Compound Probabilistic Filtering for Vehicle Localization and Map Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle localization and map update systems face challenges in accurately determining measurement noise, particularly in autonomous driving applications, due to varying satellite constellations and multipath interference, leading to computational complexity and divergence in probabilistic filters, and insufficient utilization of multiple sensors.
Innovation Solution
A system and method that leverages multiple sensors, such as GNSS and cameras, to fuse feature data and adjust measurement noise using internal variables of probabilistic filters, employing a compound probabilistic filter to jointly estimate vehicle location and map state, adapting measurement noise in real-time based on sensor quality and historical performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If measurement noise is adjusted in real-time to improve localization accuracy, then localization accuracy is improved, but computational complexity increases and filter divergence may occur
Solution Approach 1:
The system segments the measurement noise adjustment task by maintaining multiple parallel probabilistic filters, each with predetermined measurement noise values representing different hypotheses. This divides the complex real-time adjustment problem into manageable discrete segments that can be evaluated in parallel and combined through weighted averaging.
Solution Approach 2:
The system dynamically adapts to varying measurement conditions by adjusting the weights of different filter hypotheses based on their performance. The measurement noise values and corresponding weights are updated in real-time based on the correlation between predicted and actual measurements, allowing the system to respond dynamically to changing satellite constellations and multipath interference.
2Measurement precision
If multiple sensors are utilized to improve localization accuracy, then localization accuracy is improved, but device complexity increases
Solution Approach 1:
The system merges data from multiple sensors (GNSS, cameras, and other vehicle sensors) into a unified probabilistic filtering framework. By combining measurements from different sensor types with complementary characteristics, the system achieves improved localization accuracy while managing complexity through a standardized processing approach.
Solution Approach 2:
The compound probabilistic filter system serves multiple functions simultaneously: it processes data from various sensor types, performs localization, estimates map states, and adapts to different measurement conditions. This multi-functionality reduces the need for separate specialized systems for each task.
3Productivity
If predetermined measurement noise values are used to simplify computation, then computational complexity is reduced, but localization accuracy deteriorates under varying conditions
Solution Approach 1:
The system uses parameter changes by maintaining multiple predetermined measurement noise values corresponding to different measurement conditions and satellite constellations. Instead of using a single fixed noise value, the system selects and weights different noise parameters based on current conditions, achieving both computational efficiency and accuracy.
Solution Approach 2:
The system incorporates feedback mechanisms where the performance of each filter hypothesis is evaluated based on the correlation between predicted and actual measurements. This feedback is used to dynamically adjust the weights of different hypotheses, ensuring that the most accurate models are prioritized while maintaining computational efficiency.
4Measurement precision
If real-time measurement noise estimation is performed to improve accuracy, then localization accuracy is improved, but filter divergence occurs
Solution Approach 1:
The system prepares for potential filter divergence by maintaining multiple predetermined measurement noise hypotheses in advance. When measurement conditions change or divergence is detected, the system can switch to alternative hypotheses that are pre-configured to be more stable, cushioning against the effects of divergence before they occur.
Solution Approach 2:
The compound probabilistic filter acts as an intermediary between individual filter hypotheses and the final localization output. It combines multiple hypotheses with different measurement noise values, smoothing out extreme variations and preventing any single hypothesis from causing divergence, thereby stabilizing the overall system.
Data Source
AI summary
The present disclosure discloses a system and a method for jointly controlling a vehicle and updating a map using multiple probabilistic filters. The method comprises collecting a sequence of measurements indicative of the state of the vehicle at different control steps. The method also comprises executing multiple probabilistic filters configured to jointly track a current state of the location of the vehicle represented by coordinates of the vehicle and a current state of the map represented by coefficients of a polynomial forming a spline fitting representation of the map. The method including determining the location of the vehicle based on a first weighted combination of current states of the location and updating the map based on a second weighted combination of current states of the map, such that the weights of the first weighted combination and the second weighted combination are different.