Autonomous vehicles, industrial robots, drones need to “sense” the environment and react to changes in real time. The data from sensors needs to be processed instantly to enable such response. Low control latency and low power consumption are key requirements to make these devices truly autonomous and safe. Traditional computing architectures process all the input data regardless of context which can be extremely inefficient at the edge. In addition, moving data between memory and the processor introduces high latency and power penalties. 

At GrAI Matter Labs, we developed a neuromorphic computing architecture for sensor analytics and machine learning, inspired by the biological brain. It enables real-time response for autonomous action. The unique combination of a Dataflow architecture and near memory compute exploits all forms of sparsity at the edge and can deliver autonomy at low power. It is fully programmable with standard frameworks and languages.