Dynamic Covariance Matrix for Vehicle Pose Estimation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current pose estimation methods for unmanned systems are unreliable due to varying uncertainties in sensor data, such as GPS and inertial tracking, which lead to cumulative errors and inaccurate heading information, as they fail to account for changes in operating conditions and the behavior of different measurement techniques.

Innovation Solution

A computer system calculates multiple preliminary estimates using sensor data from GPS, inertial, and wheel odometry, adjusting a covariance matrix based on time, distance, and velocity to dynamically weight the estimates and apply process noise covariance matrices, thereby reducing overall uncertainty and improving accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple data acquisition methods are combined to reduce uncertainty, then measurement precision is improved, but device complexity increases

Engineering Contradiction:
Improvepose estimation accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines multiple data acquisition methods (GPS, inertial tracking, wheel odometry) into a unified pose estimation system. The computer system integrates data from these diverse sources, processing them together to produce a consolidated pose estimate that leverages the strengths of each method while compensating for their individual weaknesses.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The computer system performs multiple functions: it acquires data from various sensors, calculates individual pose estimates from each sensor type, adjusts covariance matrices dynamically, and generates a final consolidated pose estimate. This multi-functional approach allows a single system to handle diverse data types and processing requirements.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Device complexity

If fixed covariance matrices are used in pose estimation, then device complexity is reduced, but measurement precision deteriorates due to cumulative errors

Engineering Contradiction:
Improveprocessing simplicityVSAvoidpose estimation accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent implements dynamic covariance matrix adjustment where the covariance values are not fixed but are continuously updated based on the current operating state of the platform. The system monitors platform state (such as whether the platform is stationary or moving) and adjusts the covariance matrices accordingly, allowing the estimation process to adapt to changing conditions and reduce cumulative errors.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system uses feedback from the platform state to continuously adjust the covariance matrices. By monitoring whether the platform is stationary or in motion, and using this information to modify the covariance values, the system creates a closed-loop control mechanism that improves estimation accuracy over time by learning from actual operating conditions.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP2856273B1Pose estimation
Publication Date: 2018.11.14 CSX TRANSPORTATION INC
  • EP2856273B1 patent drawingFigure 1
  • EP2856273B1 patent drawingFigure 2
  • EP2856273B1 patent drawingFigure 3A

AI summary

A solution for estimating the pose of a platform, such as a vehicle, is provided. Data from a plurality of types of sensing devices located on the platform can be used to independently calculate a plurality of preliminary estimates corresponding to the pose. A plurality of estimates corresponding to the pose can be generated using the preliminary estimates and at least one covariance matrix. One or more entries in the covariance matrix are adjusted based on an uncertainty for the corresponding preliminary estimate. The uncertainty can vary based on time, distance, and/or velocity of the platform.