digi:arm-embedded:linux:dey:deycustomrootfs:ros2 [Digi产品中文Wiki]

DEY是以Yocto的方式开发定制嵌入式Linux系统，这意味着除了添加Meta-Digi外，您也可以添加其它Yocto项目组的layer，包括在生产自动化和物流行业常用ROS机器人操作系统。在github中我们可以看到meta-ros在一直保持开发和演进状态。

Yocto 4.0 Kirkstone (LTS)对应支持ROS2 Humble (LTS)是个长期支持的版本，也是我们要集成的对象。官方文档位于：https://docs.ros.org/en/humble/。更确切地说，在嵌入式Yocto源代码层面安装的文档是： https://github.com/ros/meta-ros/wiki/OpenEmbedded-Build-Instructions

1. 预安装包查缺补漏通常在安装DEY时，应该已经把大部分依赖的安装包都安装好了，但为了防止ROS需要一些额外的包，参考文档，再执行一下它的依赖包安装：

sudo apt -y update 
sudo apt -y install gawk wget file git diffstat file unzip texinfo gcc gcc-multilib build-essential chrpath socat cpio python3 python3-pip python3-pexpect python3-distutils python3-subunit xz-utils debianutils iputils-ping python3-git python3-jinja2 libegl1-mesa libsdl1.2-dev pylint xterm  mesa-common-dev g++-multilib locales lsb-release time zstd locales liblz4-tool
sudo locale-gen en_US.utf8
sudo apt -y install python-is-python3

2.安装DEY ALL In One开发环境，本开发环境会把meta-digi和meta-ros都和官方同步，用户只需在配置文件对镜像进行配置。首先要安装repo，在ubuntu 22.04中可直接用apt install repo来安装，如果您的发行版不支持这种安装方式，请从官方下载安装repo:

sudo -H tmux
curl -o /usr/local/bin/repo http://commondatastorage.googleapis.com/git-repo-downloads/repo
chmod a+x /usr/local/bin/repo
exit

然后设置好git

git config --global user.name  “yourname”   
git config --global user.email "you@email.com“

接着就可以用repo安装DEY AIO了

cd
mkdir dey-aio
cd dey-aio
repo init -u https://github.com/peyoot/dey-aio-manifest.git -b china
repo sync

3. 配置ROS支持

我们还需要把这个layer添加到conf/bblayers.conf中，在文件的适当位置插入这三句：

  /home/robin/dey-aio/dey4.0/sources/meta-ros/meta-ros-common \
  /home/robin/dey-aio/dey4.0/sources/meta-ros/meta-ros2 \
  /home/robin/dey-aio/dey4.0/sources/meta-ros/meta-ros2-humble \

并在bblayers.conf的配置文件最后添加这两行变量定义：

ROS_DISTRO = "humble"
ROS_DISTRO_TYPE = "ros2"

此时可以尝试编译ros-core

bitbake ros-core

5. 检查机器学习相关包

cat ../../sources/meta-digi/meta-digi-dey/dynamic-layers/meta-ml/recipes-fsl/packagegroup/packagegroup-imx-ml.bbappend
最新的bb文件应该如下所录，使用tensorflow-lite，如果不是请按下面注释掉原来的一行，并添加这一新行
#ML_PKGS:mx9-nxp-bsp:remove = "deepview-rt-examples"
# Only install tensorflow-lite to save space
ML_PKGS:mx9-nxp-bsp:mx93-nxp-bsp = "tensorflow-lite"

另外，在conf/local.conf中添加机器学习包 IMAGE_INSTALL:append = “ packagegroup-imx-ml”

6. 创建带ros支持的镜像

如果要编译ros的镜像，在meta-ros-common中有image相关的recipe，其中： cat ros-image-core可以观察，要定义的东西：

require ${COREBASE}/meta/recipes-core/images/core-image-minimal.bb

SUMMARY = "A small image just capable of starting core ROS."
DESCRIPTION = "${SUMMARY}"

inherit ros_distro_${ROS_DISTRO}
inherit ${ROS_DISTRO_TYPE}_image

IMAGE_INSTALL:append = " \
    ros-core \
"

从上面可以看出，和core-image-base的bbappend类似，这个ros-image-core也是参考poky/meta/recipes-core/images的基础镜像core-image-minimal添加一些包。

core-image-base和core-image-minimal的差别在于：core-image-base使用meta/classes/core-image.bbclass中的IMAGE_INSTALL定义的包，而core-image-minimal则在bb文件里定义来覆盖原来core-image下的包，移除了packagegroup-base-extended，得到一个更小的镜像。所以，ros2的镜像，应该以core-image-base为基础才有意义。

从网络找到的一些参考： https://gitee.com/zxd2021-imx/meta-robot-platform/blob/Hardknott-5.10/imx/meta-robot/recipes-core/images/imx-robot-core.bb

5. 定制core-image-ros 我们给镜像取这个名字，是因为core-image-base能支持cc93的BSP，我们只需加入ros支持即可，这个镜像的bb文件可以放在meta-digi/meta-digi-dey下的recipe-core/images。先一步一步来，从基础的core-image-base中添加ros相关的支持。

首先测试一下没加layer前的效果，也就是bb文件在meta-custom是否能正常编译，将core-image-base.bbappend文件复制出来，重命名为core-image-ros.bb，并加上一句inherit core-image，整个过程如下

cd sources/meta-custom
mkdir -p recipe-core/images
nano core-image-ros.bb

#
# Copyright (C) 2016-2022 Digi International.
#

IMAGE_FEATURES += " \
    dey-network \
    eclipse-debug \
    ssh-server-dropbear \
    ${@bb.utils.contains('DISTRO_FEATURES', 'gstreamer', 'dey-gstreamer', '', d)} \
    ${@bb.utils.contains('MACHINE_FEATURES', 'alsa', 'dey-audio', '', d)} \
    ${@bb.utils.contains('MACHINE_FEATURES', 'bluetooth', 'dey-bluetooth', '', d)} \
    ${@bb.utils.contains('MACHINE_FEATURES', 'wifi', 'dey-wireless', '', d)} \
"

# Remove graphical packages for non-graphical platforms
IMAGE_FEATURES:remove = "${@oe.utils.conditional('IS_HEADLESS', 'true', ' splash ', '', d)}"

CORE_IMAGE_BASE_INSTALL += "dey-examples-digiapix"

# The connectcore demo was removed from 'packagegroup-dey-core' for the
# 6UL (because of rootfs space limits). Add it here, to install it in the
# non-graphical core-image-base.
CORE_IMAGE_BASE_INSTALL:append:ccimx6ul = " connectcore-demo-example"

# SDK features (for toolchains generated from an image with populate_sdk)
SDKIMAGE_FEATURES ?= "dev-pkgs dbg-pkgs staticdev-pkgs"

# Add our dey-image tweaks to the final image (like /etc/build info)
inherit core-image
inherit dey-image

可以成功编译，接下来就是修改这个bb文件，加入ros相关的包。

DESCRIPTION = "A Robot image with base-featured + ROS core Linux system functionality installed."

LICENSE = "MIT"


IMAGE_FEATURES += " \
    dey-network \
    eclipse-debug \
    ssh-server-dropbear \
    ${@bb.utils.contains('DISTRO_FEATURES', 'gstreamer', 'dey-gstreamer', '', d)} \
    ${@bb.utils.contains('MACHINE_FEATURES', 'alsa', 'dey-audio', '', d)} \
    ${@bb.utils.contains('MACHINE_FEATURES', 'bluetooth', 'dey-bluetooth', '', d)} \
    ${@bb.utils.contains('MACHINE_FEATURES', 'wifi', 'dey-wireless', '', d)} \
"

# Remove graphical packages for non-graphical platforms
IMAGE_FEATURES:remove = "${@oe.utils.conditional('IS_HEADLESS', 'true', ' splash ', '', d)}"

CORE_IMAGE_BASE_INSTALL += "dey-examples-digiapix"


# SDK features (for toolchains generated from an image with populate_sdk)
SDKIMAGE_FEATURES ?= "dev-pkgs dbg-pkgs staticdev-pkgs"

# Add our dey-image tweaks to the final image (like /etc/build info)

inherit core-image
inherit dey-image
inherit ros_distro_${ROS_DISTRO}
inherit ${ROS_DISTRO_TYPE}_image

IMAGE_INSTALL:append = " \
    ros-core \
"

新建一个项目myrobot，编译首个core-image-ros，并排除错误。

bitbake core-image-ros

其它可能有用参考：https://community.nxp.com/t5/i-MX-Processors/I-MX8M-plus-AI-Robot-platform/m-p/1303809#M176520

ROS2机器人操作系统的DEY镜像定制