search

rosin-project / metacontrol_sim

2 stars 4 forks source link

Adaptation rule based on instantaneous power as QA energy triggers reconfiguration when there is a sporadic peak in battery load #14

Closed chcorbato closed 4 years ago

chcorbato commented 4 years ago

@darkobozhinoski maybe you can analyze whether this is acceptable or not and propose solutions?

Important: mros1_reasoner is not yet working 100% ( https://github.com/rosin-project/mros1_reasoner/commit/203760fded386bc4d584228cd6a40471730a9d2d ), but I have already detected that reconfiguration can be triggered by sporadic values of power consumption/battery load, which does not seem a desired behavior to me.

This is the observed behavior when I start MVP_metacontrol_world.launch and the rest of elements for metacontrol, with the reasoner initialized with a navigation objective with an associated obj.nfr_qa_energy = 2.0:

  1. An initial goal is sent tomove_base > #15
  2. rosgraph_monitor monitor reports periodically on the QualityObserver
  3. rosgraph_monitor monitor goes from 0.20000000298 to values ~1.45, with some isolated peak of 2.02
  4. mros1_reasoner can miss that peak, if not, it triggers reconfiguration.
    • [x] check if the peak is indeed not missed

UPDATE:

In MVP d92e96e348b5839754d08d885c85cfc90ec4ffa6 (but launching the metacontrol interfaces in separate terminals see #20) this is the output of in the previous scenario :-)

starting QualityObserver...
1.67026388645
1.6880903244
1.45484399796
1.69497358799
2.42952299118
2.28727555275
1.45558822155
1.45585799217
2.5255382061
2.18126916885
1.64175379276      <-- mros1_reasoner reacts to this value and commands reconfiguration
0.20000000298
0.20000000298
0.450133442879
0.880685329437
0.683739185333
0.647721648216
0.615718364716
0.524811983109
0.51432299614
0.734792113304
0.735808312893
0.734537005424
0.735540270805
0.734255671501
1.05744349957
0.734531581402
0.734742462635
1.01701796055
0.734637200832
0.734486818314
0.734632194042
0.73434394598
0.734391331673
0.734278976917
0.733835756779
chcorbato commented 4 years ago

This is an example of experimental data we need from https://github.com/rosin-project/metacontrol_sim/issues/26

Confirm after the minimum experiments if this is an issue

chcorbato commented 4 years ago

This has been solved in the implementation of the energy observed