Camera-Based Vehicle Motion Estimation Without Inertial Sensor Drift
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inertial sensors used for measuring vehicle motion state are prone to measurement errors due to temperature changes, leading to integration errors and inaccuracies in determining velocity and rotation angles, necessitating additional sensors for correction.
Innovation Solution
Estimate vehicle motion state using camera data processed by a machine learning algorithm, such as a Kalman filter or artificial neural network, trained with historical camera data and reference motion data, to convert camera data into motion data without requiring additional sensor equipment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If inertial sensors are used to measure vehicle motion state, then measurement data can be obtained, but measurement errors increase due to temperature changes and integration errors occur over time
Solution Approach 1:
The patent replaces inertial sensors with a vision-based system using cameras and machine learning algorithms to estimate vehicle motion state. This substitution eliminates the temperature sensitivity and integration errors inherent in inertial sensors, achieving stable and accurate motion estimation without mechanical sensing components.
Solution Approach 2:
The patent introduces camera data as an intermediary to infer vehicle motion state. Instead of directly measuring acceleration and velocity with inertial sensors, the system uses visual information from cameras processed through machine learning models to indirectly but accurately determine motion parameters, avoiding the accumulation of integration errors.
2Measurement precision
If position data from global navigation satellite system and wheel speed sensors are added to correct strapdown calculation, then accuracy is improved, but device complexity increases
Solution Approach 1:
The patent replaces the complex multi-sensor correction system (satellite positioning + wheel speed sensors + Kalman filter) with a streamlined vision-based machine learning system. This substitution maintains high accuracy for velocity and rotation angle measurement while significantly reducing the number of required sensors and system complexity.
Solution Approach 2:
The machine learning model trained on camera data serves multiple functions simultaneously: it estimates velocity, rotation angles, and motion state without requiring separate correction mechanisms. This multi-functionality eliminates the need for additional satellite and wheel speed sensors that would otherwise be needed for accurate measurement.
3Device complexity
If camera data is processed through machine learning algorithm, then additional sensor equipment is avoided, but computational processing requirements increase
Solution Approach 1:
The patent performs preliminary action by training the machine learning algorithm offline using historical camera data and reference motion data. This pre-training creates a ready-to-use model that can be deployed in the vehicle, shifting the computational burden from real-time operation to offline preparation, thereby reducing in-vehicle energy consumption and processing requirements.
Solution Approach 2:
The patent uses historical camera data as copies to train the machine learning model. By learning from these training examples, the system creates a knowledge base that enables accurate motion estimation without requiring complex real-time processing or additional sensors during actual vehicle operation.
Data Source
AI summary
A method is for estimating a motion state of a vehicle, which is equipped with a camera for sensing a surroundings of the vehicle. The method includes receiving camera data that has been generated by the camera and inputting the camera data into an algorithm in order to convert the camera data into motion data that describe a current motion state of the vehicle. The algorithm has been trained with historical camera data and with reference motion data associated with the historical camera data. The method further includes determining an estimated motion state of the vehicle from the motion data using a state estimator.