Lane Selection Decision Model for Autonomous Vehicles

Solution Overview

Problem

Current lane change selection mechanisms in unmanned driving scenarios are inefficient, particularly in separating discretionary lane change (DLC) and mandatory lane change (MLC), leading to suboptimal decision results and prolonged response times, which are not suitable for the stringent requirements of unmanned driving.

Innovation Solution

A method and apparatus for lane selection that utilize decision models to calculate utility values for candidate lanes based on lane change times, distance to junctions, speed information, and road network conditions, integrating DLC and MLC models to ensure precise and timely lane changes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If the judgment mechanism separates DLC and MLC analysis, then the mechanism can simulate driver driving motion, but the lane change selection precision deteriorates and response time increases

Engineering Contradiction:

ImproveAbility to simulate driver driving motionVSAvoidLane change selection precision

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent segments the lane change selection process into two distinct phases: MLC judgment phase (where junction conditions are evaluated first) and DLC selection phase (where optimal lane is selected from candidates). This segmentation allows the system to simulate driver behavior by prioritizing safety-critical junction decisions while maintaining efficient lane change selection through subsequent optimization based on speed and traffic conditions.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary filtering of candidate lanes by first evaluating MLC conditions at junctions before proceeding to DLC optimization. This preliminary action eliminates unsuitable lanes early in the decision process, reducing the search space for subsequent optimization and improving overall response time while maintaining selection precision.

Inventive Principle:

Principle #10Preliminary action

2Adaptability or versatility

If the judgment mechanism separates DLC and MLC analysis, then the mechanism can simulate driver driving motion, but the response time increases

Engineering Contradiction:

ImproveAbility to simulate driver driving motionVSAvoidResponse time

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent segments the lane change selection process into two distinct phases: MLC judgment phase (where junction conditions are evaluated first) and DLC selection phase (where optimal lane is selected from candidates). This segmentation allows the system to simulate driver behavior by prioritizing safety-critical junction decisions while maintaining efficient lane change selection through subsequent optimization based on speed and traffic conditions.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary filtering of candidate lanes by first evaluating MLC conditions at junctions before proceeding to DLC optimization. This preliminary action eliminates unsuitable lanes early in the decision process, reducing the search space for subsequent optimization and improving overall response time while maintaining selection precision.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If comprehensive lane change selection is implemented considering all conditions, then lane change selection accuracy improves, but the decision model complexity increases

Engineering Contradiction:

ImproveLane change selection accuracyVSAvoidDecision model complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the lane change selection process into two distinct phases: MLC judgment phase (where junction conditions are evaluated first) and DLC selection phase (where optimal lane is selected from candidates). This segmentation allows the system to simulate driver behavior by prioritizing safety-critical junction decisions while maintaining efficient lane change selection through subsequent optimization based on speed and traffic conditions.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies different evaluation criteria to different stages of the decision process: MLC conditions (junction safety, mandatory lane changes) are evaluated with high priority in the first phase, while DLC parameters (speed optimization, traffic flow) are optimized in the second phase. This local quality approach ensures comprehensive consideration of all conditions while managing model complexity through staged processing.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS11059485B2Lane selection method, target vehicle and computer storage medium

Publication Date: 2021.07.13 TENCENT TECHNOLOGY (SHENZHEN) CO LTD

AI summary

A method and an apparatus for lane selection are provided. In the disclosed method that is implemented by the apparatus, a decision model used for a lane change selection is obtained based on a first model configured to decide a junction lane change and a second model configured to decide a travelling speed. In addition, travelling information of a target vehicle and target information related to the target vehicle is acquired in real time. The target information is configured to represent travelling information of one or more vehicles around the target vehicle. A target lane is subsequently defined based on the decision model, the target information related to the target vehicle and the travelling information of the target vehicle that is acquired in real time.