Multi-source Fusion Positioning Technology For Emergency Rescue In Underground Environments

With the continuous expansion of the scope of human production and life,underground resources are constantly being excavated,and the gradual migration of human activities to the underground has become a major trend.Once a disaster occurs in this environment and emergency rescue is required,the current emergency rescue system that mainly relies on the GNSS system to provide location information support is difficult to play an effective role.Facing the problem of how to quickly build a large underground space accurate positioning system,this paper proposes a multi-source fusion positioning method suitable for the underground environment by analyzing the special needs of emergency rescue in the post-disaster underground environment,which has high temporal resolution and anti-multipath performance.Strong and fast layout.The research contents of the paper are as follows:(1)In the post-disaster underground scene,a large number of nonline-of-sight environments will cause the positioning accuracy of the multisource fusion positioning system to decrease.This paper proposes a channel impulse response(CIR)-based non-line-of-sight Discrimination method.Wavelet transform is used to convert CIRs with similar waveforms in the time domain into time-frequency diagrams with significant differences.After that,in view of the high data density and high noise characteristics of CIR time-frequency images compared with traditional images,and the high real-time requirements of post-disaster positioning scenarios,this paper makes targeted improvements to the convolutional neural network(CNN)used for general image recognition..Finally,in view of the insufficient amount of data in emergency rescue scenarios and insufficient time to train the model,the MAML method in meta-learning is used to train the model.The above improvements have achieved good results on both public datasets and self-collected datasets.(2)Aiming at the dynamic optimization problem of the weight distribution of each sensor and the elimination of abnormal measurement value interference in the multi-source fusion positioning system,a factor graph model for weight adaptive adjustment is proposed in this paper.It has the advantages of plug and play,and can integrate asynchronous heterogeneous navigation source data,and can process abnormal measurement values in real time.Aiming at the problem of high computational complexity when new measurement values are input,a local update method is proposed,which can reduce computational redundancy without affecting the positioning accuracy,thereby significantly improving real-time performance.(3)In order to verify the method proposed in this paper,simulation and field experiments are carried out.Both types of experiments prove that the method proposed in this paper has a significant improvement in the discrimination rate,positioning accuracy and running time compared with the benchmark method.