Lane-Level AV Routing Using Predicted Lane Performance

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current autonomous vehicle navigation systems lack the ability to effectively select routes based on lane-level performance data, which can lead to inefficient routing and increased risk due to unpredictable traffic patterns and obstacles, as they primarily rely on roadway-level data rather than lane-specific conditions.

Innovation Solution

The system collects and analyzes lane-level performance data from a fleet of autonomous vehicles, using clustering algorithms and machine learning to predict the performance of candidate lanes and select the most suitable route by aggregating data from similar lanes, even when individual lane data is sparse.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If navigation systems use roadway-level data for routing, then the routing system is simpler to implement, but the routing accuracy and safety are reduced due to inability to account for lane-specific conditions

Engineering Contradiction:
Improverouting accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the roadway into individual lanes and evaluates performance metrics at the lane level rather than roadway level. Each lane is assigned a performance score based on autonomous vehicle trip data, allowing the navigation system to select specific lanes for routing decisions. This segmentation enables more precise routing accuracy by accounting for lane-specific conditions such as visibility, traffic patterns, and collision rates.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary layer of lane performance evaluation that bridges the gap between simple roadway-level routing and complex lane-level analysis. The system collects trip data from autonomous vehicles, processes it through machine learning models to generate lane performance scores, and uses these scores to inform routing decisions. This intermediary performance evaluation layer enables accurate lane selection without requiring direct complex control of each lane parameter.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If the system collects and analyzes lane-level performance data from fleet vehicles, then the routing safety and performance prediction accuracy improve, but the data processing complexity and computational requirements increase

Engineering Contradiction:
Improverouting safetyVSAvoiddata processing complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges data from multiple autonomous vehicles in the fleet to establish lane performance metrics. By aggregating trip data across numerous vehicles and lanes, the system creates a comprehensive performance evaluation that improves routing safety through statistically significant patterns. The machine learning models combine multiple data sources including trip outcomes, traffic conditions, and vehicle behaviors to generate reliable lane performance scores.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The system uses data generated by autonomous vehicles during normal operation to continuously improve lane performance metrics without requiring separate dedicated data collection missions. Each trip by an autonomous vehicle contributes to the performance evaluation of the lanes it traverses, allowing the system to self-upgrade its routing capabilities using operational data already being collected for other purposes.

Inventive Principle:
Principle #25Self-service

3Measurement precision

If the system uses machine learning and clustering algorithms to predict lane performance, then the ability to handle sparse data and make accurate predictions improves, but the computational load and processing time increase

Engineering Contradiction:
Improveperformance prediction accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary processing of lane performance data by collecting and analyzing trip outcomes from fleet vehicles to establish baseline performance metrics for each lane before routing decisions are needed. The system pre-computes performance scores and stores them for quick retrieval during navigation. This preliminary action allows the system to make rapid routing decisions without performing complex real-time analysis, reducing processing time while maintaining prediction accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system uses partial action by focusing computational resources on evaluating only the candidate lanes that appear in potential routing options, rather than analyzing all lanes in the service area. The clustering algorithms process data for relevant lanes with higher priority, allowing the system to achieve sufficient prediction accuracy without the excessive computational burden of comprehensive analysis of every possible lane.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4153944B1Routing autonomous vehicles based on lane-level performance
Publication Date: 2024.07.17 GM CRUISE HOLDINGS LLC
  • EP4153944B1 patent drawingFigure 1~2
  • EP4153944B1 patent drawingFigure 3~5
  • EP4153944B1 patent drawingFigure 6

AI summary

A map database stores data describing a set of connected roadways, each having one or more lanes, A navigation system selects a route for an autonomous vehicle (AV) by identifying a set of candidate lanes for routing the AV in the map database and retrieving at least one parameter of each candidate lane from the map database. The navigation system determines a performance metric for each of the candidate lanes, the performance metric for a given candidate lane based on the retrieved at least one parameter of the given candidate lane. The navigation system selects a candidate lane for routing the AV based at least in part on the determined performance metrics.