by iteratively evaluating and adjusting the compression process to meet performance thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Volume of stationary object
If traditional compression techniques are used to reduce map data size, then storage volume and transfer time are improved, but critical performance information is lost
Solution Approach 1:
The patent applies local quality by differentiating between critical and non-critical portions of map data. The system identifies and preserves essential features (such as road boundaries, intersections, and navigation-critical elements) while applying higher compression to less critical areas. This selective preservation ensures that autonomous vehicle control performance is maintained while achieving effective data compression.
Solution Approach 2:
The patent employs parameter changes by using machine learning models to dynamically adjust compression parameters based on the importance of different map data features. The system learns to allocate different compression levels to different data regions, changing the compression parameters adaptively to preserve performance-critical information while maximizing overall compression efficiency.
2Productivity
If lossy compression is applied to reduce data file size, then storage and transfer efficiency are improved, but data accuracy deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models on large datasets to learn the characteristics of performance-critical map data features before actual compression occurs. The system performs offline training to establish compression strategies that will preserve accuracy for essential features during runtime compression, enabling efficient real-time operation without sacrificing precision.
Solution Approach 2:
The patent implements feedback mechanisms where the system evaluates the impact of compression on autonomous vehicle performance and uses this information to refine compression parameters. The machine learning model learns from performance outcomes and adjusts its compression strategy accordingly, creating a closed-loop system that continuously improves accuracy preservation while maintaining compression efficiency.
Data Source
AI summary
Aspects of the present disclosure provide techniques for training a machine-learning model to compress map data for use online by an autonomous vehicle and techniques for compressing map data using the trained machine-learning model. A system includes a computing device configured to deploy a simulation environment initiating an instance of a virtual vehicle, execute iterations of a simulation of the virtual vehicle, wherein each iteration: deploys a set of map data compressed by the machine-learning compression model and causes the virtual vehicle to execute control operations based on the deployed set of map data, evaluate performance of the executed control operations by the virtual vehicle based on the compressed map data for each iteration, and train the machine-learning compression model to compress map data such that the evaluated performance of the executed control operations by the virtual vehicle exceeds a performance threshold.