QA about cognifly-python's functions and features

[lida2003]

As readme suggests usages about cognifly-python
a) Manual control
b) Python API (Pro API/School API/Streaming)

Google AIY Vision Kit (Camera + Vision Bonnet) ----> RPI Zero W <--- TTL ---> FlightControl <--- TTL ---> Matek 3901

As we know if we need school API for controlling the iNav/BF quads, there is a lot of work to do, especially pull data out of the FC and build a postion estimater.

So my question as follows:

1. All school API (which tello does), especially "forward/backward/up/down/left/right/flip/ccw/cw" is tested on RPI right?
2. API is same with tello Tello API right ?

Just port changes

1. tello (8889) ==> cognify(8988)
2. tello (8890) ==> cognify(8989)
3. tello (11111) ==> cognify(8990)

3. Where is UDP server (listens command) and UDP client (sends status back) , is it in the code of cognifly-python?
   ==>cognifly-python, handles PS4/keyboard/python API
   ==>YAMSPy, handles MSP protocol throught TTL
   ==>AIY-Vision-Kit-Utils, handle video analysis

[yannbouteiller]

Hello,

You are right, (1) all the school API is tested on RPI, and (2) the school API largely mimics the easytello API.

The version that is currently under development on the repo (yb/stable branch) works with INAV 6.

(3) the UDP logic is in the code of cognifly-python. The client/server per-se is defined here

[lida2003]

Really glad to see quick response.

Further question here.
Does "forward/backward/up/down/left/right/flip/ccw/cw" tested on RPI, with proper env, has (more or less, about)1cm drift?

[yannbouteiller]

It depends a lot on the conditions. If you are using the optical flow sensor over a well-textured and well-lit floor, yes. If using the optical flow sensor and the texture is bad, no, and the drone can drift badly. And if using a custom estimator with perfect position estimation and adapted PID values, there is no drift at all.

[lida2003]

Well, I'm planning to use lastest iNav (released version). And my preliminary test is based on iNav(5.0.0) + Matek 3901 module.

And now I'm going to use cognify softwares (RPI + Google AI camera) + iNav(5.0.0) + matek 3901 to setup the test. I think it should be much better than the previous test video

What do u think?

[yannbouteiller]

The 'yb/stable' branch is not compatible with inav5, only with the yet unreleased inav6. This is because this branch uses inav6 MSP messages to fake inav-required sensors that Cognifly does not use, or even possibly all sensors (except IMU) when using a custom estimator.

The Google stuff is not used by cognifly-python when using a RPI. It may simply use a raspberry camera to stream video, that the remote can then process in Python if they wish.

[lida2003]

It really helps a lot.

a) inav6 + yb/stable // MSP protocol compatible
b) Currently, video (AI) has nothig to do with cognifly control link // if video, no ai required, raspberry camera is good to go.