MAPE-K loop based autonomy

Please refer to docments in the directory doc for the detailed design idea.

Build

This branch is based on the release 9 of OceanWATERS testbed (a version of noetic-devel branch. Details can be found in the release note of v2.0.0).

• It will first need to follow the "Get Started" instruction in nasa/ow_simulator to install software prerequisites.
• Create a catkin workspace mkdirs -p ~/oceanwater_ws/src
• pull down the ow9 branch of the repos (ow_simulator, ow_autonomy, ow_europa and irg_ope) from nasa-raspberry-si git organization into ~/oceanwater_ws/src. After checking out the ow9 branch of nasa-raspberry-si/ow_simulator, create a soft link .rosinstall inside ~/oceanwater_ws/src using the command below and then use wstool to update code bases.
  • ln -s ~/oceanwater_ws/src/ow_simulator/oceanwaters/workspaces/oceanwaters_ow9_rs.rosinstall ~/oceanwater_ws/src/.rosinstall
  • wstool update
• Pull down nasa-raspberry-si/autonomy into ~/oceanwater_ws/src
• Source ROS noetic envirionment
  • source /opt/ros/noetic/setup.bash
• Build using catkin
  • catkin build

Run

• Run ow_simulator in one terminal
  • cd ~/oceanwater_ws && source devel/setup.bash
  • roslaunch ow europa_terminator.launch
• Run ow_autonomy (actually an PLEXIL executive) in another terminal
  • cd ~/oceanwater_ws && source devel/setup.bash
  • roslaunch ow_plexil ow_exec.launch
• Run the Raspberry-SI autonomy in a new terminal
  • setup environment
    • cd ~/oceanwater_ws && source devel/setup.bash
  • Export three environment variables
    • export EVALUATION\_ROOT\_DIR=< evaluation directory >
    • export OW\_PLEXIL\_LIB\_SOURCE\_DIR=< PLEXIL library source directory >
    • export PLEXIL\_LIB\_COMPILED\_PLAN\_DIR=< the direcotry of compiled PLEXIL plan in devel dir >
  • Run with an example mission specification file, mission1.txt
    • roslaunch rs_autonomy rs_autonomy.launch mission_spec_filename:=mission1.txt
  • If the evaluation directory is decided to be inside this code base, run.bash script can be used to avoid the above two steps.

Code Organization

• Mission Control node and MAPE-K nodes (except knowledge node)
  • autonomy/src/mape-k
• Knowledge Node and other service nodesow8-rosnodes
  • autonomy/script
• Python packages for support services in autonomy/src
  • acknowledgement management (for knowledge node): maintain machine learning models and task-dependent runtime information.
  • prism planning: given some runtime info, automatically generate a PRISM model for a task and use it to get policy and actions
  • plan translation: based on the actions (i.e., high-level plan), generate a PLEXIL plan using a task-related template and then compile it
  • fault management: a ROS server to inject and clear arm fault in ow_simulator
• The organization of evaluation directory
  • An example of directory structure: autonomy/evaluation
  • For detailed description, please refer to autonomy/doc/structure_of_evaluation_directory.txt

Current Adaptation Cases

• List of cases
  • case 1: digging fails due to an incorrectly estimated excavation probability
  • case 2: the beliefs of supporting machine-learning models drop below predefined
  • case 3: surface vibration caused by earthquake
  • case 4: an arm fault is detected during executing a plan for excavation task
  • case 5: a manual plan sent from the Earth center
  • case 6: has a new task and need to terminate the current task
  • case 7: the current task finishes and has a new task needs
• The categorization of above cases (excluding case 7)
• Demo of adaptation cases:
  • Adaptation case 1 and case 2

Docker Image

• Build oceanworld and rs_autonomy images
  • Build oceanworld image, aisys/raspberry-oceanworld:v9
    cd docker/oceanworld && ./build.sh noetic v9 aisys
  • Build rs_autonomy image, aisys/raspberry-autonomy:v2
    cd docker/rs_autonomy && ./build.sh aisys/raspberrysi-oceanworld:v9 noetic v2 aisys ow9
  • Two available docker images are available for play at:
    • aisys/raspberrysi-oceanworld:v9
    • aisys/raspberrysi-autonomy:v2
• Run a docker container using the rs_autonomy image
  • Create the docker container, oceanworld, with the GPU support.
    xhost +local:root
docker run --name oceanworld --rm -it --gpus all -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY -e QT_X11_NO_MITSHM=1 aisys/raspberrysi-autonomy:v2 bash
  • Open two terminals. In each terminal, access to the docker container, oceanworld, by using the following command. One terminal is for running the PLEXIL Executive (ow_plexil package) while the other one is for running rs_autonomy.
    docker exec -it oceanworld bash