Architecture Advice for Large-Scale Radar-Based Volume Estimation

Hello! I am working on an IoT project where I need to use 100 radars on a dome surface to estimate the volume of what is inside.

Due to my background in robotics, I am familiar with micro-ROS and ROS 2, and theoretically they could work. However, in practice I am unsure about the system architecture.

One possible approach is to have the radars communicate via CAN-SPI with Teensy boards running as micro-ROS clients. Ideally, these microcontrollers could be connected via LAN, and a PC would act as the agent.

Another approach is to connect 8 radars per Teensy and use 3 or 4 Teensy boards with a Jetson device acting as the agent. Then, I would have a Jetson-based network where a PC subscribes to all topics and performs the volume computation.

My concern with the second option is that it’s more centralized—if one Jetson fails, the data from 24 radars would be lost. On the other hand, I am not sure how to implement and test a global LAN setup with one agent and many clients.

Any feedback or suggestions would be greatly appreciated, since this project is outside my usual area of expertise.

Thanks!