Lane Change Intention Prediction Network for Autonomous Route Updates

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing vehicle lane change intention prediction methods are inaccurate and rigid, failing to consider sufficient surrounding vehicle information, leading to suboptimal autonomous driving decisions and increased traffic accident risk.

Innovation Solution

A lane change intention prediction network with multiple sub-networks processes vehicle state, interaction, and road network information in parallel, performing feature extraction and fusion to accurately predict lane change intentions in various scenarios without excessive preset rules.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional lane change prediction methods are used, then the system is simple to implement, but the prediction accuracy is low and cannot consider sufficient surrounding vehicle information

Engineering Contradiction:
Improvelane change intention prediction accuracyVSAvoidprediction network structure complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The prediction network is divided into multiple independent sub-networks, each responsible for processing specific types of information (vehicle state, interaction, road network). This segmentation allows each sub-network to specialize in extracting features from particular data sources, improving overall prediction accuracy while maintaining manageable complexity through modular design

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The multi-sub-network architecture serves multiple functions simultaneously: it processes diverse information types (vehicle dynamics, spatial relationships, road context), handles different feature extraction tasks, and integrates various data sources. This multi-functionality enables comprehensive consideration of surrounding vehicle information without requiring separate systems for each task

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Measurement precision

If multiple types of vehicle information are processed in parallel, then the prediction accuracy is improved, but the computational complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidcomputational power consumption
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

By segmenting the computational workload across multiple specialized sub-networks processing different information types in parallel, the system can optimize resource allocation for each sub-network. This reduces redundant computations and enables more efficient use of computational power compared to a single monolithic network processing all information sequentially or in a unified manner

Inventive Principle:
Principle #1Segmentation

3Adaptability or versatility

If excessive preset rules are used, then the system can handle various scenarios, but the system becomes rigid and loses adaptability

Engineering Contradiction:
Improvescenario handling capabilityVSAvoidsystem flexibility
Core Design Contradiction:
Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The system replaces traditional mechanical rule-based decision-making with an intelligent neural network model. Instead of relying on excessive preset rules that make the system rigid, the neural network learns adaptive patterns from data, enabling the system to handle various scenarios with flexibility and adaptability while maintaining stability through the learned representations

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP4009300B1Vehicle automatic control method and lane change intention prediction network training method
Publication Date: 2026.01.21 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP4009300B1 patent drawingFigure 1~2
  • EP4009300B1 patent drawingFigure 3
  • EP4009300B1 patent drawingFigure 4~5

AI summary

A vehicle automatic control method, a lane change intention prediction network training method, a vehicle automatic control apparatus, and a lane change intention prediction network training apparatus, and an electronic device. The method comprises: receiving multiple types of vehicle traveling information of a target vehicle (702, 703, 704) (S800); inputting the multiple types of vehicle traveling information of the target vehicle (702, 703, 704) into a lane change intention prediction network, wherein the lane change intention prediction network comprises multiple subnetworks in one-to-one correspondence with the multiple types, and the lane change intention prediction network is used for predicting a lane change intention of a vehicle in a traveling state (S810); respectively performing feature extraction on each vehicle traveling information by means of each subnetwork, and outputting a feature extraction result (S820); performing feature fusion on the feature extraction results output by the subnetworks, and predicting a lane change intention of the target vehicle (702, 703, 704) according to a feature fusion result (S830); and updating an autonomous driving route of the current vehicle (701) according to the lane change intention of the target vehicle (702, 703, 704) (S840).