Construction Robot Localization Using Semantic Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for localizing mobile construction robots on construction sites are not robust or precise, especially in environments with interfering objects, and often require manual interventions or inflexible BIM data-based solutions.

Innovation Solution

A method that captures environment data using sensors like LIDAR scanners, performs semantic segmentation to classify data into background and foreground, and updates a semantic model to improve localization accuracy, allowing the robot to autonomously discard noisy data and enhance its positioning based on BIM data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If BIM data and distance measurements are used for localization, then the method is simple to implement, but the localization precision deteriorates due to interfering objects like persons, power tools, and material depositions

Engineering Contradiction:
Improveease of implementationVSAvoidlocalization precision
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent applies segmentation by dividing the environment data into semantic classes (foreground and background) to separate interfering objects from relevant localization features. The semantic classifier segments point cloud data into classes such as persons, furniture, walls, and floors, allowing the system to selectively use only background elements for localization, thereby maintaining precision in cluttered environments while keeping the overall approach manageable

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts only the necessary subset of environment data for localization by filtering out foreground objects that interfere with positioning. The semantic classifier identifies and extracts background elements (walls, floors, ceilings) from the complete environment scan, discarding interfering objects like persons and equipment, thus achieving precise localization using only relevant data

Inventive Principle:
Principle #2Taking out (Extraction)

2Reliability

If a semantic classifier with updatable model is used to filter environment data, then the localization robustness improves in cluttered environments, but the device complexity increases

Engineering Contradiction:
Improvelocalization robustnessVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The semantic classifier implements self-service through automatic model updating using the environment data itself. The system learns from the captured environment data without requiring manual training or external intervention, automatically adapting to different construction sites and configurations. This self-updating capability maintains high robustness while reducing the operational complexity of deploying the system in varied environments

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary semantic classification and model updating before localization occurs. By pre-processing the environment data to identify and classify objects, and updating the semantic model in advance based on captured data, the system prepares the filtered environment data for more accurate and robust localization execution

Inventive Principle:
Principle #10Preliminary action

3Quantity of substance

If all environment data is used for localization, then the data availability is maximized, but the localization precision deteriorates due to noisy and interfering data from construction site clutter

Engineering Contradiction:
Improvedata availabilityVSAvoidlocalization precision
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent extracts only the necessary subset of environment data for localization by filtering out foreground objects that interfere with positioning. The semantic classifier identifies and extracts background elements (walls, floors, ceilings) from the complete environment scan, discarding interfering objects like persons and equipment, thus achieving precise localization using only relevant data

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different quality treatments to different parts of the environment data. Background elements receive high-quality processing for localization, while foreground objects are filtered out or down-weighted. This local differentiation in data quality handling ensures that precise localization is achieved using only the high-quality background data while maintaining comprehensive environmental awareness

Inventive Principle:
Principle #3Local quality

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

This approach enables robust and precise localization of mobile construction robots, even in cluttered environments, by selectively using environment data and continuously updating the semantic model, thereby improving the robot's autonomy and accuracy.

Implementation Method 1

The environment data may be obtained by distance measurement sensors, e. g. by a LIDAR scanner or a time-of-flight camera.

Methodology Applied
Scientific EffectTime of flight: Time of Flight

Data Source

PatentEP4330784B1Method for localizing a mobile construction robot on a construction site using semantic segmentation, construction robot system and computer program product
Publication Date: 2024.11.27 HILTI AG
  • EP4330784B1 patent drawingFigure 1
  • EP4330784B1 patent drawingFigure 2A~2D
  • EP4330784B1 patent drawingFigure 3

AI summary

The invention concerns a method (10) of localizing a mobile construction robot (102) on a construction site (101), wherein environment data (12) of the construction site is captured, wherein a pose of the mobile robot (102) is inferred using the environment data (12), wherein the method (10) comprises a step of semantic segmentation, wherein a semantic classifier (18) having an updatable semantic model classifies the environment data (12) into at least two semantic classes, wherein the semantic model is updated at least once. Furthermore, the invention concerns a mobile construction robot system (100) and a computer program product (108). The invention provides solutions for a robust and precise localization of the mobile construction robot (102).