Context-Specific Driving Models via Distributed Vehicle Learning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing AI or ML models for autonomous driving face challenges in training due to privacy concerns, bandwidth costs, and the lack of contextual data, which affects their accuracy and efficiency in various operating conditions.

Innovation Solution

A distributed learning framework is implemented where vehicles share context-specific training data with roadside units (RSUs) or nearby vehicles, allowing for localized model generation and aggregation, preserving privacy and optimizing bandwidth usage.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If centralized training with large amounts of sensory data is used, then model accuracy is improved, but privacy protection deteriorates and bandwidth costs increase

Engineering Contradiction:
Improvemodel accuracyVSAvoidprivacy risk
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent segments the centralized training process into distributed local training at multiple vehicles. Each vehicle trains its own model locally using its sensory data, preventing centralization of sensitive data while still achieving collaborative learning through model aggregation at the roadside unit.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The roadside unit acts as an intermediary that aggregates models from multiple vehicles without directly handling raw sensory data. This intermediary structure enables collaborative training while maintaining data privacy at the source vehicles.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Adaptability or versatility

If contextual data is transported for training, then model adaptability to different contexts is improved, but bandwidth usage increases and privacy issues arise

Engineering Contradiction:
Improvecontext-specific accuracyVSAvoidbandwidth consumption
Core Design Contradiction:
Adaptability or versatilityVSLoss of energy

Solution Approach 1:

The patent extracts only the essential model parameters and gradients needed for training from the vehicles, rather than transmitting complete contextual sensory data. This extraction approach maintains context-specific adaptability while significantly reducing bandwidth consumption.

Inventive Principle:
Principle #2Taking out (Extraction)

3Reliability

If more training data is collected from multiple contexts, then model reliability is improved, but data privacy protection becomes more difficult

Engineering Contradiction:
Improvemodel reliabilityVSAvoidprivacy exposure
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent implements local quality by keeping training data processing at the local vehicle level where privacy is naturally protected. Each vehicle processes its own contextual data locally to generate model updates, ensuring that privacy protection is maintained at the source while still achieving diverse contextual training.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4021027B1Distributed learning to learn context-specific driving patterns
Publication Date: 2025.08.20 INTEL CORP
  • EP4021027B1 patent drawingFigure 1
  • EP4021027B1 patent drawingFigure 2
  • EP4021027B1 patent drawingFigure 3

AI summary

A method of implementing distributed AI or ML learning for autonomous vehicles is disclosed. An AI or ML model specific to a location or a type of the location is generated at a vehicle. In response to a detection that the vehicle is within a proximity to a road side unit (RSU) associated with the location or the type of the location or the vehicle is within a proximity to an additional vehicle that is present or anticipated to be present at the location or the type of the location, causing an AI or ML model transmission to the additional vehicle or the RSU. Based on the causing the AI or ML model reception, causing deployment of an additional AI or ML model in the vehicle to optimize the vehicle for the location or the type of the location.