BrainChip Holdings (ASX:BRN) has been awarded a US$1.8m contract with the US Air Force Research Laboratory (AFRL) to develop neuromorphic radar signal processing algorithms. The deal is part of the Small Business Innovation Research program.
The total contract amount will be paid over the 12-month term of the agreement.
The tech
BrainChip specialises in the development of “neuromorphic” computing. Unlike traditional processors (like CPUs and GPUs), which process information sequentially, neuromorphic processors aim to mimic the way brains function.
BrainChip’s flagship product is the Akida 2.0 processor, which has military, autonomous vehicle, consumer electronics and industrial IoT applications. The chip allows such devices to process data themselves (called “edge computing”) rather than sending it to a cloud server. This approach allows for faster response times, better privacy and significantly lower energy consumption.
Traditional AI models require large, pre-trained neural networks that must be fed huge datasets to “train” them to recognise patterns. The Akida 2.0 chip, by contrast, allows for continuous learning and can adapt to new data on the fly. This is especially useful in dynamic environments.
The TENNs (Temporal Event Neural Network) algorithm framework is designed to work with Akida and focuses on time-based data processing, meaning it excels with video, audio, radar and sensor data.
Contract details
The project will focus on developing and optimising radar processing algorithms tailored to BrainChip’s proprietary hardware.
The SBIR contract, titled “Mapping Complex Sensor Signal Processing Algorithms onto Neuromorphic Chips”, builds on earlier work in which a multinational aerospace and defence customer successfully demonstrated the feasibility of running radar processing algorithms on BrainChip’s hardware.
The current contract will apply this work to a specific type of radar processing known as micro-Doppler signature analysis, which enables the discrimination of different types of activity by analysing subtle movement patterns. For instance, the processing could detect whether a person is walking, running or standing still.
This capability has potential applications in military, robotics and space-based platforms.
Shares closed 11.36% higher at 24 cents yesterday.