Get Started Guide

  • 2021.3
  • 10/21/2021
  • Public Content

Release Notes

EI for AMR 2021.3 does not include the latest functional and security updates. EI for AMR 2021.4.1, released in December 2021, includes additional functional and security updates. Customers should update to the latest available version.

EI for AMR Features

Fast Mapping Algorithm
  • Removed -march compiler option
  • Fixed issue with pose not provided by the SLAM algorithm in the tf tree
  • Minor fixes for ROS 2 nodes
Wandering AI Algorithm
  • Added Gazebo* simulation for an industrial room to run with the Wandering application
  • Updated navigation and scan parameters to improve obstacle avoidance
  • Added additional launch files for: integrated_navigation, pengo_wander_integrated, integrated_rtabmap
  • Added callback navigation time
  • Optimizations introduced on MapEngine to improve movement
  • Decreased robot rotation speed to allow the map to get updated
  • Added option to move the robot to the furthest free cell
  • Added rviz2 configuration file
  • Pass robot radius as ROS argument to Wandering node (radius size also updated)
AMR Containers
  • Added the requirement to upgrade the kernel to Intel’s Linux* LTS kernel, version 5.10.47
  • Added the OpenVINO™ model_server demo client
  • Updated with an extra parameter with a predefined set of users that have different permissions (you can run the script from any location)
  • Added the evo tool for measuring location accuracy from the SLAM algorithm
  • Introduced code optimizations to reduce the containers size
  • Updated the Intel® RealSense™ SDK to version 2.48 and the ROS 2 wrapper to version 3.2.2
Intel® Edge Software Device Qualification (Intel® ESDQ) for Autonomous Mobile Robots (AMR)
  • Initial release of the Intel® Edge Software Device Qualification (Intel® ESDQ) for Autonomous Mobile Robots (AMR). For more information, go to the Intel® ESDQ for AMR article.

Known Issues/Limitations

  • Due to the End of Life (EOL) of the LIDAR L515 camera, L515 is not supported anymore by EI for AMR
  • During the Docker* build process, packages are installed from public apt repositories. Intel has no control over the public repositories. Specific versions of packages might be removed by the owners at any time, which might break the docker image build. The Docker* build targets the latest available versions of software packages from the public repositories but always keeping the same major version.
    Between EI for AMR releases, it is possible that packages in public apt repositories get upgraded to newer versions. Although it is possible for these upgraded software packages to work without issues with the current EI for AMR release, you assume all risks associated with the use of the upgraded packages.
    Contact your Intel representative if you encounter issues that are a result of using different versions of software packages from public apt repositories than the ones provided with this release.
    In the EI for AMR installer package, EIforAMR_2021.3_apt_components_list.pdf contains the complete list with all software packages from public apt repositories used to validate this release.
  • It currently takes approximately 80 minutes to install the EI for AMR package. The time varies depending on the speed of your internet connection and system’s capabilities.
  • The installed TensorFlow version in the AMR SDK contains AVX instructions. These AVX instructions are not supported by the Intel® Atom® CPU like the CPU in the Elkhart Lake platform. Any action, including Run OpenVINO™ Sample Applications in Docker* Container, fails on a platform with an Intel® Atom® CPU. To be able to run TensorFlow on an Intel® Atom® CPU, it must be re-compiled without the AVX instructions using the steps from: How to Build and Install the Latest TensorFlow Without CUDA GPU
  • While navigating, the Robot loses or resets the costmap it previously created. Nav2 path planner creates a path through the wall and hits it. Known bug in Nav2 and similar to what has been reported at: No fix for ROS Foxy is available. Please contact Intel® Developer Catalog Support Forum to get assistance with fixing this issue.
Wandering AI Application
  • RTAB_MAP is not best suited for indoor navigation, therefore it might happen that some obstacles are not detected with the highest accuracy, due to reflections, etc.
  • The object detection node does not filter visualized objects on ROS /camera/detected_objects Topic. This results in having the same object reported multiple times.

Release Content

Subproject (component)
EI for AMR Containers
Fast Mapping
Wandering AI Application
EI for AMR Simulations

Product and Performance Information


Performance varies by use, configuration and other factors. Learn more at