Robust Indoor Positioning of Automated Guided Vehicles in Internet of Things Networks With Deep Convolution Neural Network Considering Adversarial Attacks

The effectiveness of positioning techniques that utilize the receiver signal strength (RSS) is highly dependent on the instability of the received signal strength indicator (RSSI). Up to now, there is no strategy that effectively lowers the influence of such instability on the accuracy of positioning. Moreover, recent studies showed that indoor positioning techniques are vulnerable to noise in RSSI data and cyber-attacks, which make them more expensive. In this study, a new Internet of Things (IoT) paradigm is proposed for the indoor positioning of automated guided vehicles (AGVs) using a deep convolution neural network (CNN). The proposed method handles signal processing by converting the RSSI signal into an image. In which, the 1-D RSSI signal is converted into 2-D image data in order to generate the new features based on continuous wavelet transform (CWT), and then the proposed deep CNN is implemented for the indoor positioning system. The test results show that the proposed model can outperform other state-of-the-art positioning techniques with small position errors. Furthermore, the robustness of the proposed model is validated against various adversarial attacks. In addition, the proposed method can have a lower impact on RSSI change compared with other methods.