Autonomous Vehicle Simulation Database Using Public Data

Solution Overview

Problem

Conventional methods for training autonomous vehicles are costly and inefficient, relying on direct data collection by proprietary fleets, which is time-consuming and expensive, and do not effectively utilize publicly available online resources for data.

Innovation Solution

A mechanism is developed to create a World Model Database by using publicly available online resources, such as videos and images, to simulate real-world scenarios, allowing for the creation of a neutral file format, compression, and storage, which reduces data collection time and costs, and enables faster algorithm development and validation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If proprietary fleets are used to collect training data directly from the real world, then the quality and authenticity of training data is improved, but the time and cost required for data collection increases significantly

Engineering Contradiction:

Improvequality of training dataVSAvoiddata collection time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent creates synthetic copies of real-world driving scenarios through simulation environments. Instead of collecting data directly from physical vehicles, the system generates virtual training data by copying real-world road geometries, traffic patterns, and environmental conditions into a digital simulation space, thereby eliminating the time-consuming data collection process while maintaining data quality

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The system performs preliminary actions by pre-building detailed digital twins of real-world environments, including road networks, intersections, and traffic patterns, before actual training begins. This preparatory work allows the simulation to generate unlimited training scenarios without requiring additional real-world data collection expeditions

Inventive Principle:

Principle #10Preliminary action

2Reliability

If proprietary fleets are used to collect training data directly from the real world, then the authenticity of training scenarios is improved, but the cost of data collection increases significantly

Engineering Contradiction:

Improveauthenticity of training scenariosVSAvoiddata collection cost

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The patent creates synthetic copies of real-world driving scenarios through simulation environments. Instead of collecting data directly from physical vehicles, the system generates virtual training data by copying real-world road geometries, traffic patterns, and environmental conditions into a digital simulation space, thereby eliminating the time-consuming data collection process while maintaining data quality

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The simulation system is self-sufficient in generating training data without requiring expensive proprietary vehicle fleets. The virtual environment automatically generates diverse driving scenarios, including edge cases and rare events, that would be difficult and costly to capture in the real world, making the training process independent of physical data collection infrastructure

Inventive Principle:

Principle #25Self-service

3Quantity of substance

If vast amounts of publicly available online data are utilized, then the quantity of training data is improved, but the relevance and quality control of the data becomes more difficult

Engineering Contradiction:

Improvequantity of training dataVSAvoiddata quality control

Core Design Contradiction:

Quantity of substanceVSManufacturing precision

Solution Approach 1:

The simulation environment acts as an intermediary layer between raw public data and the training algorithm. Publicly available data such as satellite imagery, map data, and traffic statistics are fed into the simulation, which then processes this information to generate standardized, high-quality training scenarios with consistent formatting and annotated ground truth, thereby maintaining quality control while leveraging large data quantities

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent creates synthetic copies of real-world driving scenarios through simulation environments. Instead of collecting data directly from physical vehicles, the system generates virtual training data by copying real-world road geometries, traffic patterns, and environmental conditions into a digital simulation space, thereby eliminating the time-consuming data collection process while maintaining data quality

Inventive Principle:

Principle #26Copying

Data Source

PatentUS11584390B2System and method for simulation of autonomous vehicles

Publication Date: 2023.02.21 INTEL CORP

AI summary

Various systems and methods for creating and managing a simulation database system are described herein. A system for creating and managing a simulation database system includes a processor subsystem; and memory including instructions, which when executed by the processor subsystem, cause the processor subsystem to: access an operating scenario, the operating scenario including parameters defining an environment for a simulated autonomous vehicle; access a rule set, the rule set controlling a search mechanism; search for content, the search controlled by the rule set and operable to find content relevant to the operating scenario, the search resulting in search results; process the search results for training autonomous vehicle operation; and store the processed search results.