Bio-Inspired Sensor Predicts Object Movement with Photo Memristors

Key Highlights

• The bio-inspired sensor integrates sensing, memory, and processing in a single device, allowing it to detect motion accurately and predict trajectories.
• Using an array of photo memristors, the sensor can record a dynamic memory of preceding instants, making it more energy-efficient and compact than current motion detection systems.

Researchers at Aalto University have developed a new bio-inspired sensor to detect moving objects in a single video frame and predict their movements. This technology could be useful in various fields, including dynamic vision sensing, autonomous driving technology, industrial process control, automatic inspection, and robotic guidance.

The Integration of Sensing, Memory, and Processing in a Single Device Enables Compact and Efficient Motion Detection

Current motion detection systems need numerous components and complex algorithms for frame-by-frame analyses, making them inefficient and energy-intensive. Inspired by the human visual system, the new technology integrates sensing, memory, and processing in a single device, enabling it to detect motion and predict trajectories.

At the heart of the technology is an array of photo memristors, electrical devices that produce an electric current in response to light. Photo memristors can effectively “remember” whether they’ve been exposed to light recently, which means that a sensor made from an array of photo memristors can record instantaneous information about a scene and a dynamic memory of the preceding instants.

1. The technology can integrate a series of optical images in one frame, with the information of each image embedded in the following images as confidential information.
2. The researchers can detect motion earlier in the video by analyzing only the final frame with a simple artificial neural network. 
3. As a result, the sensor is compact and efficient.

The Potential Impact of the Bio-Inspired Sensor on Self Driving

In tests, the researchers showed that their sensor could recognize motion and predict the next frames with nearly 100% accuracy. The technology could help self-driving vehicles accurately predict how objects will move and guide their decisions.

Their compact in-sensor memory and computing solution could provide new opportunities in autonomous robotics and human-machine interactions. The in-frame information attained in the system using photo memristors avoids redundant data flows, enabling energy-efficient decision-making in real time.