Mobile Robot Pose Localization Using Contour Gradient Histograms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mobile robot localization methods, such as particle filter algorithms, suffer from low accuracy due to random sampling, which may not include the true or similar pose of the robot in sampled poses, reducing the precision of pose determination in global maps.
Innovation Solution
A pose determining method for mobile robots that uses sensor data to build local and global maps through SLAM technology, rotates initial local maps to detect environmental contour points and gradients, constructs angle histograms, and predicts the target pose based on gradient directions and appearance counts, ensuring accurate localization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If particle filter algorithm with random sampling is used, then the method can determine pose without external localization system, but the localization accuracy is reduced because true pose may not be included in sampled poses
Solution Approach 1:
The patent performs preliminary actions by rotating the local map to multiple candidate poses before final pose determination, and pre-processes environmental data to extract contour points and gradient information. This preliminary preparation ensures that the true pose is included among the candidate poses being evaluated, resolving the issue of random sampling potentially missing the true pose while maintaining systematic exploration of the pose space.
Solution Approach 2:
The patent segments the pose determination process into distinct stages: generating multiple candidate poses through local map rotation, extracting environmental features (contour points, gradients) for each candidate pose, and evaluating poses based on feature matching. This segmentation transforms the single random sampling step into multiple structured sampling steps, ensuring comprehensive coverage of possible poses while improving accuracy.
2Measurement precision
If multiple candidate poses are generated through map rotation, then pose determination accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent extracts key environmental features (contour points and gradient directions) from the maps and uses these extracted features for pose evaluation instead of processing complete map images. This extraction approach significantly reduces computational complexity and processing time while maintaining pose determination accuracy, as only the most discriminative features are used for comparison.
Solution Approach 2:
The patent generates a limited number of candidate poses (e.g., 5-10 poses) through local map rotation rather than exhaustively searching all possible poses. This partial action approach provides sufficient accuracy for practical applications while avoiding excessive computational time, balancing between thoroughness and efficiency.
Data Source
AI summary
The present disclosure provides a pose determining method for a mobile robot as well as an apparatus and a mobile robot thereof. The method includes: obtaining a first position of a mobile robot in each local map after building an initial local map corresponding to a current environment and rotating; determining first environmental contour points of each of the local maps and corresponding first gradient directions, and obtaining a relative position of each of the first environmental contour points and the corresponding first position; building an angle histogram in each of the local maps; determining a second position of second environmental contour points of a global map and corresponding second gradient directions; and predicting a third position in the global map of the mobile robot, counting an appearance amount of the third positions, and determining a target pose of the mobile robot in the global map.