Robotic State Mapping for Real-Time 3D Pose Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for constructing a three-dimensional representation of an environment using a robotic device face challenges in real-time processing due to computational requirements, particularly with unpredictable motion and non-planar landscapes, leading to issues like drift and inaccurate mapping.
Innovation Solution
A mapping system that optimizes the state of a robotic device using kinematic, odometric, and geometric errors to update its pose and the environment model, employing a state engine to jointly optimize current and previous states without a pose graph, allowing for real-time dense mapping even with 'loopy' or 'choppy' trajectories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dense mapping techniques are used to generate three-dimensional representations, then mapping precision is improved, but computational requirements increase making real-time processing difficult
Solution Approach 1:
The patent segments the dense mapping problem into two parts: (1) sparse feature extraction and tracking to determine camera pose, and (2) separate dense reconstruction using pre-computed depth maps. This segmentation allows real-time pose estimation while maintaining the option for high-quality dense mapping without real-time computational burden.
Solution Approach 2:
The patent performs preliminary action by pre-computing depth maps from multi-view stereo reconstruction offline, before real-time operation. During real-time mapping, these pre-computed depth maps are used directly, eliminating the need for computationally intensive real-time dense reconstruction while maintaining mapping precision.
2Measurement precision
If traditional SLAM methods with pose graphs are used, then mapping accuracy is maintained, but device complexity and computational burden increase
Solution Approach 1:
The patent extracts and removes the pose graph component from traditional SLAM systems. Instead of maintaining a pose graph structure, the system directly optimizes camera poses and map points using bundle adjustment, simplifying the system architecture while maintaining mapping accuracy through direct geometric optimization.
Solution Approach 2:
The patent uses pre-computed depth maps as copies of three-dimensional information, replacing the need for complex real-time depth estimation mechanisms. These depth map copies provide accurate geometric information for dense reconstruction without requiring complex real-time processing.
3Loss of information
If continuous image capture is performed during robotic device movement, then mapping completeness is improved, but data processing requirements and computational load increase
Solution Approach 1:
The patent extracts only the essential information from continuous image captures - specifically sparse feature points for pose estimation - while discarding redundant pixel-level data. This extraction approach maintains mapping completeness through sufficient feature sampling while dramatically reducing processing requirements.
Solution Approach 2:
The patent performs preliminary action by pre-processing images offline to extract depth information and create depth maps before real-time operation. During real-time mapping, only lightweight feature matching and pose optimization are performed on captured images, maintaining completeness while ensuring processing speed.
Data Source
AI summary
Certain examples described herein enable a robotic device to accurate map a surrounding environment. The robotic device uses an image capture device and at least one of the image capture device and the robotic device move within the environment. Measurements associated with movement of at least one of the image capture device and the robotic device are used to determine a state of the robotic device. The state of the robotic device models the image capture device and the robotic device with respect to a model of the environment that is constructed by a mapping engine. By comparing the state of the robotic device with a measured change in the robotic device, an accurate representation of the state of the robotic device may be constructed. This state is used by the mapping engine to update the model of the environment.