Hardware Implementation of Ultra‐Fast Obstacle Avoidance Based on a Single Photonic Spiking Neuron

Abstract

Visual obstacle avoidance is widely applied to unmanned aerial vehicles (UAVs) and mobile robot fields. A simple system architecture, low power consumption, optimized processing, and real‐time performance are extremely needed due to the limited payload of some mini UAVs. To address these issues, an obstacle avoidance system harnessing the rate encoding features of a photonic spiking neuron based on a Fabry–Pérot (FP) laser is proposed, which simulates the monocular vision. Here, time to collision is used to describe the distance of obstacles. The experimental results show that the FP laser excites ultra‐fast spike responses in real time for the following cases, facilitating the generation of control commands by motor neurons to realize accurate decision‐making. Four cases of mobile obstacle avoidance scenarios, including “Constant velocity approach”, “Approach and retreat”, “The motion state involving stays”, and “Approach with different velocities”, and obstacle avoidance problems with multiple stationary obstacles appearing simultaneously are experimentally analyzed. The system exhibits a spike response rate of up to 5 GHz. This work proves the feasibility of applying the ultra‐fast photonic obstacle avoidance system to UAVs and other fields in the future and highlights the potential of photonic neuromorphic processor platforms.

Photonic Reservoir Computing System for Pattern Recognition Based on an Array of Four Distributed Feedback Lasers

“…However, such an architecture is inefficient for computational models that are distributed, massively parallel, and adaptive, most notably those used for neural networks in machine learning (ML). − ML is an attempt to achieve a human-level approach to tasks that are challenging for traditional computers but easy for humans. Furthermore, the human brain, which could be abstracted as a neural network, is a dynamic system characterized by high parallelism and adaptability. − Reservoir computing (RC), as a ML approach based on the principles of dynamic system theory, attempts to mimic these neural network characteristics, although its structure is relatively simpler compared to the human brain. − Compared with those of other ML models, the input and reservoir weights of RC are randomly generated and fixed. Only the output weight, applied to connect the reservoir and output layer to implement RC, needs to be trained, which greatly reduces the difficulty and computational complexity of model training. , …”