Gaussian Process Visual Positioning for Noisy Outdoor Robot Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robot positioning methods, such as bundle adjustment and filtering, suffer from poor accuracy in outdoor and noisy environments, failing to effectively track and predict robot trajectories due to their limitations in handling spatial dependencies and historical data.
Innovation Solution
A visual positioning method based on a Gaussian process that collects image information and extracts global and semantic features, processing them to establish a Gaussian process observation model, which is then used to reconstruct a Bayes filtering framework for improved trajectory prediction and navigation guidance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bundle adjustment method is used for robot positioning optimization, then the ability to optimize the whole trajectory is improved, but the tracking of moving trajectory is lost and positioning error increases in outdoor changing and noisy environment
Solution Approach 1:
The patent segments the positioning optimization into two distinct components: bundle adjustment for global trajectory optimization and filtering for real-time state estimation and tracking. This segmentation allows each method to perform its strength while compensating for the other's weaknesses in outdoor environments
Solution Approach 2:
The patent implements a dynamic switching mechanism that adapts between bundle adjustment and filtering methods based on environmental conditions and robot motion states. The system dynamically adjusts the positioning strategy to maintain both global optimization and local tracking reliability
2Ease of operation
If filtering method is used for robot positioning optimization, then the latest state estimation is improved, but the positioning accuracy is reduced by not considering historical state
Solution Approach 1:
The patent performs preliminary action by using bundle adjustment to optimize the entire trajectory beforehand, establishing a globally optimized path. This pre-optimized trajectory then serves as a reference for the filtering method to perform real-time state estimation, ensuring both historical context and current accuracy
3Device complexity
If only latest state is estimated without considering historical state, then the computational complexity is reduced, but the positioning accuracy is inevitably reduced
Solution Approach 1:
The patent introduces an intermediary mechanism where the bundle adjustment results serve as a mediator between historical trajectory data and current state estimation. The filtering method uses both the pre-optimized global trajectory and current sensor data, effectively incorporating historical information without requiring direct processing of all historical states
Data Source
AI summary
The present provides a visual positioning method and system based on a Gaussian process and a storage medium. The method includes: collecting image of surrounding environment and moving trajectory points while traveling (S100); extracting global features and semantic features from images (S200); processing the extracted global features and semantic features and the moving trajectory points according to a preset processing rule to obtain a Gaussian process observation model (S300); and reconstructing a Bayes filtering framework according to the Gaussian process observation model, endowing a current trajectory with an initial position point, and generating a next position point of the current trajectory, the next position point being used for providing a positioning guidance for navigation (S400). An association between a current state and a historical state can be established, so that the accuracy of a predicted next position point is improved, and accurate navigation can be provided for a robot motion.