Autonomous Vehicle Simulation Tuning for Cost-Performance Balance

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Deep learning and simulation for autonomous vehicles (AVs) consume significant resources, limiting their development and optimization.

Innovation Solution

The optimization of deep learning models and simulation for AVs is achieved by modifying parameters such as sensor fidelities and hyperparameters, using reinforcement learning to balance the cost of training and simulation with performance scores.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If larger control models with more layers are used to make better decisions, then AV performance is improved, but resource consumption for training and applying the model increases

Engineering Contradiction:

ImproveAV decision-making performanceVSAvoidresource consumption for deep learning

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent modifies parameters of the control model including number of layers, filter sizes, and other architectural parameters to find an optimal configuration that achieves sufficient performance while reducing computational cost and resource consumption

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If sensors with higher fidelity are used, then quality of sensor data is improved, but resource consumption for simulation increases

Engineering Contradiction:

Improvesensor data qualityVSAvoidresource consumption for simulation

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent modifies sensor fidelity parameters in the simulation environment to find an optimal level that provides sufficient training data quality without excessively increasing simulation computational requirements and resource consumption

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12233887B2Optimization of deep learning and simulation for autonomous vehicles

Publication Date: 2025.02.25 GM CRUISE HOLDINGS LLC

AI summary

An operation of an AV in a scene is simulated. The AV includes sensors detecting the scene and generating sensor data. The sensor data can be input into a control model that outputs control signals, in accordance with which the AV operates. A learning cost, i.e., a cost of training or applying the control model, is determined. A performance of the AV during the simulation is evaluated. A simulation cost, i.e., a cost of running the simulation is determined. The control model can be optimized based on the learning cost and the performance of the AV. Settings of the sensors can be optimized based on the simulation cost and the performance of the AV. The optimization of the control model or the settings of the sensors can be done through reinforcement learning.