Autonomous Driving Model Training With Roadside Resource Sharing

Solution Overview

Problem

Connected and Automated Vehicles (CAVs) face limitations due to high costs and performance constraints related to numerous sensors and computation devices, which hinder the effective implementation of automated driving systems.

Innovation Solution

A systematic intelligent system (SIS) manages and coordinates resources among roadside intelligent units (RIUs) and vehicle intelligent units (VIUs) by optimizing task and computing resource allocation, providing unified data specifications, and enabling pre-trip, en route, and post-trip services through a distribution manager module.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If numerous sensors and computation devices are provided on CAV to improve automated driving performance, then the functional capabilities and reliability are improved, but the capital costs and energy costs increase significantly

Engineering Contradiction:

Improveautomated driving performanceVSAvoidnumber of sensors and computation devices

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The patent merges sensing, computation, and control functions between vehicle-mounted devices and roadside infrastructure. The roadside server consolidates computational tasks for multiple vehicles, while roadside sensors provide environmental data to multiple vehicles, reducing the need for each vehicle to have complete independent systems.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The roadside infrastructure serves multiple vehicles simultaneously with unified data processing and computation services. The same roadside sensors and servers support multiple CAVs, making the infrastructure multi-functional and reducing per-vehicle equipment requirements.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Reliability

If numerous sensors and computation devices are provided on CAV to improve automated driving performance, then the functional capabilities are improved, but the capital costs increase significantly

Engineering Contradiction:

Improveautomated driving performanceVSAvoidcapital costs

Core Design Contradiction:

ReliabilityVSEase of manufacture

Solution Approach 1:

The patent merges sensing, computation, and control functions between vehicle-mounted devices and roadside infrastructure. The roadside server consolidates computational tasks for multiple vehicles, while roadside sensors provide environmental data to multiple vehicles, reducing the need for each vehicle to have complete independent systems.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The roadside infrastructure serves multiple vehicles simultaneously with unified data processing and computation services. The same roadside sensors and servers support multiple CAVs, making the infrastructure multi-functional and reducing per-vehicle equipment requirements.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Adaptability or versatility

If each ADS uses different formats and specifications for primitive data and interfaces to optimize individual system performance, then the adaptability to specific conditions is improved, but the difficulty of coordinating and managing multiple ADS increases

Engineering Contradiction:

Improvesystem customizationVSAvoidcoordination complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent introduces a standardized interface layer between diverse ADS systems and the roadside infrastructure. This intermediary layer translates between different ADS proprietary formats and unified communication protocols, enabling coordination without requiring each system to adopt another's format.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The roadside server provides universal data processing capabilities that handle multiple ADS formats through standardized interfaces, allowing the same infrastructure to serve diverse automated driving systems with different proprietary specifications.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20250296599A1Training and optimization of autonomous driving models

Publication Date: 2025.09.25 CAVH LLC

AI summary

The technology relates to a systematic intelligent system (SIS) configured to train and optimize trip profile models through dynamic distribution of computing resources and model parameters across vehicle intelligent units (VIUs) and roadside intelligent units (RIUs). The SIS comprises a system intelligent unit (SIU) configured to generate and maintain a trip profile model using aggregated historical trip data. Based on this model, the SIU coordinates distributed training, schedules model updates, allocates computing resources, and issues task assignments to VIUs and RIUs provided by multiple automated driving system service providers. The system further supports pre-trip planning, en-route updates, and post-trip feedback to continuously refine training processes and optimize deployment. Communication among SIS components is enabled through unified data interfaces and formats to ensure cross-system coordination, efficient model calibration, and adaptive resource management.