Split-Inference Architecture for Device-Centric Radio Sensing in a Networked Robotics Scenario

In this paper, we present an architecture for neuromorphic device-edge co-inference, with application to radio sensing for environment mapping in networked robotics scenarios. The developed demonstration setup integrates a radar sensor, radar preprocessing unit and neuromorphic processing unit (NPU) implemented on-device, while the sensing task (object detection, tracking and environment mapping) is completed at the edge server (access point).