Multi-Camera Visual Localization for Outdoor Mobile Robots

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Localizing a mobile robot in unstructured outdoor environments is challenging due to the imprecision of publicly available maps and the dynamic nature of these environments, where existing solutions often rely on GPS systems with limited precision for autonomous navigation.

Innovation Solution

A mobile robot equipped with at least two cameras that take visual images, process them using edge detection and line extraction algorithms, and compare the extracted features with stored map data using an iterative probabilistic algorithm, such as a particle filter, to accurately determine its pose and navigate autonomously.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If GPS systems are used for robot localization, then the robot can navigate in outdoor environments, but the localization precision is limited to 1-10 meters

Engineering Contradiction:
Improvelocalization precisionVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces GPS satellite-based positioning with a ground-based visual localization system using multiple cameras and image processing. Instead of relying on electromagnetic signals from satellites, the system uses visual features (lines, edges, corners) captured by cameras and matched against pre-built maps to determine robot position and orientation, achieving centimeter-level precision

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

Solution Approach 2:

The system uses multiple cameras that can serve both localization and mapping functions. The same camera hardware and processing algorithms are used for both building the environmental map and determining robot pose, eliminating the need for separate specialized sensors

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

2Adaptability or versatility

If visual techniques are used for SLAM, then the robot can autonomously generate and update maps in dynamic environments, but the computational complexity increases

Engineering Contradiction:
Improveenvironmental adaptabilityVSAvoidcomputational complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the visual processing task into distinct stages: edge detection, line extraction, corner detection, and feature matching. By breaking down the complex SLAM problem into these manageable sub-tasks, the system can process visual information more efficiently while maintaining adaptability to dynamic environments

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs preliminary actions by pre-building detailed maps of the environment containing extracted visual features (lines, corners, edges) before localization. This pre-processing allows the robot to quickly match features during operation without performing full SLAM computations in real-time, reducing computational load

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If multiple cameras are used for visual localization, then the localization accuracy improves, but the device complexity and processing requirements increase

Engineering Contradiction:
Improvelocalization accuracyVSAvoidcamera system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent merges the data from multiple cameras into a unified coordinate system and processes them together in the localization algorithm. By combining visual information from multiple viewpoints and integrating the features across all camera images, the system achieves more accurate and robust localization than single-camera systems

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3371670B1Device and method for autonomous localisation
Publication Date: 2024.05.22 STARSHIP TECH OU
  • EP3371670B1 patent drawingFigure 1
  • EP3371670B1 patent drawingFigure 2
  • EP3371670B1 patent drawingFigure 3

AI summary

Disclosed is a mobile robot comprising at least one memory component comprising at least map data; at least two cameras adapted to take visual images; and at least one processing component adapted to at least extract straight lines from the visual images taken by the at least two cameras and compare them to the map data to at least localise the robot. Further disclosed is a localisation method comprising taking visual images with at least two cameras; extracting straight lines from the individual visual images with at least one processing component; comparing the extracted features with existing map data; and outputting a location hypothesis based on said comparison.