Development kit for the AI deck

Developing for the AI Deck of the nano-quadcopter Crazyflie requires programming three different chips, each requiring its own toolchain, debug setup and SDKs:

• the GAP8 chip by Greenwaves Technologies, using the GAP8 SDK currently at v 3.9.1,
• the NINA-W102 chip by ESP -- you'll need the SDK at v 3.3.1 -- and
• the STM32F405 on the quadcopter itself, using the ARM toolchain

Installing all this in parallel on the same machine easily leads to a lot of trouble with different python versions, incompatible packages etc.

This repo provides

• a dockerfile for each chip to build a self-contained development environment for the respective chip
• example bash scripts how to run and use these docker images
• hints on how to integrate these docker images with VS Code to make debugging a breeze.
• example hook files if you want to use these images in a Github CI setup

Dependencies

You need Docker and a standard Linux machine. This was tested and set up on Ubuntu 20.04.

Creating the docker images

You can fork this repo and set up your own builds on Docker Hub, or you can use the script create.sh to build your own images. If you don't want to build your own, you can download them from the profile simatai on Docker Hub using

docker pull simatai/stm-arm:1.1
docker pull simatai/espidf:3.3.1
docker pull simatai/gapsdk:3.9.1

Getting into a docker image and using the toolchain

Use the bash scripts in bin to log into a container based on the images. Make sure to update the USB port in the scripts depending on where your chip programmer is connected.

Check the respective dockerfile to understand where everything was installed inside the different images. In general, each image contains the SDK for the chip and an appropriate version of OpenOCD and GDB. The working directory for every image is /module.

Installing the autotiler for GAP8

If you want to use the autotiler for the GAP8 chip, you need to accept the license conditions by Greenwaves. This is done inside the container.

Open up the container to install the auto tiler

export GAP_SDK_VERSION=3.9.1
docker run --rm -it gapsdk:${GAP_SDK_VERSION} /bin/bash

Then in the container write:

cd /gap_sdk
source configs/ai_deck.sh
make autotiler

This will install the autotiler, which requires you to register your email and get a special URL token to download and install the autotiler.

In a separate terminal on your local machine, commit the changes to the image by first looking up the container ID status:

docker ps

Copy and past the container ID and replace the on the line here below, then run it in the separate terminal (also adapt the SDK version if you did before)

export GAP_SDK_VERSION=3.9.1
docker commit <container id> gapsdk:${GAP_SDK_VERSION}

This will save the install of the autotiler on your image. You can close the container in the other terminal with 'exit'. This step cannot be automated in my opinion -- if you know how that could work, let me know.

Setting up VS Code

Refer to the VS Code tutorial on developing inside containers for general pointers on how to use VS Code with the Docker images set up here.

Setting up Github CI

You can use the docker images above to set up continous integration in your github repo. The example yaml files in the ci folder show how to do that.

Credits

This work has heavily borrowed from work by the Bitcraze team. The stm-arm image, used for programming the STM32F405 chip on the Crazyflie itself, is a modified version of the build image by Bitcraze -- in addition to the Bitcraze version it includes OpenOCD for STM chips to enable debugging from the docker image; but it excludes the toolchains for the cflib.