Reasoning

Reasoning in #bayrob is performed using the bayrob.core.base.BayRoB class, which triggers the query of the respective action models (of type jpt.trees.JPT) with a passed bayrob.core.base.Query object. This object requires information about made observations (evidence) and the variables (queryvars) one is interested in.

import os
from jpt.base.intervals import ContinuousSet
from bayrob.core.base import BayRoB, Query
from bayrob.utils import locs

preset = {
     "evidence": {
         'detected(milk)': False,
         'x_in': ContinuousSet(58, 68),
         'y_in': ContinuousSet(70, 80),
         'nearest_furniture': 'fridge'
     },
     "queryvars": ['daytime', 'open(fridge_door)']
 }

bayrob = BayRoB()
bayrob.adddatapath([os.path.join(locs.examples, 'demo', "perception")])
allvars_ = {v.name: v for v in bayrob.models['perception'].variables}

qo = Query()
qo.model = bayrob.models['perception']
qo.evidence = {allvars_[k]: v for k, v in preset['evidence'].items()}
qo.queryvars = [bayrob.models['perception'].varnames[k] for k in preset['queryvars']]

bayrob.query = qo
bayrob.query_jpts()
cond, post = bayrob.result.result
print(cond)
print(post)

The result of a BayRoB query is a tuple (cond, post) where cond is the conditional tree (jpt.trees.JPT) and post is a mapping of the variables (jpt.variables.Variable) specified in queryvars to their respective posterior distributions (jpt.distributions.univariate.numeric.Numeric, jpt.distributions.univariate.multinomial.Bool, jpt.distributions.univariate.multinomial.Multinomial, …):