自动驾驶域控制器nvidia环境搭建

nvidia安装ros和深度学习环境搭建步骤总结

#############todesk ################

sudo dpkg -i todesk_4.1.0_aarch64.deb

####################################

#############500g ##################

lsblk

# /dev/nvme0n1

fdisk /dev/nvme0n1

n

p

enter

enter 

sudo mkfs -t ext4 /dev/nvme0n1p1

sudo gedit /etc/fstab 

/dev/nvme0n1p1 /media/rosbag ext4 defaults 0 0

####################################

sudo apt-get update

sudo apt-get install python3-pip

JetPack:5.0.2

jetson-stats 

 sudo -H pip3 install -U jetson-stats

 jtop

sudo systemctl restart jtop.service

jetpack:

 a. sudo apt update  

 b. sudo apt upgrade

 c. sudo apt install nvidia-jetpack -y

 查看版本：/etc/apt/sources.list.d/nvidia-l4t-apt-source.list 文件中

######################ubuntu20.04 ros noetic install############################

sudo sh -c ‘. /etc/lsb-release && echo “deb http://mirrors.ustc.edu.cn/ros/ubuntu/ $DISTRIB_CODENAME main” > /etc/apt/sources.list.d/ros-latest.list’

sudo apt-key adv –keyserver ‘hkp://keyserver.ubuntu.com:80’ –recv-key C1CF6E31E6BADE8868B172B4F42ED6FBAB17C654

sudo apt update

sudo apt install ros-noetic-desktop-full

sudo pip install rosdepc // sudo pip3 install rosdepc

sudo rosdepc init

rosdepc update

echo “source /opt/ros/noetic/setup.bash” >> ~/.bashrc

source ~/.bashrc

tf2_sensor_msgs: 

 sudo apt-get install ros-noetic-tf2-sensor-msgs

sudo apt install libqt5serialport5-dev libudev-dev

sudo apt-get install ros-noetic-geographic-msgs 

vim ~/.bashrc:

 source /opt/ros/noetic/setup.bash

 export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/ros/noetic/lib

ros-numpy:

 sudo apt-get install ros-noetic-ros-numpy

catkin_make : catkin_make -DPYTHON_EXECUTABLE=/usr/bin/python3

catkin_make -DCMAKE_BUILD_TYPE=Release -DPYTHON_EXECUTABLE=/usr/bin/python3 

cuda: /usr/local

 a. CUDA 检查是否安装成功 

 nvcc -V

 如果报错，需要把nvcc添加到环境变量。

 vim ~/.bashrc

 export LD_LIBRARY_PATH=/usr/local/cuda/lib64

 export PATH=$PATH:/usr/local/cuda/bin

 source ~/.bashrc

 出现如下则表示安装正确：

 nvcc: NVIDIA (R) Cuda compiler driver

 Copyright (c) 2005-2022 NVIDIA Corporation

 Built on Wed_May__4_00:02:26_PDT_2022

 Cuda compilation tools, release 11.4, V11.4.239

 Build cuda_11.4.r11.4/compiler.31294910_0

安装过程

conda:

 root: sh Miniforge-pypy3-4.14.0-2-Linux-aarch64.sh

 conda create -n py38 -y

 sudo apt-get install libopenblas-base libopenmpi-dev libomp-dev

 conda activate py38 

pytorch:py38 ———copy torch-1.11.0-cp38-cp38-linux_aarch64.whl from usb

 pip3 install torch-1.11.0-cp38-cp38-linux_aarch64.whl

torchvision: ———–copy vision-0.12.0.zip from usb

 cd torchvision

 export BUILD_VERSION=0.12.0

 python3 setup.py install –user

 cd ../

pycuda:

 nvidia@tegra-ubuntu:/$ python3 -m pip install ‘pycuda<2021.1'

cupy:

 pip install cupy # only good use for (base)root

 if can not import cupy: pip uninstall numpy & pip install numpy==1.23.5 

(py38)root: python3

# import tensorrt

# import torch

# import pycuda

# import cupy

(base)root: python3

import cupy

import numpy

import pycuda

import 

untitled test:

 (base)root@tegra-ubuntu: catkin_make -DPYTHON_EXECUTABLE=/usr/bin/python3 

numpy:

 pip uninstall numpy 

 pip install -U numpy==1.23.5

rviz:

 frame: zvision_lidar1

protobuf 3.0.0 :

 gmock-1.7.0