Passive Laser Array Positioning: Real-Time Object Positioning System Based on Transformer

Currently, positioning primarily relies on sources such as Bluetooth, Wi-Fi, or visual cues. However, these methods have limitations in precision, sampling rate, and environmental requirements, potentially restricting their applicability across various scenarios. To enhance real-time positioning performance, this study devised a system based on laser arrays for positioning. By leveraging a Transformer model, it made real-time predictions on collected temporal signal sequences, achieving passive real-time object positioning. The study compiled a dataset from sequential signals generated by 24 sets of laser tube pair arrays. It compared the predictive performances of methods including Transformer, LSTM, GRU, and CNN + GRU. Results show that the proposed system attained a positioning accuracy of 0.2 cm and a positioning delay of 4.74 ms, demonstrating superior attributes: high precision, sampling rate, low latency, robust interference resistance, and cost-effectiveness. This provides a viable and effective solution for real-time object positioning.