Bio-Plausible Collision Detection in Micro-Robots: A Comparative Study of LGMD-Inspired Models

Reliable collision detection is a core challenge in the field of autonomous micro-robots. Although LiDAR-based or deep-learning-based visual algorithms exhibit high-precision advantages, their computational complexity and energy consumption are difficult to adapt to resource-constrained micro-robots. This study implements an on-line horizontal comparison of three representative LGMD-inspired models (inspired by the locust's Lobula Giant Movement Detector neuron) on a lightweight, low-cost, low-power micro-robot platform. Despite differences in their performance, all three models demonstrate the feasibility of bio-plausible collision detection systems in resource-constrained scenarios. Their low computational-load and high energy-efficiency characteristics provide practical verification for embedded dynamic obstacle detection algorithms, demonstrating the engineering applicability of bio-plausible visual models in micro-robotic platforms.