meta-control/mc_mros_reasoner

library for metacontrol-based self-adaptation using ontological reasoning, with wrappers for robotic systems based on ROS1 and ROS2

mc_mros_reasoner's Issues

Error executing the reasoner when the observers are loaded

I get the following error:

bad callback: <function callbackDiagnostics at 0x7f392df50e18>
Traceback (most recent call last):
  File "/home/nhg/rosin-paper/rospy3_melodic/src/ros_comm/rospy/src/rospy/topics.py", line 750, in _invoke_callback
    cb(msg)
  File "/home/nhg/rosin-paper/mros1_reasoner_ws/src/mros1_reasoner/mros1_reasoner/scripts/mros1_reasoner.py", line 220, in callbackDiagnostics
    updateQA(diagnostic_status)
  File "/home/nhg/rosin-paper/mros1_reasoner_ws/src/mros1_reasoner/mros1_reasoner/scripts/mros1_reasoner.py", line 238, in updateQA
    fg = onto.search_one(solvesO=onto.search_one(iri="*" + "o_navigateA")) #TODO read objective from diagnostic_status
  File "/home/nhg/rosin-paper/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/namespace.py", line 380, in search_one
    def search_one(self, **kargs): return self.search(**kargs).first()
  File "/home/nhg/rosin-paper/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/util.py", line 62, in first
    if len(self) != 0: return self[0]
  File "/home/nhg/rosin-paper/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/triplelite.py", line 1470, in __len__
    sql, params = self.sql_request()
  File "/home/nhg/rosin-paper/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/triplelite.py", line 1451, in sql_request
    transits, sql, params = self.sql_components()
  File "/home/nhg/rosin-paper/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/triplelite.py", line 1722, in sql_components
    sql = "SELECT DISTINCT q%s.s FROM %s WHERE %s" % (self.target, ", "   .join(tables), " AND ".join(conditions))
AttributeError: target

I am pretty sure I have to load somewhere a model (owl?) that set the target/objective, or maybe set the information manually but I don't know where

STEPS TO REPRODUCE THE ISSUE:

Installation:

mros1_resoner

I followed: https://github.com/rosin-project/mros1_reasoner/tree/mvp
plus:

pip3 install catkin_pkg
pip3 install pyyaml

metacontrol_sim

I followed: https://github.com/rosin-project/metacontrol_sim/tree/MVP_world

Execution:

roslaunch metacontrol_sim MVP_metacontrol_world.launch
roslaunch mros1_reasoner run.launch
rosservice call /load_observer EnergyQualityObserver
rosservice call /load_observer SafetyQualityObserver

The error come periodically and just after call the observers services.

No suitable FD found for adaptation

CONTEXT

Packages in their most recent branches for MODELS paper: mvp, master or observers

Experiment command:

./run_single_sim.sh -i 2 -g 2 -n standard -r true -p 1.3

OUTPUT

...
OBJECTIVE	|  STATUS	|  NFRs
o_navigateA 	|   INTERNAL_ERROR 	|   [('energy', 0.15), ('safety', 0.8)]
...
== Checking FDs for Objective with NFRs type:  energy and value:  0.15
	 WARNING FD has no expected value for this QA
	 WARNING FD has no expected value for this QA
#### WARNING: no FDs meetf NFRs
== FunctionDesigns also meeting NFRs:  []
*** OPERATOR NEEDED, NO SOLUTION FOUND ***
*** Objective  o_navigateA cannot be realised ***

Explore ontology queries to implement configuration selection (instead of phyton code)

ERROR: FG not found when received QA observation

TEST

run_single_sim.sh -i 1 -g 1 -n f1_v3_r1 -r true

OUTPUT

ERROR] [1590236368.127794]: = RECONFIGURATION SUCCEEDED =
[INFO] [1590236368.131871]: Exited timer_cb for metacontrol reasoning
[INFO] [1590236368.141647]: Entered timer_cb for metacontrol reasoning
[INFO] [1590236368.146012]:   >> Started MAPE-K ** Analysis (ontological reasoning) **
* Owlready2 * Running Pellet...
...
[INFO] [1590236368.368852]: received QA observation
ERROR: FG not found
[INFO] [1590236368.377154]: received QA observation
ERROR: FG not found
[INFO] [1590236369.053949]: received QA observation
ERROR: FG not found
[INFO] [1590236369.060824]: received QA observation
ERROR: FG not found
[INFO] [1590236369.070854]: received QA observation
ERROR: FG not found
[INFO] [1590236369.596342]: received QA observation
ERROR: FG not found
[INFO] [1590236370.137616]: received QA observation
ERROR: FG not found
[INFO] [1590236370.143824]: received QA observation
ERROR: FG not found
[INFO] [1590236370.747393]: received QA observation
ERROR: FG not found
[INFO] [1590236371.426202]: received QA observation
ERROR: FG not found
[INFO] [1590236371.437687]: received QA observation
ERROR: FG not found <--------------------------------------------------------------------------
[ERROR] [1590236371.854689]: Reasoning error: <--------------------------------------
[INFO] [1590236371.868696]:      >> Finished ontological reasoning)

Make QAs application specific

Separate QA definition from the metacontroller, make QA application specific:
the developer should define the QAs needed for the application: energy, safety or qa_x

For this, we need to create the generic QA class in tomasys so that we can have individuals in the application model for the application specific QAs

Missing location of dependencies (repositories) for installation instructions

After following the steps on README.md there are some dependencies repositories that need to be added to the reasoner_ws, they are on the package.xml but there's no indication on where to find them

These are the repositories needed and their locations

cheops_system_state_msgs
https://github.com/rosin-project/cheops/tree/master/cheops_system_state_msgs

cheops_graph_manipulation_msgs
https://github.com/rosin-project/cheops/tree/master/cheops_graph_manipulation_msgs
mc_mdl_tomasys
https://github.com/rosin-project/mc_mdl_tomasys

mc_mdl_abb
https://github.com/rosin-project/mc_mdl_abb

Status of objectives need to be reset after reconfiguration

Otherwise in the next metacontrol cycle reconfiguration will be done again (but maybe different since the configuration that failed to solve the objective (fg.typeFD) would be a different one)

Problem migrating to Melodic

I tried to emulate Gij's instructions for Kinetic with Python3 support, to setup a ROS Melodic installation with Python3 support for the metacontrol reasoner, see branch melodic

I think I managed to properly setup my Python3 virtual environment, build ROS Melodic bare-bones in a workspace, and then overlay my workspace for mros1_reasoner.
However, when trying to launch the reasoner I get:

(venv3.6_ros) chcorbato@tud0038655:~/abb_metacontrol_ws$ roslaunch mros1_reasoner run.launch 
Traceback (most recent call last):
  File "/home/chcorbato/rospy3_melodic/devel/bin/roslaunch", line 15, in <module>
    exec(compile(fh.read(), python_script, 'exec'), context)
  File "/home/chcorbato/rospy3_melodic/src/ros_comm/roslaunch/scripts/roslaunch", line 34, in <module>
    import roslaunch
ModuleNotFoundError: No module named 'roslaunch'

ModuleNotFoundError: No module named 'Cryptodome' when trying to use rostopic pub in launchfile

I have tried to create a mock diagnostics using rostopic pub ni the mros1_reasoner launch file, but I get the following error:

Traceback (most recent call last):
  File "/home/chcorbato/rospy3_melodic/src/ros_comm/rostopic/scripts/rostopic", line 35, in <module>
    rostopic.rostopicmain()
  File "<string>", line 2118, in rostopicmain
  File "/home/chcorbato/rospy3_melodic/devel/lib/python3/dist-packages/rosbag/__init__.py", line 34, in <module>
    exec(__fh.read())
  File "<string>", line 33, in <module>
  File "/home/chcorbato/rospy3_melodic/src/ros_comm/rosbag/src/rosbag/bag.py", line 53, in <module>
    from Cryptodome.Cipher import AES
ModuleNotFoundError: No module named 'Cryptodome'

This is the code that causes it when added in the launchfile:

 <node name="fake_m_diagnostics" pkg="rostopic" type="rostopic" args="pub /metacontrol_diagnostics DiagnosticArray.msg
    '{[status: {level: 1, name: '', message: '', values: []}]
    }' -r 2"/>

For the moment, I will use a launch file in the abb_cheops_ws for this implementation of mock_metacontrol_diagnostic with rostopic

Metacontrol is not triggering any adaptation

Test performed:

Test

Terminal 1:

OUTPUT

FD selected that does not meet the NFR threshold

Actually it's switching to f1_v3_r1 which is strange because that one doesn't meet the safety requirements

Originally posted by @marioney in #53 (comment)

mros1_reasoner does not store qa values for FG properly in the KB

The desired behavior is that only one value for each QA type is stored in the KB, and that value is continuously updated¹

Example of desired printouts:

FG Statuses:
fg_standard      None    [('obs_mvp.safety', 1.0), ('obs_mvp.energy', 0.685)

But we are getting this:

FG Statuses:
fg_standard      None    [('obs_mvp.safety_18', 1.0), ('obs_mvp.safety_36', 1.0), ('obs_mvp.safety_54', 1.0), ('obs_mvp.safety_70', 1.0), ('obs_mvp.safety_88', 1.0), ('obs_mvp.safety_106', 1.0), ('obs_mvp.energy_133', 6.20881714097e-10), ('obs_mvp.safety_124', 1.0)]

and apparently in some runs no energy value is associated to the FG (to be confirmed)

[1] Note: this is not how ontologies are supposed to work, so this requires further analysis from an ontological perspective

implement metacontrol reasoning in python

Update the README of this repo to indicate how to parse the OWL file

Ontology file should be given as arg to the run.launch file

e.g.

roslaunch mros1_reasoner run.launch onto:=mvp.owl

Metacontrol selects configuration based on QAs AND expected performance

In the MAPE-K PLAN step, the metacontroller should select, amongst the FDs that fullfi the energy and safety NFRs, the one with better performance.

That is, Utility function for the metacontrol (the one used for selection) could be:
U = QAperformance_expected

we could also consider to include QAenergy_expected and QAsafety_expected

Sometimes no QA values are received by mros1_reasoner

Context:

• all metacontrol monitoring input is sent via /diagnostics (probably not the best idea I ever had). This input is mainly messages about QA values for energy and safety
• in our system, /ekf_localization is also publishing in /diagnostics
• a callback in mros1_reasoner process these messages to update the KB (runtime ontology in memory)

Problem

Sometimes no QA values at all are received by mros1_reasoner
See TEST 1 below

This is tricky because it only happens sometimes, especially for energy (maybe because they are published less frequently)
See TEST 2

I have checked that sometimes energy and safety messages are published see TEST 3

Maybe this is related to the messages published in topic /diagnostics

Could it be that the diagnostic messages from /ekf_localization are "jamming" the topic?

@marioney @gavanderhoorn advice from ROS experts is needed here.

Tests

TEST 1: no QA value messages in /diagnostics at all? At least not received by mros1_reasoner

Terminal 1: simulation + navigation + rosgraph_monitor+ rosgraph_manipulator

roslaunch metacontrol_sim MVP_metacontrol_world.launch

OUTPUT analysis: All seems to work as expected

Terminal 2: rostopic echo /diagnostics

header: 
  seq: 20
  stamp: 
    secs: 19
    nsecs:  60000000
  frame_id: ''
status: 
  - 
    level: 0
    name: "ekf_localization: Filter diagnostic updater"
    message: "The robot_localization state estimation node appears to be functioning properly."
    hardware_id: "none"
    values: []
  - 
    level: 0
    name: "ekf_localization: odometry/filtered topic status"
    message: ''
    hardware_id: "none"
    values: 
      - 
        key: "Events in window"
        value: "502"
      - 
        key: "Events since startup"
        value: "952"
      - 
        key: "Duration of window (s)"
        value: "10.040000"
      - 
        key: "Actual frequency (Hz)"
        value: "50.000000"
      - 
        key: "Minimum acceptable frequency (Hz)"
        value: "43.200000"
      - 
        key: "Maximum acceptable frequency (Hz)"
        value: "57.200000"
... __AND ONLY MESSAGES LIKE THE ONE ABOVE__

OUTPUT analysis: All seems to work as expected

Terminal 3: mros1_reasoner

roslaunch mros1_reasoner run.launch onto:=mvp.owl
 ...
RELEVANT OUTPUT
----------------------------
FG Statuses:
fg_standard      None    []

OUTPUT analysis: The callback for message reception does not seem to be triggered, since there is no output like this:
[INFO] [1590055021.497386, 38.601000]: received QA observation

TEST in which only QA values on safety are received

Terminal 1: simulation + navigation + rosgraph_monitor+ rosgraph_manipulator

roslaunch metacontrol_sim MVP_metacontrol_world.launch

OUTPUT analysis: All seems to work as expected

Terminal 2: rostopic echo /diagnostics
Unfortunately not recorded, trying to reproduce

Terminal 3: mros1_reasoner

roslaunch mros1_reasoner run.launch onto:=mvp.owl
 ...
RELEVANT OUTPUT
----------------------------
[INFO] [1590055014.832565, 36.205000]: received QA observation
received QA about:  mvp.fg_standard     TYPE:  safety
              __5-6 LIKE THE ONE ABOVE on each reasoning period__
...
FG Statuses:
fg_standard      None    [('obs_mvp.safety_10', 1.0), ('obs_mvp.safety_14', 1.0), ('obs_mvp.safety_24', 1.0), ('obs_mvp.safety_33', 1.0), ('obs_mvp.safety_41', 1.0), ('obs_mvp.safety_52', 1.0), ('obs_mvp.safety_61', 1.0)]

OUTPUT analysis: The callback for message reception is triggered by the received messages with QA safety values

TEST 3

Terminal 1: simulation + navigation + rosgraph_monitor+ rosgraph_manipulator

roslaunch metacontrol_sim MVP_metacontrol_world.launch

OUTPUT analysis: All seems to work as expected

Terminal 2: rostopic echo /diagnostics

...
---
header: 
  seq: 4
  stamp: 
    secs: 3
    nsecs:         0
  frame_id: ''
status: 
  - 
    level: 0
    name: "ekf_localization: Filter diagnostic updater"
    message: "The robot_localization state estimation node appears to be functioning properly."
    hardware_id: "none"
    values: []
  - 
    level: 0
    name: "ekf_localization: odometry/filtered topic status"
    message: ''
    hardware_id: "none"
    values: 
      - 
        key: "Events in window"
        value: "147"
      - 
        key: "Events since startup"
        value: "147"
      - 
        key: "Duration of window (s)"
        value: "3.000000"
      - 
        key: "Actual frequency (Hz)"
        value: "49.000000"
      - 
        key: "Minimum acceptable frequency (Hz)"
        value: "43.200000"
      - 
        key: "Maximum acceptable frequency (Hz)"
        value: "57.200000"
---
header: 
  seq: 16
  stamp: 
    secs: 2
    nsecs: 802000000
  frame_id: ''
status: 
  - 
    level: 0
    name: ''
    message: "QA status"
    hardware_id: ''
    values: 
      - 
        key: "safety"
        value: "1.0"
---
header: 
  seq: 17
  stamp: 
    secs: 2
    nsecs:   4000000
  frame_id: ''
status: 
  - 
    level: 0
    name: ''
    message: "QA status"
    hardware_id: ''
    values: 
      - 
        key: "energy"
        value: "0.233684937159"

...

Could not start EnergyQualityObserver

I will come back to check in detail, but this is happening to me all the time when I run:
./run_single_sim.sh -i 2 -g 2 -n standard -r true -p 1.3

[INFO] [1590164922.395471, 0.000000]: Controller frequency: 1.0
[INFO] [1590164922.457184, 0.000000]: Waiting for service...
energy
EnergyQualityObserver
[INFO] [1590164922.602716, 0.000000]: Waiting for service...
Could not start EnergyQualityObserver

Decide how to execute reconfiguration

Can be based on:

• cheops_graph_manipulation

• the one from Rainbow

• Carlos' previous metacontroller

• Implementation in rosin-project/metacontrol_sim#7

Adaptation is not triggered with qa individuals

The rule in tomasys.owl is:
if fg.qa.value < obj.nfr.value
then obj in INTERNAL_ERROR

nfr.value = 1.5 https://github.com/rosin-project/mros1_reasoner/blob/382f4a0fae4233aaf79ff3a876cd8cc508d58bbe/mros1_reasoner/scripts/init_models.py#L49

updated QA energy values for fg (* for those above nfr value):
FG Statuses:
fg_nav_fast None [('obs_energy_1', '0.20000000298'), ('obs_energy_2', '**1.89199912548**'), ('obs_energy_3', '1.41609191895'), ('obs_energy_4', '1.42725098133'), ('obs_energy_5', '1.43280863762'), ('obs_energy_6', '1.43841362'), ('obs_energy_7', '1.45673406124'), ('obs_energy_8', '1.45814871788'), ('obs_energy_9', '1.45667350292'), ('obs_energy_10', '1.45740365982'), ('obs_energy_11', '1.45468640327'), ('obs_energy_12', '1.45513892174'), ('obs_energy_13', '1.45447206497'), ('obs_energy_14', '1.45391464233'), ('obs_energy_15', '1.45491302013'), ('obs_energy_16', '1.4551589489'), ('obs_energy_17', '1.45762276649'), ('obs_energy_18', '1.45493543148'), ('obs_energy_19', '1.45621967316'), ('obs_energy_20', '1.45462548733'), ('obs_energy_21', '1.45513737202'), ('obs_energy_22', '1.45356249809'), ('obs_energy_23', '1.4541246891'), ('obs_energy_24', '1.45473957062'), ('obs_energy_25', '1.45236492157'), ('obs_energy_26', '**1.68376469612**'), ('obs_energy_27', '**1.68711805344**')]

Potential causes of the undesired behavior:

• Problem with data types updating the KB
• Problem the SWRL rule QA1 in tomasys
  • Values need to be normalized for :

0  - worst case value
1  - desired value

Complete installation and execution instructions in README

We need to be able to deploy the mros1_reasoner in any machine.

As info, this is the excerpt of what @gavanderhoorn did in my machine to get it running

  940  mkdir $HOME/ros_kinetic_py3
  941  cd /home/chcorbato/ros_kinetic_py3/
  942  apt install python3-venv libpython3-dev python-catkin-tools
  943  sudo apt install python3-venv libpython3-dev python-catkin-tools
  944  sudo apt clean
  945  sudo apt update
  946  sudo apt-key del 421C365BD9FF1F717815A3895523BAEEB01FA116
  947  sudo -E apt-key adv --keyserver 'hkp://keyserver.ubuntu.com:80' --recv-key C1CF6E31E6BADE8868B172B4F42ED6FBAB17C654
  948  sudo apt clean && sudo apt update
  949  python3.5 -m venv venv3.5_ros
  950  source venv3.5_ros/bin/activate
  951  pip3 install wheel
  952  pip3 install -U rosdep rosinstall_generator wstool rosinstall
  953  rosinstall_generator --deps --tar --rosdistro=kinetic ros_base > 
  954  rosinstall_generator --deps --tar --rosdistro=kinetic ros_base > ros_base.rosinstall
  955  wstool init src -j8 ros_base.rosinstall
  956  catkin build -DCATKIN_ENABLE_TESTING=0 -DCMAKE_BUILD_TYPE=Release -DPYTHON_VERSION=3.5
  957  cd
  958  cat .bashrc 
  959  vim .bashrc 
  960  nano .bashrc 
  961  source .bash
  962  ls -al
  963  cd ..
  964  cd ros_kinetic_py3/
  965  cd mros_ws/src/
  966  git clone [email protected]:rosin-project/cheops.git
  967  cd mros_ws/src/mros1_reasoner/mros1_reasoner/scripts/
  968  python owl_reasoner.py 
  969  mmons-lang3-3.3.2.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/xml-apis-1.4.01.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/sesame-rio-datatypes-2.7.16.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/commons-csv-1.0.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/guice-multibindings-4.0.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/pellet-query-2.4.0-SNAPSHOT.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/pellet-jena-2.4.0-SNAPSHOT.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/pellet-cli-2.4.0-SNAPSHOT.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/commons-codec-1.6.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/pellet-explanation-2.4.0-SNAPSHOT.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/xz-1.5.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/sesame-rio-ntriples-2.7.16.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/stringtemplate-3.2.1.jar:/home/chcorbato/.local/share/virtualenvs/owlready2/lib/python3.5/site-packages/owlready2/pellet/jena-core-2.13.0.jar pellet.Pellet realize --infer-prop-values --infer-data-prop-values --ignore-imports /tmp/tmpuf3wjjzc
  970  Traceback (most recent call last):
  971  python owl_reasoner.py 
  972  Bindings Statuses:
  973  python owl_reasoner.py 
  974  git merge owlready2 
  975  git push
  976  python mros1_reasoner.py 
  977  python owl_reasoner.py 
  978  git fetch
  979  git checkout master
  980  git pull master
  981  git pull
  982  git checkout reasoner_integration 
  983  git rebase master
  984  git rebase --continue
  985  python mros1_reasoner.py 
  986  cd ros_kinetic_py3/
  987  ls -al
  988  catkin clean -y
  989  clear
  990  ls -al
  991  clear
  992  source venv3.5_ros/bin/activate
  993  env | less
  994  env | grep -i ros  less
  995  env | grep -i ros | less
  996  catkin build -DCATKIN_ENABLE_TESTING=0 -DCMAKE_BUILD_TYPE=Release -DPYTHON_VERSION=3.5
  997  pip3 install empy
  998  catkin build -DCATKIN_ENABLE_TESTING=0 -DCMAKE_BUILD_TYPE=Release -DPYTHON_VERSION=3.5
  999  source devel/setup.bash 
 1000  python
 1001  roscore
 1002  pip3 install defusedxml netifaces
 1003  roscore
 1004  pip3 install owlready2
 1005  cd ..
 1006  mkdir -p ros_py3_ws/src
 1007  rm -rf ros_py3_ws/
 1008  ls -al
 1009  cd mros_ws/
 1010  ls -al
 1011  catkin clean
 1012  catkin clean -y
 1013  ls -al
 1014  catkin config
 1015  catkin config --extend $HOME/ros_kinetic_py3
 1016  catkin config --extend $HOME/ros_kinetic_py3/devel
 1017  clear
 1018  catkin config
 1019  ls -al
 1020  catkin ls
 1021  ls -al
 1022  ls -al src/
 1023  touch src/rosin_diagnosis_experiments/CATKIN_IGNORE
 1024  catkin ls
 1025  catkin 
 1026  catkin b
 1027  catkin clean -y
 1028  catkin config --help
 1029  catkin config --no-extend
 1030  clear
 1031  ls -al
 1032  clear
 1033  ls -al
 1034  cd ..
 1035  mkdir -p abb_mros/src
 1036  cd abb_mros/
 1037  catkin init
 1038  catkin config --extend $HOME/ros_kinetic_py3/devel
 1039  catkin b
 1040  mc
 1041  catkin b
 1042  cd src/
 1043  ls -al
 1044  cd mros1_reasoner/
 1045  ls -al
 1046  git status
 1047  gitk --all
 1048  ls -al
 1049  cd mros1_reasoner/
 1050  ls -al
 1051  nano -w Cm
 1052  nano -w CMakeLists.txt 
 1053  nano -w package.xml 
 1054  cler
 1055  clear
 1056  ls -al
 1057  cd ..
 1058  ls -al
 1059  cd ..
 1060  cd cheops/
 1061  ls -al
 1062  git status
 1063  git branch -av
 1064  gitk --all
 1065  git branch -av
 1066  git checkout graph_manipulation 
 1067  gitk --all
 1068  git checkout master 
 1069  git ft
 1070  git fetch 
 1071  git status
 1072  git checkout graph_manipulation 
 1073  git reset --hard origin/graph_manipulation 
 1074  clear
 1075  gitk
 1076  cd ..
 1077  ls -al
 1078  clear
 1079  catkin clean -y
 1080  catkin b
 1081  source devel/setup.bash 
 1082  python
 1083  clear
 1084  cd 
 1085  source venv3.5_ros/bin/activate
 1086  ls
 1087  cd ros_kinetic_py3/
 1088  ls
 1089  source venv3.5_ros/bin/activate
 1090  cd ../abb_mros/
 1091  source devel/setup.bash 
 1092  roslaunch mros1_reasoner run.launch 
 1093  cd src/mros1_reasoner/
 1094  python owl_reasoner.py
 1095  cd mros1_reasoner/scripts/
 1096  python owl_reasoner.py
 1097  roslaunch mros1_reasoner run.launch 
 1098  python owl_reasoner.py
 1099  roslaunch mros1_reasoner run.launch 
 1100  ifconfig 
 1101  source /opt/ros/kinetic/setup.bash 
 1102  rosrun rqt_graph rqt_graph 
 1103  env | grep -i ros | sort
 1104  rosnode list
 1105  roswtf
 1106  clear
 1107  env | grep -i ros | sort
 1108  export ROS_MASTER_URI=http://192.168.131.1:11311

Explore use of python scripts in combination with the ontology OWL API

Mixing Python and OWLin owlready2:

• Adding Python methods to an OWL Class
• Generating Instance names
• Forward declarations
• Associating a Python module to an OWL ontology

Define the scenario for energy:

• what is the navigation mission
  • navigate from A to B
• what is the adaptation trigger?
  • battery consumption gets higher than a threshold
• what is the adaptation?
  • the metacontrol reconfigures move_base to the low_power config
  • FDs only involve change in move_base params max_vel and max_accel

Should the metacontroller performs only 1 reconfiguration?

Seems that reconfiguration is launched only once per experiment.

For instnce, I'm running

./run_single_sim.sh -i 1 -g 1 -n f3_v3_r3 -r true -p 2.0 -e 0.8 -s 0.4

There's a reconfiguration due to safety at the beginning, but even if the nfr_energy is not meet at the end of the experiment, there's no reconfiguration launched

@chcorbato is this expected?

Need proper functional hierarchy diagnosis

• If a needed objective is in error, not the entire hierarchy upstream needs to be fixed, but only the FG solving that objective
  • We need to differentiate FG errors: internal (due to binding errors or solved objective in error) or external (due to needed obj error)
  • FD.errorLog should only be updated in the case of FG internal error

move metacontrol repo (root) on github https://github.com/organizations/tud-cor

transfer ownership of the repository so it is maintained later in the CoR github organization

Rosinstall error

@marioney , can you check this:

I am getting the following error:

/home/darko/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/wstool/config_yaml.py:74: YAMLLoadWarning: calling yaml.load() without Loader=... is deprecated, as the default Loader is unsafe. Please read https://msg.pyyaml.org/load for full details.
  yamldata = yaml.load(stream)
ERROR in config: Invalid multiproject yaml format in [https://github.com/rosin-project/mros1_reasoner/blob/mvp/mros1_reasoner/mros1_reasoner.rosinstall]: mapping values are not allowed here
  in "<file>", line 115, column 73

Furthermore, the link to rosinstall is also wrong in this command:
wstool init src https://raw.githubusercontent.com/rosin-project/mros1_reasoner/mvp/mros1_reasoner.rosinstall

Remove reasoner output from terminal

The Pellet reasoner called with owlready2 produces and output that is annoying when debugging (see below)
I could not figure out how to remove it from the output. Owlready2 documents do not explain how to handle that
I tried this to temporarily remove it form the stdout, but it did not work.

Any ideas are welcome @gavanderhoorn @ipa-hsd @ipa-nhg @marioney

https://github.com/rosin-project/mros1_reasoner/blob/8de2e484650154245ed1eae0efebaa9ee711fe1b/mros1_reasoner/scripts/mros1_reasoner.py#L230

OUTPUT

* Owlready2 * Running Pellet...
    java -Xmx2000M -cp /home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/jena-arq-2.10.0.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/antlr-3.2.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/jena-iri-0.9.5.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/pellet-2.3.1.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/httpclient-4.2.3.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/antlr-runtime-3.2.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/owlapi-distribution-3.4.3-bin.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/jgrapht-jdk1.5.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/xercesImpl-2.10.0.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/slf4j-log4j12-1.6.4.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/xml-apis-1.4.01.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/jcl-over-slf4j-1.6.4.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/slf4j-api-1.6.4.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/jena-tdb-0.10.0.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/jena-core-2.10.0.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/commons-codec-1.6.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/aterm-java-1.6.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/httpcore-4.2.2.jar:/home/chcorbato/rospy3_melodic/venv3.6_ros/lib/python3.6/site-packages/owlready2/pellet/log4j-1.2.16.jar pellet.Pellet realize --loader Jena --input-format N-Triples --infer-prop-values --infer-data-prop-values --ignore-imports /tmp/tmp6__87twd
* Owlready2 * Pellet took 1.3000226020812988 seconds
* Owlready * (NB: only changes on entities loaded in Python are shown, other changes are done but not listed)

TypeError: '>' not supported between instances of 'IndividualValueList' and 'float'

https://github.com/rosin-project/mros1_reasoner/blob/ba6ec6c1d25bda7fd778adae293f1a4e3ede88ea/mros1_reasoner/scripts/mros1_reasoner.py#L74

This is error is usually related to to fd.fd_qa_tradeoff=[] because no value has been assigned by any SWRL rule, for example when the only QA we are considering is energy

Add rospy.logerror() messages in mros1_reasoner to chatch debug info in log files

runtime error missing 'java'

This seems related to missing dependencies related to owl reasoning #14
Maybe just install Protege is a workaround.

RECONFIGURATION UNKNOWN issue running single exp script WITH metacontrol

mros1_reasoner in 353c3be

Command executed:
./run_single_sim.sh -i 2 -g 2 -n standard -r true -p 1.3

INFO] [1590138841.922033]:   >> Finished MAPE-K ** EXECUTION **
== RECONFIGURATION UNKNOWN ==

Metacontrol loop takes around 1 sec?

We are calling the reasoner synchronously because of this strong limitation.

The time can be checked observing the rosout.info by the mros1_reasoner node:

In this run with NO ADAPTATION needed it takes over 1 sec

[INFO] [1590080065.797720]: Entered timer_cb for metacontrol reasoning
[INFO] [1590080065.799938]:   >> Started MAPE-K ** Analysis (ontological reasoning) **
...
[INFO] [1590080066.991114]:      >> Finished ontological reasoning)
...
[INFO] [1590080066.995500]:   >> Finished MAPE-K ** ANALYSIS **
[INFO] [1590080066.997103]:   >> Started MAPE-K ** PLAN adaptation **
-- NO ADAPTATION NEEDED --
[INFO] [1590080066.998722]: Exited timer_cb for metacontrol reasoning

With ADAPTATION needed but crashed (edit issue to add the trace of a successful adaptation):

[INFO] [1590080067.798098]: Entered timer_cb for metacontrol reasoning
[INFO] [1590080067.803917]:   >> Started MAPE-K ** Analysis (ontological reasoning) **
...
[INFO] [1590080068.819368]:      >> Finished ontological reasoning)
...
[INFO] [1590080068.822836]:   >> Finished MAPE-K ** ANALYSIS **
[INFO] [1590080068.824089]:   >> Started MAPE-K ** PLAN adaptation **
...
[INFO] [1590080068.825849]:   >> Finished MAPE-K ** Plan adaptation **
[INFO] [1590080068.827475]:   >> Started MAPE-K ** EXECUTION **
[WARN] [1590080068.829207]: New Configuration requested: fd_navigate_safe
^C[reasoner-2] killing on exit

Error in log? - Maybe caused by rospy.logerr

It seems to be an error in printing the log

File "/usr/lib/python3.6/threading.py", line 916, in _bootstrap_inner
    self.run()
  File "/home/mario/ros/rospy3_melodic/src/ros_comm/rospy/src/rospy/timer.py", line 234, in run
    self._callback(TimerEvent(last_expected, last_real, current_expected, current_real, last_duration))
  File "/home/mario/ros/reasoner_metacontrol_ws/src/mros1_reasoner/mros1_reasoner/scripts/mros1_reasoner.py", line 264, in timer_cb
    fd = selectFD(objectives_internal_error[0])
  File "/home/mario/ros/reasoner_metacontrol_ws/src/mros1_reasoner/mros1_reasoner/scripts/mros1_reasoner.py", line 126, in selectFD
    fd = obtainBestFunctionDesign(o)
  File "/home/mario/ros/reasoner_metacontrol_ws/src/mros1_reasoner/mros1_reasoner/scripts/mros1_reasoner.py", line 60, in obtainBestFunctionDesign
    rospy.loginfo("== FunctionDesigns available for obj: ", [fd.name for fd in fds])
  File "/home/mario/ros/rospy3_melodic/src/ros_comm/rospy/src/rospy/core.py", line 190, in loginfo
    _base_logger(msg, args, kwargs, level='info')
  File "/home/mario/ros/rospy3_melodic/src/ros_comm/rospy/src/rospy/core.py", line 183, in _base_logger
    logfunc(msg, *args, **kwargs)
  File "/usr/lib/python3.6/logging/__init__.py", line 1308, in info
    self._log(INFO, msg, args, **kwargs)
  File "/usr/lib/python3.6/logging/__init__.py", line 1444, in _log
    self.handle(record)
  File "/usr/lib/python3.6/logging/__init__.py", line 1454, in handle
    self.callHandlers(record)
  File "/usr/lib/python3.6/logging/__init__.py", line 1516, in callHandlers
    hdlr.handle(record)
  File "/usr/lib/python3.6/logging/__init__.py", line 865, in handle
    self.emit(record)
  File "/home/mario/ros/rospy3_melodic/src/ros_comm/rosgraph/src/rosgraph/roslogging.py", line 245, in emit
    record_message = _defaultFormatter.format(record)
  File "/usr/lib/python3.6/logging/__init__.py", line 577, in format
    record.message = record.getMessage()
  File "/usr/lib/python3.6/logging/__init__.py", line 338, in getMessage
    msg = msg % self.args
TypeError: not all arguments converted during string formatting

After this the reasoner basically stops working

Read NFRs from rosparams in mros1_reasoner

param names:

\nfr_energy
\nfr_safety

Better owlready2 code

• use search_one() instead of search()[0]
• create individuals with keyword parameters for properties

SWRL rules with ^^xsd:string in an antecedent atom does not seem to work

(required for the implementation to solve #4 )
Example rule S1a:
fg_status(?fg, "INTERNAL_ERROR"^^xsd:string), FunctionGrounding(?fg), solvesO(?fg, ?o), Objective(?o) -> o_status(?o, "INTERNAL_ERROR"^^xsd:string)

However, the rules with a logical value for the atom seem to be working.
e.g. rule S3:
Binding(?b) ^ hasBindings(?fg, ?b) ^ b_status(?b, false) -> fg_status(?fg, "INTERNAL_ERROR")

TEST:
=====
- branch: navigation
- mc_mdl_tomasys - branch: master
- mc_mdl_abb - branch: navigation
       ontology file: abb_dualarm_mm_complete.owl

Commands:
---------
roslaunch mros1_reasoner run.launch
roslaunch cheops_mock_diagnostics mock_metacontrol_monitoring_input.launch
rosservice call /mock_camera/is_in_error "data: true"

EXPECTED RESULT:
=================
FG Statuses:
fg_build_2arms None
fg_locate_tag None
fg_detect_tag INTERNAL_ERROR

Objectives Statuses:
o_build_pyramid EXTERNAL_ERROR
o_tag_localization EXTERNAL_ERROR
o_detect_tag INTERNAL_ERROR

RESULT:
=======
FG Statuses:
fg_build_2arms None
fg_locate_tag None
fg_detect_tag INTERNAL_ERROR

Objectives Statuses:
o_build_pyramid None
o_tag_localization None
o_detect_tag None

Don't overlay Py2 workspace on Py3 (as suggested in the readme)

The readme currently states:

Since mros1_reasoner depends on cheops_system_state_msgs, you will need workspace overlaying to run mros1_reasoner. You can do the following:

source abb_cheops_ws/cheops/devel/setup.bash
cd abb_metacontrol_ws
catkin b

this is not a good idea, as this would overlay a Python 2 workspace on top of a Python 3 workspace.

Suggestion: copy the cheops_system_state_msgs package (and just that package) into the Python 3 workspace (ie: abb_metacontrol_ws), so it now exists in both the Python 2 as well as the Python 3 workspaces.

That is completely fine and supported.

In essence the Python 2 and Python 3 workspaces are completely separate ROS installations. They may even be run on different machines (but they just happen to run on the same machine).

Re-send nav action goal after re-launching move_base

What about the navigation goal?
It is hold by the action server, which originally is the move_base, but we kill it
SOLUTION: intermediary action server in the metacontrol that is the action server that the user/app_logic sees, and to which is sends the nav_goal
The the metacontroller manages that this goal is a metacontrol objective and is responsible to send it again to new configurations

Missing to reconfigure despite Observed qa_energy above threshold for some messages

This is related to the current architecture of the mros_reasoner (see discussions 1 and 2), that is:

• async reception of input data
• sync execution of the reasoning with that data (only most recent value)

I propose to keep the highest value received on each timer_cb cycle for reasoning

@marioney @gavanderhoorn what do you think? We need a quick fix now for these experiments to provide useful results for the paper, but also a proper solution maybe later, rethinking the execution model of the metacontroller.

Unknown configuration request desired_configuration_name: "fg_fd_navigate_safe"

(in 0885143)
This requires a proper fix in mros1_reasoner.py and maybe rosgraph_manipulator.py

[INFO] [1588331252.970999]: Rosgraph Manipulator Action Server received goal desired_configuration_name: "fg_fd_navigate_safe"
[INFO] [1588331252.974950]: Unknown configuration request desired_configuration_name: "fg_fd_navigate_safe"

Reasoning error

Something did not go well, probably inconsistency in the ontology.
When this error is triggered I'm saving the ontology in ~/.ros/error.owl, so I will inspect it with Protege to check what happened.

TEST

run_single_sim.sh -i 1 -g 1 -n f1_v3_r1 -r true

OUTPUT

...
[ERROR] [1590235376.721149]: Reasoning error:
...

Estimate reasonable values for the QAs of each FD/configuration

• @darkobozhinoski please add link to relevant QAs definitions (safety, energy)
• Run basic experiments
• obtain experimental data on the evaluation metrics (see goc) and metacontrol_sim#26

QA diagnostics of the same QA from Observer are accumulated as a list of QA values for reasoner

See: https://github.com/rosin-project/mros1_reasoner/blob/d94814671bda1d13d8ce52b6a2572f4054c863f6/mros1_reasoner/scripts/mros1_reasoner.py#L231

• Since there is no time info attached for the reasoner, probably we rather update the QA value for the FG, rather than append a new value to a list.

FG Statuses:
fg_nav_fast      None    [('obs_energy_1', '0.20000000298'), ('obs_energy_2', '1.89199912548'), ('obs_energy_3', '1.41609191895'), ('obs_energy_4', '1.42725098133'), ('obs_energy_5', '1.43280863762'), ('obs_energy_6', '1.43841362'), ('obs_energy_7', '1.45673406124'), ('obs_energy_8', '1.45814871788'), ('obs_energy_9', '1.45667350292'), ('obs_energy_10', '1.45740365982'), ('obs_energy_11', '1.45468640327'), ('obs_energy_12', '1.45513892174'), ('obs_energy_13', '1.45447206497'), ('obs_energy_14', '1.45391464233'), ('obs_energy_15', '1.45491302013'), ('obs_energy_16', '1.4551589489'), ('obs_energy_17', '1.45762276649'), ('obs_energy_18', '1.45493543148'), ('obs_energy_19', '1.45621967316'), ('obs_energy_20', '1.45462548733'), ('obs_energy_21', '1.45513737202'), ('obs_energy_22', '1.45356249809'), ('obs_energy_23', '1.4541246891'), ('obs_energy_24', '1.45473957062'), ('obs_energy_25', '1.45236492157'), ('obs_energy_26', '1.68376469612'), ('obs_energy_27', '1.68711805344')]

Also:

• All values printed are the same > not the case
  https://github.com/rosin-project/mros1_reasoner/blob/d94814671bda1d13d8ce52b6a2572f4054c863f6/mros1_reasoner/scripts/mros1_reasoner.py#L249
  not sure whether the error is printing or storing the new value from diagnostics

Build mros_reasoner as a rospy package without ros installation

From @gavanderhoorn (thanks!) in #6

I would perhaps just try to use rospypi/simple.

Create a venv, just as you've done earlier, run the two install commands shown in the readme -- make sure to install additional packages if/when needed.

Now install whatever else you need for the reasoner.

roslaunch is probably not needed. The .launch file on starts a single node, which you could start as a regular program inside a terminal with the venv active.

This is an interesting idea to simplify the release and deployment of mros_reasoner

Remove dependency on https://github.com/rosin-project/mc_mdl_abb.git

• Create example of metacontrol use in the same repository like in System Modes (preferably a separate package). These requires:
• minimal rosgraph (either dumb node(s)) or launchfile using other repos (not preferred, since will add more dependencies)
• owl file with the model of the minimal example

Automatic generation of owl file from RosModel

Using the RosModel model files of the system as input, a python script could automatically generate the .owl file for the metacontrol.

Such script requires similar resources to the mros1_reasoner.py (the metacontrol), that is why I placed a skeleton for such a script in this repository:
https://github.com/rosin-project/mros1_reasoner/blob/rosmodel2owl/mros1_reasoner/scripts/rosmodel2owl.py

The skeleton contains comments on how to create the owl elements from the RosModel

@ipa-hsd can you help me create this script today? I saw your code to parse RosModel and looks great, but it is beyond my current skills.
@ipa-nhg can help clarify.

Officially I have meetings till 16.00, but hopefully I'll be available again earlier, for 14.00

Avoid using a separate ws with manual ROS compilation for Python3

Create a singularity image with ROS Melodic version with Python3 to avoid creating a virtual environment.

desired_configuration needs to be sync between metacontrol nodes

The metacontrol MAPE-K elements need to sync on what is the desired configuration at each time:

• ros_graph_parser (and rosgraph_monitor? ) use it to know what .rossystem to compare with the one extracted from the rosgraph => not needed now, see below
• mros1_reasoner needs to initialize what is the initial configuration, so it can assign the current QA values and configuration diagnostics from monitoring to it desired_configuration
  • done on the mros1_reasoner.py side https://github.com/rosin-project/mros1_reasoner/blob/31def16da53deea7377f6c950c53a0462f82897b/mros1_reasoner/scripts/mros1_reasoner.py#L316
  • rosin-project/metacontrol_experiments#8
• rosgraph_manipulator.py receives it to select which configuration to deploy
  =>This one is done via the MvpReconfiguration action goal received from mros1_reasoner

I propose to use a ROS global parameter to sync this value and also to initialize it.
What do you think @ipa-nhg @ipa-hsd @marioney

Problem finding the expected QA for FDs

Testing with ddbb21d

== Checking FDs for Objective with NFRs type:  energy and value:  0.15
         WARNING FD has no expected value for this QA

Metacontroller only considers the most recent QA values for reconfiguration

The QAs values are received asynchronously:
https://github.com/rosin-project/mros1_reasoner/blob/0a8022676b597f59fb8b6880f3d05955628f8643/mros1_reasoner/scripts/mros1_reasoner.py#L128

but the reasoner to decide if an adaptation is needed is executed sync
https://github.com/rosin-project/mros1_reasoner/blob/0a8022676b597f59fb8b6880f3d05955628f8643/mros1_reasoner/scripts/mros1_reasoner.py#L193

As a result, I think that we are not considering some QA values for reconfiguration.

• Do we need to reduce increase the metacontrol freq?
  • see https://github.com/rosin-project/mros1_reasoner/blob/b6d0a836f96ea8e85353b11ef5d4d6dd0e589dc8/mros1_reasoner/scripts/mros1_reasoner.py#L325
• Shall we do it? How much?
  • Reasoner needs over 1 sec for each loop see #38

We need to confirm if this is a problem and then decide what to do.