Memristor-based adaptive neuromorphic perception in unstructured environments

Efficient operation of control systems in robotics or autonomous driving targeting real-world navigation scenarios requires perception methods that understand and adapt to unstructured environments with high accuracy, adaptation, and generality. This work presents a memristor-based differential neuromorphic computing, perceptual signal processing, and online adaptation method that provides neuromorphic-style adaptation to external sensory stimuli. Adaptation ability and generality are validated in two scenarios: object grasping and autonomous driving. In grasping, a robot hand achieves safe, stable grasping via fast (~1 ms) adaptation based on tactile object features using a single memristor. In autonomous driving, decision-making information across 10 unstructured environments is extracted with 94% accuracy using a 40×25 memristor array. By mimicking human low-level perception mechanisms, the electronic neuromorphic circuit-based method achieves real-time adaptation and high-level reactions to unstructured environments.