Robust Sensor Fusion for SLAM Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for sensor fusion, such as particle filters, often fail to accurately update estimates of a robot's pose after kidnapping events due to limited sample representation and unknown statistical properties of measurements, leading to slow convergence or incorrect estimates, especially in environments with noisy or incomplete data.
Innovation Solution
The implementation of Visual Simultaneous Localization and Mapping (VSLAM) technologies using multiple particles to maintain multiple hypotheses, which combines data from visual sensors and dead reckoning sensors to create and update maps, allowing for robust recovery from kidnapping events by correcting drift in dead reckoning measurements with visual landmarks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a relatively large number of samples (particles) is used to improve estimation accuracy, then the estimation precision is improved, but the computational requirement increases significantly
Solution Approach 1:
The patent segments the particle filter algorithm into multiple parallel processors, where each processor handles a subset of particles. This segmentation allows the system to maintain high estimation precision with a large number of particles while distributing the computational load across multiple processors, thereby reducing the computational requirement on any single processor and enabling parallel computation.
2Reliability
If the error in the initial estimate is large, then the system must handle kidnapping events, but the estimate converges slowly or not at all
Solution Approach 1:
The patent implements dynamic particle management where the system adapts to large initial errors by dynamically adjusting particle distribution and weights. When kidnapping events are detected, the system dynamically resamples particles and redistributes them across the state space, enabling fast convergence even when the initial estimate is completely wrong. This dynamic adaptation allows reliable recovery from kidnapping within a short time frame.
3Reliability
If the statistical properties of measurements are not known with accuracy, then the performance of particle filter breaks down, but using robust treatment allows faster and more reliable recovery
Solution Approach 1:
The patent employs parameter changes by adaptively adjusting particle weights, variances, and distribution parameters based on the observed measurement data rather than relying on pre-specified statistical properties. When measurements arrive, the system dynamically updates the particle parameters to match the actual data distribution, enabling robust performance even when initial statistical assumptions are incorrect. This adaptive parameter adjustment maintains processing efficiency while achieving reliable recovery.
Data Source
AI summary
This invention is generally related to methods and apparatus that permit the measurements from a plurality of sensors to be combined or fused in a robust manner. For example, the sensors can correspond to sensors used by a mobile device, such as a robot, for localization and/or mapping. The measurements can be fused for estimation of a measurement, such as an estimation of a pose of a robot.