Presentation Information
[N-2-21]Classification of LiDAR data using photonic reservoir computing based on a semiconductor laser
◎Ryotaro Kibe1, Keisuke Kase1, Felix Koester1, Tomoki Yamagami1, Atsushi Uchida1 (1. Saitama University, Department of Information and Computer Sciences, Uchida Laboratory)
Keywords:
Reservoir computing,Semiconductor lasers,LiDAR
In recent years, neural networks have demonstrated high performance in areas such as image recognition and natural language processing; however, recurrent neural networks (RNNs), which are used for time-series processing, require significant computational resources to learn connection weights. In contrast, reservoir computing reduces computational load by training only the output layer, thereby enabling high-speed processing. Furthermore, in optical sensing—including LiDAR—it is necessary to process vast amounts of point cloud data in real time, yet computational resources on the sensors are limited.Therefore, in this study, aiming to reduce latency in both data acquisition and processing, we created a dataset using 3D LiDAR and performed object classification by processing it with photonic reservoir computing based on backlight phase modulation. By combining distance data with reflected light intensity and performing data compression through averaging, we achieved a high classification accuracy of 99.56%.