Distance calibration method, device and equipment in non-line-of-sight scene

Abstract

The embodiment of the present invention provides a distance calibration method, device and equipment in a non-line-of-sight scene. The method comprises the following steps: acquiring a first raw distance and a first channel impulse response waveform of ultra-wideband radio frequency RF ranging between a first target point and a second target point; inputting the first raw distance and the first channel impulse response waveform into a well-trained deep learning model, and obtaining a distance calibration result and a confidence degree between the first target point and the second target point,wherein the well-trained deep learning model is a deep learning framework composed of a variational self-encoder and a probabilistic neural network. The embodiment of the invention performs calibration of the ranging result and gives the confidence degree of the calibration result through the deep learning model, which gives a lower confidence degree to the distance that is difficult to be correctly calibrated and reduces the probability that an unreliable prediction result causes a serious error of the system.

Description

technical field [0001] Embodiments of the present invention relate to the fields of deep learning and pattern recognition, and in particular to a distance calibration method, device and equipment in a non-line-of-sight scene. Background technique [0002] Due to the complex indoor environment and the influence of non-line-of-sight transmission and multipath transmission, the ranging algorithm based on electromagnetic wave propagation time is not accurate enough. [0003] Currently, there have been some works on indoor odometry correction using traditional machine learning algorithms. However, these traditional machine learning algorithms do not work well. First, they lack sufficient model capacity and generalization ability to provide a sufficiently accurate prediction of the true distance based on environmental information; second, due to the limitations of the model itself, the prediction cannot be given confidence level. In many indoor positioning applications that requ...

AI Technical Summary

Problems solved by technology

However, these traditional machine learning algorithms do not work well. First, they lack sufficient model capacity and generalization ability to provide a sufficiently accurate prediction of the true distance based on environmental information; second, due to the limitations of the model itself, the prediction cannot be given Confidence of

In many indoor positioning applications that require high precision and high robustness, traditional machine learning algorithms cannot meet the precision requirements, and the output results lack sufficient reliability.

The complex indoor environment makes it almost impossible to collect enough data in various possible environments, which leads to the traditional machine learning model having to give predictions in various never-before-seen scenarios. When this new When the statistical distribution of the scene is different from the statistical distribution of the model training data, the model will give very unreasonable prediction results

Images