TY - JOUR
T1 - How does a box work? A study in the qualitative dynamics of solid objects
AU - Davis, Ernest
N1 - Funding Information:
✩ My thanks to David Stewart, for information about the current state of the art in scientific computation in the theory of rigid solid objects; to Steve LaValle, for discussions about the relations of this work to robotics; to Leora Morgenstern and Vladimir Lifschitz, for discussions of non-monotonic inference; and to the reviewers for helpful suggestions. This research was supported in part by NSF grants IIS-097537 and IIS-0534809. E-mail address: [email protected]. 1 As opposed to, for instance, a theory that posits that the people’s interactions with boxes can be characterized in terms of stimulus/response behavior that subjects learn from positive and negative reinforcement when they put objects into boxes and take them out.
PY - 2011/1
Y1 - 2011/1
N2 - This paper is an in-depth study of qualitative physical reasoning about one particular scenario: using a box to carry a collection of objects from one place to another. Specifically we consider the plan, plan1 "Load objects uCargo into box oBox one by one; carry oBox from location l1 to location l2". We present qualitative constraints on the shape, starting position, and material properties of uCargo and oBox and on the characteristics of the motion that suffice to make it virtually certain that plan1 can be successfully executed. We develop a theory, consisting mostly of first-order statements together with two default rules, that supports an inference of the form "If conditions XYZ hold, and the agent attempts to carry out plan1 then presumably he will succeed". Our theory is elaboration tolerant in the sense that carrying out the analogous inference for carrying objects in boxes with lids, in boxes with small holes, or on trays can reuse much of the same knowledge. The theory integrates reasoning about continuous time, Euclidean space, commonsense dynamics of solid objects, and semantics of partially specified plans.
AB - This paper is an in-depth study of qualitative physical reasoning about one particular scenario: using a box to carry a collection of objects from one place to another. Specifically we consider the plan, plan1 "Load objects uCargo into box oBox one by one; carry oBox from location l1 to location l2". We present qualitative constraints on the shape, starting position, and material properties of uCargo and oBox and on the characteristics of the motion that suffice to make it virtually certain that plan1 can be successfully executed. We develop a theory, consisting mostly of first-order statements together with two default rules, that supports an inference of the form "If conditions XYZ hold, and the agent attempts to carry out plan1 then presumably he will succeed". Our theory is elaboration tolerant in the sense that carrying out the analogous inference for carrying objects in boxes with lids, in boxes with small holes, or on trays can reuse much of the same knowledge. The theory integrates reasoning about continuous time, Euclidean space, commonsense dynamics of solid objects, and semantics of partially specified plans.
KW - Commonsense reasoning
KW - Dynamics
KW - Kinematics
KW - Naive physics
KW - Planning
KW - Qualitative physics
KW - Solid objects
UR - http://www.scopus.com/inward/record.url?scp=78649733421&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=78649733421&partnerID=8YFLogxK
U2 - 10.1016/j.artint.2010.04.006
DO - 10.1016/j.artint.2010.04.006
M3 - Article
AN - SCOPUS:78649733421
SN - 0004-3702
VL - 175
SP - 299
EP - 345
JO - Artificial Intelligence
JF - Artificial Intelligence
IS - 1
ER -