Autonomous Motion Vision with Tri-bulk-heterojunctioned Organic Adaptation Transistor

Motion vision is increasingly critical in scenarios such as autonomous driving and embodied intelligence. Existing sensors perform motion analysis and moving‑object extraction as separate tasks, which increases computation and storage needs to challenge highly efficient applications. Here, we design an organic tri-bulk-heterojunctioned adaptation phototransistor (OTAT) that integrates both functionalities in a single device. Notably, the OTAT incorporates a gate-adjacent semiconductor-insulator heterojunction as a photosensitive percolation channel, providing fine-tuned photoadaptation under illumination spanning six orders of magnitude. In imaging experiments, the device exhibits in‑sensor moving‑object extraction with a 45.3% highlight ratio, approaching the capability of the human visual system, while yielding motion analysis accuracy over 99.4% in both direction and speed detection. Importantly, the OTAT-based agents generated the lowest risk in multiple route-planning simulation. These results show that the OTAT effectively bridges the gap between motion extraction and analysis, marking a significant step forward for next-generation in‑sensor motion vision systems. Motion vision sensors typically perform motion analysis and moving-object extraction separately. Here, the authors develop an organic adaptation transistor that simultaneously extracts and analyses moving objects, achieving human-eye-like motion perception and improving safety in autonomous route-planning tasks.