Robot Localization with Dynamic Obstacle Removal in 3D SLAM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robot localization systems struggle in dynamic environments, such as factories, where dynamic obstacles are difficult to distinguish from static objects, leading to inaccurate mapping and localization, and manual editing of 2D maps fails to address 3D obstacles, necessitating time-consuming and error-prone manual map updates.
Innovation Solution
A robot localization system that simultaneously performs mapping and localization using SLAM, detects and removes dynamic obstacles, and edits 3D SLAM data by user manipulation, optimizing pose graphs and generating 2D grid maps from updated 3D data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If map building and localization are performed separately with manual map correction, then localization accuracy is improved in static environments, but the system cannot handle dynamic environments where objects frequently change
Solution Approach 1:
The patent implements dynamic map building that automatically adapts to environmental changes by continuously detecting and classifying objects as static or dynamic during the mapping process. The system adjusts its mapping strategy in real-time based on object mobility detection, allowing it to handle both static and dynamic environments effectively without requiring separate manual correction steps.
Solution Approach 2:
The system performs automatic dynamic object detection and classification without requiring manual intervention. The robot autonomously identifies moving objects, distinguishes them from static structures, and adjusts the mapping process accordingly, eliminating the need for users to manually correct maps or determine whether objects are dynamic.
2Measurement precision
If manual map editing is performed to correct map errors, then map accuracy is improved, but time consumption and user error risk increase
Solution Approach 1:
The system automatically detects and classifies dynamic objects during the mapping process, eliminating the need for manual map editing. The autonomous classification and handling of dynamic objects reduces both the time required and the potential for user errors in map correction.
Solution Approach 2:
The system continuously monitors the environment during mapping, providing real-time feedback about detected objects and their mobility status. This feedback mechanism allows the system to automatically adjust the mapping process and correct potential errors without requiring manual intervention or post-processing.
3Device complexity
If 2D grid maps are used for localization, then the system is simpler to implement, but 3D obstacles cannot be edited or properly represented
Solution Approach 1:
The patent transitions from traditional 2D grid mapping to 3D point cloud-based mapping while maintaining computational efficiency. By utilizing depth information from sensors like LiDAR or stereo cameras, the system creates accurate 3D representations of the environment including obstacles, while still enabling efficient localization algorithms to operate on the enriched spatial data.
4Stability of the object's composition
If dynamic obstacles are not removed from the map, then the map remains complete, but the robot plans unnecessary detour paths or experiences localization difficulties
Solution Approach 1:
The system applies different processing treatments to different regions of the map based on object classification. Static objects are permanently incorporated into the map structure, while dynamic objects are identified and handled separately - either temporarily marked or removed from the navigable space. This localized quality adjustment allows the robot to plan efficient paths by avoiding only the necessary dynamic obstacles while maintaining awareness of the complete environment.
Data Source
AI summary
An apparatus for localizing a robot having robustness to a dynamic environment includes a map building unit which builds a map based on SLAM; a localizing unit which acquires first feature from sensor data acquired by a sensor mounted in a robot and localizes the robot using the first feature acquired from the sensor data based on the map built by the map building unit; and a map updating unit which reduces an error caused by the movement of the robot by correcting the first feature using an estimated position of the robot with regard to a feature obtained from a static object, among the first features acquired by the localizing unit.