Camera-Lidar Navigation Fusion for Robust Autonomous Vehicle Guidance

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing end-to-end machine learning algorithms for autonomous driving struggle with machine learning algorithm confusion, particularly in dynamic environments with changing conditions such as weather and unexpected obstacles, and fail to generalize well to new environments.

Innovation Solution

Integrate a lidar system with a camera system to generate navigational commands using a trained artificial neural network that combines camera images with lidar pointclouds, allowing the network to dynamically adapt routes and avoid obstacles based on real-time environmental data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If end-to-end machine learning algorithms are used for autonomous driving, then the system can directly map sensory input to driving commands, but the algorithm fails to generalize to new environments and changing conditions

Engineering Contradiction:
Improveautonomous driving capabilityVSAvoidgeneralization to new environments
Core Design Contradiction:
Extent of automationVSAdaptability or versatility

Solution Approach 1:

The patent segments the autonomous driving system into multiple independent perception algorithms (object detection, lane detection, mapping) rather than using a single end-to-end model. Each algorithm is trained separately on specific tasks, allowing the system to maintain specialized expertise for different environmental aspects while working together through rule-based integration.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system dynamically adapts by combining outputs from multiple perception algorithms with rule-based decision making. The rule-based module can adjust behavior based on contextual understanding of navigational commands and environmental conditions, allowing the system to generalize to new situations beyond training data.

Inventive Principle:
Principle #15Dynamics

2Measurement precision

If multiple independent perception algorithms are used for different tasks, then the system can handle specific detection tasks well, but the complexity of combining outputs increases significantly

Engineering Contradiction:
Improveobject detection accuracyVSAvoidsystem integration complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces a rule-based decision module as an intermediary between the multiple perception algorithms and the driving command generation. This intermediary layer processes the outputs from object detection, lane detection, and mapping algorithms according to predefined rules derived from navigational commands, simplifying the integration complexity while maintaining detection precision.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Reliability

If the algorithm is trained on specific environments, then it performs well in those environments, but it fails to detect and avoid unexpected temporary blockages like work zones

Engineering Contradiction:
Improveperformance in training environmentVSAvoiddetection of unexpected obstacles
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary detection using multiple perception algorithms to identify potential obstacles and environmental features before making driving decisions. By continuously monitoring the environment with object detection, lane detection, and mapping algorithms, the system can detect unexpected temporary blockages like work zones in advance and plan appropriate responses based on navigational commands.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements feedback loops where the outputs from multiple perception algorithms are continuously fed into the rule-based decision module, which adjusts driving commands based on current environmental conditions. This feedback mechanism allows the system to adapt to unexpected obstacles by comparing detected features against navigational commands and adjusting behavior accordingly.

Inventive Principle:
Principle #23Feedback

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

Enhances the reliability and generalization of autonomous driving by improving the network's ability to handle changing environmental conditions and unexpected obstacles, ensuring safe and efficient navigation.

Implementation Method 1

a camera image representing an environment of the motor vehicle is generated by a camera system of the motor vehicle

Methodology Applied
Scientific EffectLight reflection: Reflection

Implementation Method 2

A lidar pointcloud representing the environment is generated by a lidar system of the motor vehicle

Methodology Applied
Scientific EffectLight emission and detection: LIDAR

Data Source

PatentEP4012338B1Guiding a motor vehicle based on a camera image and a navigational command
Publication Date: 2026.01.28 VALEO SCHALTER & SENSOREN GMBH
  • EP4012338B1 patent drawingFigure 1~2
  • EP4012338B1 patent drawingFigure 3~4
  • EP4012338B1 patent drawingFigure 5A~5D

AI summary

According to a method for guiding a motor vehicle (1) at least in part automatically, a camera image (9) representing an environment of the motor vehicle (1) is generated by a camera system (3) of the motor vehicle (1). At least one driving command for guiding the motor vehicle (1) at least in part automatically is generated depending on the camera image (9) and depending on a navigational command (8) by using a computing unit (4) of the motor vehicle (1). A lidar pointcloud (21) representing the environment is generated by a lidar system (5) of the motor vehicle and the navigational command (8) is determined by the computing unit (4) depending on the lidar pointcloud.