GNSS-INS Navigation Using Neural Network Motion Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The GNSS and INS integrated navigation system faces challenges in maintaining accurate and reliable positioning when GNSS signals are obstructed, leading to increased positioning errors over time, especially in complex environments like urban canyons, and existing solutions such as auxiliary sensors, motion constraint algorithms, and neural networks have limitations in adaptability and real-time performance.
Innovation Solution
A method combining a BP neural network with a motion constraint algorithm to predict transverse vehicle velocity and correct velocity errors using a Kalman filter, incorporating forward and heading angular velocities as inputs, and adjusting constraints based on vehicle motion states to improve positioning accuracy and reliability after GNSS losing lock.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If auxiliary sensors (magnetometers, odometers, barometers) are added to improve positioning accuracy after GNSS losing lock, then positioning accuracy is improved, but device complexity and cost increase
Solution Approach 1:
The patent uses the IMU sensors already present in the system to generate motion constraint information without requiring additional auxiliary sensors. The system serves itself by utilizing its existing inertial measurement capabilities to provide the constraint information needed for accurate positioning during GNSS outages.
Solution Approach 2:
The patent introduces a neural network algorithm as an intermediary that processes IMU data to generate motion constraint information. This intermediary component transforms raw sensor data into useful constraint information that improves positioning accuracy without requiring physical auxiliary sensors.
2Stability of the object's composition
If fixed constraint values are used in motion constraint algorithms to ensure stability, then system stability is improved, but adaptability to different vehicle motion states deteriorates
Solution Approach 1:
The patent transforms the static, fixed constraint values into dynamic, adaptive constraint values that automatically adjust according to the vehicle's actual motion state. The neural network continuously processes IMU data to generate constraint information that adapts to different driving conditions while maintaining system stability through the filtering algorithm.
Solution Approach 2:
The system implements feedback by using the neural network to continuously monitor vehicle motion states through IMU sensors and adjust motion constraint information accordingly. This closed-loop approach ensures the system adapts to changing motion conditions while maintaining stability through the extended Kalman filter.
3Measurement precision
If neural network algorithms are used to improve positioning accuracy after GNSS losing lock, then positioning accuracy is improved, but real-time performance deteriorates due to computational complexity
Solution Approach 1:
The patent applies partial action by using the neural network only for generating motion constraint information rather than for complete positioning solution. This selective application reduces computational burden while maintaining accuracy improvements. The extended Kalman filter then efficiently integrates this constraint information with INS data to produce the final positioning solution.
Solution Approach 2:
The patent merges the neural network algorithm with the extended Kalman filter algorithm to combine the advantages of both approaches. The neural network provides adaptive motion constraint information while the EKF efficiently integrates this information with INS data, achieving both high accuracy and real-time performance through algorithmic integration.
4Duration of action of stationary object
If GNSS and INS integrated navigation is used to provide continuous positioning, then continuity is improved, but positioning accuracy deteriorates when GNSS signal is obstructed
Solution Approach 1:
The patent applies beforehand cushioning by pre-establishing motion constraint models and training neural networks during normal GNSS operation. This preparation ensures that when GNSS signals are obstructed, the system immediately has available the constraint information needed to maintain accurate positioning without experiencing a sudden degradation in performance.
Data Source
AI summary
The present disclosure relates to the field of GNSS and INS integrated navigation technology, and specifically discloses a GNSS and INS integrated navigation positioning method and a system thereof. To addresses the technical problem of positioning error divergence in integrated navigation systems caused by insufficient satellite visibility or strong multipath effects in GNSS denial environments, a method combining motion constraint algorithm and neural network algorithm is proposed for robustness by the present disclosure. The motion constraint algorithm is very stable, but it cannot self-adaptively adjust the constraint threshold based on the vehicle motion state. The neural network algorithm has strong flexibility, but the obtained predicted values inevitably have outliers. The present disclosure combines motion constraints with the neural network algorithms, enabling these two algorithms to complement advantages of each other, thereby effectively improving the positioning accuracy and reliability of the integrated navigation system after GNSS losing lock.