Pedestrian Inertial Navigation Zero-Velocity Update Point Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional zero-velocity update methods in pedestrian inertial navigation systems, such as fixed and dynamic threshold methods, suffer from high cumulative errors and poor generalization performance due to their inability to adaptively select zero-velocity update points effectively across different pedestrians and motion states.
Innovation Solution
A pedestrian adaptive zero-velocity update point selection method based on a convolutional neural network is developed, which collects and preprocesses inertial navigation data to train a model that can accurately identify zero-velocity update points, integrating with the extended Kalman filter to reduce cumulative errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a fixed threshold method is used for zero-velocity update point selection, then the method is simple to implement, but the cumulative error increases and generalization performance deteriorates across different pedestrians and motion states
Solution Approach 1:
The patent transforms the static fixed threshold into a dynamic adaptive threshold by training a neural network model on gait data from multiple pedestrians. The model dynamically adjusts the zero-velocity update point selection based on individual gait characteristics and motion states, resolving the contradiction between implementation simplicity and reliability by automating the threshold adaptation process.
Solution Approach 2:
The patent changes the parameter selection approach from manual fixed threshold setting to automated neural network-based parameter optimization. The system learns optimal threshold parameters from training data and adapts them to different pedestrians and motion states, maintaining simplicity while improving cumulative error control through data-driven parameter optimization.
2Adaptability or versatility
If a dynamic threshold method is used for zero-velocity update point selection, then the adaptation to different velocities is improved, but the computational complexity increases and accuracy remains insufficient
Solution Approach 1:
The patent applies preliminary action by pre-training the neural network model offline on comprehensive gait data from multiple pedestrians and motion states. The model learns and stores optimal threshold selections in advance, so during actual navigation, the system only needs to infer from pre-learned patterns rather than performing complex real-time calculations, thus reducing online computational complexity while maintaining velocity adaptation.
Solution Approach 2:
The patent uses copying by training the neural network on gait data from multiple pedestrians and then copying the learned patterns to individual pedestrians. The model captures general gait characteristics that can be applied across different users, reducing the need for extensive individual calibration while maintaining adaptability to different velocities and motion states.
3Measurement precision
If manual selection of fixed threshold is performed for different pedestrians and motion states, then the optimization for specific conditions is improved, but the time consumption and operational difficulty increase significantly
Solution Approach 1:
The patent implements self-service by enabling the system to automatically select and optimize thresholds for each pedestrian without manual intervention. The neural network model processes individual gait data and autonomously determines optimal thresholds during the first use or through continuous learning, eliminating time-consuming manual operations while maintaining high precision threshold optimization.
Solution Approach 2:
The patent applies feedback by using the neural network to continuously learn from actual gait data and motion state information. The system receives feedback from sensor measurements and motion state detection, then automatically adjusts threshold selections based on learned patterns, achieving precise optimization without manual time investment while adapting to individual pedestrian characteristics.
Data Source
AI summary
A pedestrian adaptive zero-velocity update point selection method based on a neural network, including the following steps: S1, collecting inertial navigation data of different pedestrians in different motion modes; S2, preprocessing the inertial navigation data collected in the step S1, labeling the preprocessed data, and obtaining a training data set, a validation data set, and a test data set according to the preprocessed data and a label corresponding to the preprocessed data; S3, inputting the training data set to a convolutional neural network for training, obtaining a pedestrian adaptive zero-velocity update point selection model based on the convolutional neural network, and using the validation data set to validate the pedestrian adaptive zero-velocity update point selection model; and S4, inputting the test data set into the pedestrian adaptive zero-velocity update point selection model based on the convolutional neural network, and obtaining a selection result of pedestrian zero-velocity update points.