Construction Robot Localization by Semantic Filtering of Site Clutter
Find Innovative SolutionsGenerate 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 involving semantic segmentation of environment data using an updatable semantic classifier to distinguish between construction site background and foreground, allowing for precise localization by discarding noisy data and utilizing 3D sensors like LIDAR scanners and cameras, with continual learning and self-supervision to improve classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If BIM data and distance measurements are used for localization, then map-based positioning is achieved, but localization fails due to interfering objects like persons, power tools, and material depositions
Solution Approach 1:
The patent segments the environment data into different semantic classes (background vs. foreground objects) to distinguish between static construction elements and moving interfering objects. This segmentation allows the system to selectively use only the reliable background data for localization, filtering out the harmful foreground objects that cause measurement failures.
Solution Approach 2:
The patent extracts and removes interfering objects from the environment data by classifying them as foreground elements. By taking out these harmful factors (persons, tools, materials) from the localization calculation, the system achieves robust positioning that is not affected by temporary obstructions on the construction site.
2Measurement precision
If manual interventions are required for localization setup, then initial positioning can be achieved, but the system lacks flexibility and usability
Solution Approach 1:
The patent implements self-service by enabling the mobile construction robot to autonomously perform semantic segmentation and localization without manual intervention. The system automatically classifies environment data, identifies background versus foreground objects, and computes its position independently, thereby achieving both high precision and ease of operation.
Solution Approach 2:
The patent applies preliminary action by pre-training the semantic classifier model with construction site data before deployment. This preliminary training enables the system to automatically recognize and classify construction elements, allowing the robot to perform accurate localization from the start without requiring manual setup or calibration on-site.
3Loss of information
If all environment data is used for localization, then comprehensive positioning information is available, but noisy and interfering data reduces localization accuracy
Solution Approach 1:
The patent segments environment data into semantic classes to separate useful background information from noisy foreground objects. This segmentation preserves the complete environmental information for analysis while selectively using only the clean background data for localization calculations, thereby maintaining information completeness while improving accuracy.
Solution Approach 2:
The patent introduces a semantic classifier as an intermediary between raw environment data and the localization algorithm. This intermediary processes and filters the data, translating comprehensive environmental information into a refined set of reliable background features that can be accurately used for positioning without the noise of interfering objects.
4Adaptability or versatility
If a fixed semantic model is used for classification, then initial segmentation is achieved, but the system cannot adapt to varying construction site conditions
Solution Approach 1:
The patent applies dynamics by making the semantic classifier model updateable and adaptable to different construction sites. The system can dynamically adjust its classification parameters and retrain on site-specific data, allowing it to adapt to varying conditions while maintaining a relatively simple base architecture that avoids excessive complexity.
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
The method enhances the autonomy and precision of mobile construction robots by filtering out interfering data, improving localization accuracy by 60% and reducing median localization error by 10%, even in cluttered environments, through on-board sensor data processing and adaptive semantic model updates.
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.
Data Source
AI summary
A method of localizing a mobile construction robot on a construction site, wherein environment data of the construction site is captured, wherein a pose of the mobile robot is inferred using the environment data disclosed, wherein the method comprises a step of semantic segmentation, wherein a semantic classifier having an updatable semantic model classifies the environment data into at least two semantic classes, wherein the semantic model is updated at least once. Furthermore, the invention concerns a mobile construction robot system and a computer program product. The invention provides solutions for a robust and precise localization of the mobile construction robot.