search

azinke / mmwave

Linux tool for controlling the MMWCAS-DSP-EVM and MMWAVECAS-RF-EVM boards
MIT License
24 stars 5 forks source link

Realtime demo #3

Open andraspalffy opened 1 year ago

andraspalffy commented 1 year ago

Dear @azinke,

thank you for this repository, much needed. Did you by any chance have a look at the real time options? Seems like a hassle to start even on windows:

https://e2e.ti.com/support/sensors-group/sensors/f/sensors-forum/882929/awr2243-the-appimage-form-sdk3-7-and-sdk3-8-can-not-running-in-cascade-radar-with-awr2243/3344366#3344366

Would be really nice to have the board working with ROS for example, even if it is just the point cloud.

azinke commented 1 year ago

Hi @andraspalffy,

Thank you very much! I'm not sure if the default app running on the DSP board supports real-time streaming. My current understanding is that one needs to write its own firmware to be able to stream point clouds directly.

I haven't looked into this yet but would definitely be interested in it; because it would then be possible to test the MMW-CAS board in various use real-time cases

andraspalffy commented 1 year ago

To be honest even point cloud streaming would be a nice start - it is a bit shame that such a big company does not have such a basic demo ready to run :\

chuber1986 commented 1 year ago

Hi @andraspalffy,

recently I ask in the TI forum if it is possible to somehow get life data from the board. The answer was not uplifting. I guess the Network interface is to slow to process that amount of data and the DSP board is to underpowered to do real-time processing. There the referred me to a Retina device from "Smart Radar System", which has it's own downsides (no raw ADC data, no beamforming, ...). (https://e2e.ti.com/support/sensors-group/sensors/f/sensors-forum/1179807/mmwcas-dsp-evm-and-mmwcas-rf-evm-live-data-processing)

@azinke May I use the opportunity to ask you 2 Question regarding the DSP board?

  1. Do you know if there is a way to install or even (cross-)compile software for the board? And how?
  2. Do you know who the time synchronization between the DSP board and other sensors can be achieved with reasonable precision?

BR Christian

andraspalffy commented 1 year ago

@chuber1986 Thanks a lot! What do you mean by no beamforming?

azinke commented 1 year ago

Hi @chuber1986 ,

  1. I think the following SDKs are a good starting point. I haven't tried myself yet but I'll definitely be giving it a try. My current expectation is that it would be possible to write a radar signal processing pipeline directly on the DSP board. And eventually real-time processing

As you pointed out, there would be little to no support from TI for writing firmware. So, I think one needs to navigate through the codes and libraries on its own. Which could take some time

  1. Very good question about the synchronization with other sensors. I found it difficult with the software trigger, at least with mmwave as it is now (always configure + record) to achieve good timing. However,

    • One option would be to use the software trigger but with the possibility to have the record option separate
      • mmwave -d <capture-directory> --record
      • And then trigger all the sensors together from a single script (or program)
      • They would be some delay but would already be better than what we have now
    • The second option, probably the best, would be to use a hardware trigger for the master chip as well
chuber1986 commented 1 year ago

It is not possible to operate the Smart Radar System devices in an Tx beamforming mode as you can do with the TI board.

andraspalffy commented 1 year ago

Thanks. I guess it depends on the usecase, for me real time point cloud data would be already a win, RAD data would be tremendous success, synced data would be heaven :)

For syncing PTP would be the best.

chuber1986 commented 1 year ago

Hi @azinke ,

thanks for your response! Thanks for the links, these already help a lot.

To what I've read so far the HW trigger discussion refers to the interface between RF and DSP board and not an separate GPIO pin in the board, so it's not possible to use it without discarding the DSP. I also think that PTP would be the best option and I hope I can get it to run on the TI board.

I was just hoping someone has an idea how they did it for the ColoRadar dataset, where they obviously hat to synchronize multiple sensors.