Neuromorphic 'retina' hardware aims to speed up driverless car vision
Feb 27th 2026
A team including Northeastern professor Ravinder Dahiya developed synaptic-transistor based vision hardware that mimics the human retina to cut perception latency, showing major speed gains in simulations but facing industry hardware hurdles.
- Researchers led by Northeastern professor Ravinder Dahiya published a Nature Communications paper describing retina-inspired sensors built with synaptic transistors.
- The device mimics human temporal motion cues to focus processing on areas of change rather than full-frame images.
- In simulations of autonomous driving and robotic arms the hardware produced a 400% increase in processing speed compared with traditional image processing systems.
- Synaptic transistors emulate neural pathways to reduce the time between detecting an object and taking action.
- Possible uses include robotaxis, smart glasses and industrial robots that need fast, real-time perception.
- Wider adoption may be slow because commercial analog neuromorphic hardware is not yet widely available and the work remains mostly academic.