Autonomous Vehicle Road Region Extraction via Sensor Fusion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current autonomous vehicle navigation systems face challenges in accurately identifying and navigating through drivable road regions, especially in occluded or varied terrain and weather conditions, with existing solutions failing to provide robust and self-adaptive road region extraction.

Innovation Solution

A method and system that utilize a combination of LiDAR sensors and cameras to capture and map three-dimensional point clouds to image pixels, dynamically updating training data to predict and control the autonomous vehicle's path based on drivable road regions, excluding obstacles and adapting to different terrains and lighting conditions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple sensors and complex processing systems are used to identify road regions, then the accuracy of road region detection is improved, but the device complexity and computational requirements increase significantly

Engineering Contradiction:
Improveroad region detection accuracyVSAvoidsensor and processing system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the road region identification process into distinct functional modules: sensor data acquisition module, point cloud generation module, mapping module, and road region identification module. Each module handles a specific aspect of the processing pipeline, allowing independent optimization and reducing overall system complexity while maintaining high detection accuracy through specialized processing at each stage.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a point cloud as an intermediary representation between raw sensor data and final road region identification. The point cloud serves as a structured intermediate format that preserves spatial information while being more amenable to processing than raw sensor data, facilitating accurate road region detection without requiring direct complex processing of原始 sensor inputs.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Loss of information

If the system processes all input data including occluded objects, then comprehensive environmental awareness is achieved, but the reliability of drivable road region identification decreases due to false mapping of occluded objects

Engineering Contradiction:
Improveenvironmental information completenessVSAvoiddrivable road region identification reliability
Core Design Contradiction:
Loss of informationVSReliability

Solution Approach 1:

The patent extracts and isolates the drivable road region from the complete point cloud data by implementing specific filtering criteria. The system separates drivable surfaces from non-drivable objects and occluded regions by analyzing point distribution patterns, normal vectors, and spatial relationships, extracting only the relevant drivable road portions while excluding misleading occluded objects that would otherwise contaminate the identification results.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent implements dynamic adjustment of processing parameters based on scene characteristics. The system adapts its filtering thresholds and processing depth according to the detected environment type, weather conditions, and visibility levels, dynamically balancing between comprehensive information processing and reliable drivable region identification by adjusting the aggressiveness of occlusion filtering in real-time.

Inventive Principle:
Principle #15Dynamics

3Adaptability or versatility

If the system is designed to work on all types of roads and weather conditions, then the adaptability is improved, but the difficulty of detecting and measuring road regions increases

Engineering Contradiction:
Improveroad type and weather condition adaptabilityVSAvoidroad region detection difficulty
Core Design Contradiction:
Adaptability or versatilityVSDifficulty of detecting and measuring

Solution Approach 1:

The patent designs a universal road region identification system that handles multiple road types (paved, unpaved, icy, snowy) and weather conditions through a single integrated processing framework. The system uses multi-sensor fusion and adaptive parameter adjustment to maintain consistent performance across diverse environments, eliminating the need for separate specialized processing pipelines for different road conditions while managing detection complexity through unified algorithms.

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

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

Enables precise and robust navigation of autonomous vehicles on various road types and conditions, ensuring collision-free and smooth driving by focusing on the drivable road region, independent of structured or unstructured roads and weather.

Implementation Method 1

capturing the road region ahead of the autonomous vehicle using a plurality of sensors. The road region is captured as a three-dimensional point cloud

Methodology Applied
Scientific EffectLIDAR: LIDAR

Data Source

PatentEP3444693B1Method, system, and device for guiding autonomous vehicles based on dynamic extraction of road region
Publication Date: 2022.05.18 WIPRO LTD
  • EP3444693B1 patent drawingFigure 1
  • EP3444693B1 patent drawingFigure 2
  • EP3444693B1 patent drawingFigure 3

AI summary

A method and system for guiding an autonomous vehicle to extract drivable road region is provided. The method involves capturing the road region ahead of the autonomous vehicle using a plurality of sensors. The road region is captured as a three-dimensional point cloud. Thereafter, a plurality of images of the road ahead the autonomous vehicle is captured using a camera. The captured road region and the plurality of images are mapped and compared. The mapping involves comparing the point cloud with plurality of pixels in the images. Based on the mapping, a training data is dynamically updated to incorporate current road conditions and a drivable road region is predicted. Finally, based on the drivable region, the autonomous vehicle is controlled and guided through the road.