soargroup/rosie

Soar Agent (and SML code) that learns through situated interactive instruction in a robotic environment

rosie's Issues

Incomplete object descriptions on top-state

^state-prop attribute of objects on input-link is not copied over to objects on the top-state. Following is the trace from the state.

p o122 --depth 3
(O122 ^bbox B6 ^category location ^focus op_open-locationargument1 ^id 5
^linguistic-desc D8 ^pose P131 ^property P184 ^smem-rep R93)
(B6 ^max M12 ^min M11)
(M12 ^x 0.1 ^y 0.1 ^z 0.1)
(M11 ^x -0.1 ^y -0.1 ^z -0.1)
(D8 ^word pantry)
(P131 ^pitch 0. ^roll 0. ^x -0.2 ^y 0.2 ^yaw 0. ^z 0.)
(P184 ^name name ^type linguistic ^value V36 ^word pantry)
(V36 ^word pantry)
(R93 ^id 5 ^wme-type object)

p o19 --depth 3
(O19 ^bbox B6 ^id 5 ^name pantry ^pose P131 ^state-prop S31)
(B6 ^max M12 ^min M11)
(M12 ^x 0.1 ^y 0.1 ^z 0.1)
(M11 ^x -0.1 ^y -0.1 ^z -0.1)
(P131 ^pitch 0. ^roll 0. ^x -0.2 ^y 0.2 ^yaw 0. ^z 0.)
(S31 ^name door ^type state ^value open)

Elaborating object referents

I do not understand why the following was implemented -

(<int> ^stack.segments.segment <seg>
          ^referents.objects <objs>
         -^stack.segment <seg>)

Why are we not looking at the top-most segment to elaborate referent objects in focus?

line 95 in https://github.com/InstructoSoar/rosie/blob/master/agent/_agent/elaborations/elaborate-interaction.soar