Some questions of micro-espectre

[Leeeeeeedw]

Hello,

First of all, thank you for your contributions to this amazing project. I set up the ESP32-S3 with micro-espectre today, and it is now running normally, the detection effect is great.

But, Regarding ESP32-S3 data transmission and multi-ESP32-S3s setup, I have a few questions that I would like to clarify, and I appreciate your help in advance!

Can the ESP32-S3 run independently after disconnecting from the PC via the COM port in the micro-espectre project? When I run the command "./me run", the device first connects to a specific COM port. If it could connect to the PC through the router's LAN IP without connecting to the COM port, I believe this would greatly improve flexibility.
Furthermore, does the micro-espectre project support connecting multiple ESP32-S3 devices to achieve better detection performance now or in the future? I am very much looking forward to this feature!

BR

[francescopace]

Hi @Leeeeeeedw! Thank you for testing micro-espectre!

At the moment, I use micro-espectre mainly as an R&D/prototyping platform. In my current workflow, the board is typically managed through serial (flash/deploy/run), so this setup is intended for research and experimentation.

If you need a fully autonomous sensor (with no serial dependency in normal operation), I recommend the C++ ESPHome version of ESPectre, which is designed for production-style standalone deployments.

That said, from a technical perspective, MicroPython autostart is possible by configuring startup scripts (boot.py / main.py), so autonomous boot is feasible in theory: https://docs.micropython.org/en/latest/reference/reset_boot.html

Regarding multiple-device support:

• You can already use multiple devices as independent nodes (each with its own MQTT topic/client ID).
• Native multi-node sensor fusion (combining multiple ESP32 streams into one unified detection result) is not available yet, but it is a very interesting direction. If you want to collaborate in this area, I would be happy to review a PR.

[Leeeeeeedw]

Hi @francescopace! Thank you for your detailed reply~

I have successfully set up the ESPHome version of ESPectre using a single ESP32-S3 and encountered some issues, which are also mentioned in issues #119, and you replied to me. I will test it tomorrow based on your suggestions.

Also, does the ESPHome version currently not support native multi-node sensor fusion? I know that HA can connect multiple sensors, but the sensors themselves do not support multi-node fusion, is my understanding correct?

Regarding micro-espectre, I agree with your point that it is "mainly used as a R&D platform," and I will check the MicroPython auto-start thread. As for multi-node sensor fusion, I have only seen some concepts in the RuView project, but I did not get the desired detection results, so I am not sure if RuView has implemented a true fusion algorithm. When I have more free time, I will study this fusion topic, and if I find some technical ideas, I will discuss them with you!

[francescopace]

One important note: the repository you referenced (RuView) has been flagged as a scam/misleading project, so I would strongly recommend avoiding using it as a technical reference or baseline. Scam Report

If you end up exploring fusion approaches, that would be very interesting — especially if you look into time synchronization + probabilistic merging (e.g., Kalman filters or similar lightweight methods). Happy to discuss any ideas when you get there.

Looking forward to your test results tomorrow 👍

[Leeeeeeedw]

Yes, I've been testing RuView for weeks with no results. Last week I finally discovered it's a scam/misleading project, and the link is the one you mentioned https://github.com/deletexiumu/wifi-densepose.

I just think multi-node sensor fusion (the concept in the RuView project, which, of course, misled and fooled many people and was never actually implemented) might be a future technological direction, providing multi-view detection.

I hope our team can launch a research project on this topic in the near future, so I can dedicate more time to it and discuss it with you then!

Also, I need to confirm with you whether the statement above that "HA can connect multiple sensors, but the sensors themselves do not support multi-node fusion" is true. Thank you!

I suddenly remembered I have many meetings tomorrow, so I'll test it ASAP and give feedback~

[francescopace]

I think it helps to separate "multi-node sensor fusion" into two scenarios:

1. What works today

• You can use multiple ESPectre devices as independent nodes.
• Home Assistant can aggregate their outputs (automations, templates, Bayesian logic, dashboards).
• In this scenario, each ESP32 performs detection locally and publishes its own result.

2. What can be done tomorrow

• True multi-node fusion means combining CSI data across devices (time sync + cross-node signal processing) to produce one unified estimate.
• This is not implemented yet in ESPectre today.
• It is a very interesting direction and already on the roadmap as exploratory work.

So the statement is: HA can combine multiple sensors at decision level, but native cross-device CSI fusion is not available yet.

[Leeeeeeedw]

Thank you so much for your detailed technical explanation, I understand it much better now.

I will continue to conduct deployment testing and research on related methodologies based on the ESPectre project, and I look forward to its future development!