An Annotated Corpus of Examples of Commonsense Inference in Solid Object Dynamics

Ernest Davis
Department of Computer Science
Courant Institute of Mathematical Sciences
New York University

This document gives a structured, annotated corpus of examples of commonsensically obvious inferences in the domain of solid object dynamics. Examples are organized into similarity classes and are annotated by features.

Features
Examples
Bibliography

Features

• Subtheory: geometric/statics/kinematics/quasistatics/dynamic.
  A geometric problem deals configurations of objects at a single objects, subject to the constraint that objects do not overlap.
  In a static problem everything remains in the same place.
  In a kinematic problem the only physical constraints are that objects are rigid, move continuously, and do not overlap.
  In a quasi-static, problem, it is assumed that frictive forces are large enough that nothing moves unless pushed (directly or indirectly) by an imposed force.
  A dynamic problem involves the full physical theory.
• Deterministic vs. probabilistic vs. comparative probabilistic vs. ``tendency''. The last is the following very common framework for the conclusion of an inference: Under circumstances P, definitely A; under circumstance Q, definitely B; the closer the circumstances to Q than to P, the larger the probability of B as opposed to A. This is thus the combination of a deterministic with a comparative probabilistic inference.
• Enumerated objects (default) vs. collections of objects
• External actions or not (default).
• Direction of inference:
  • Geometric: A purely geometric constraint.
  • Prediction: Predict future behavior from (a) current state; (b) current state plus external actions; (c) past behavior.
  • Postdiction: Infer past events from current state.
  • Interpolation: From states at an earlier and a later time, infer the events that occured between.
  • Shape from behavior.
  • System identification: Characterize a system (e.g. "This is a door") given its geometric and physical specification.
  • Infer the existence of external objects or external actions from behavior.
  • Infer the existence of scenarios of a given property.
  • Metalevel: Infer a meta-level rule.
  • Abstraction: Infer that a system can be abstracted as a simpler system.
• Comments; e.g. bibliographic citation.

Content

• Foundations
  • Ontology Ontological categories, foundational and temporal predicates, notational conventions.
  • Geometry The language of geometry.
  • Probability

• Abstract Geometric and Physical Theories
  • Fitting
  • Holes
  • Kinematics
  • Dynamics
  • Abstraction, Approximation, and MetaRules

• Physical Systems
  • Boxes.
  • Rings
  • Doors
  • Doors and deadbolts.
  • Block dropped on the ground.
  • Heap of objects
  • Wobbly desk

Bibliography

E. Davis. 1988. A Logical Framework for Commonsense Predictions of Solid Object Behavior. AI in Engineering, vol. 3 no. 3, pp. 125-140.

E. Davis. 1990. Representations of Commonsense Knowledge. Menlo Park, Calif.: Morgan Kaufmann.

E. Davis. 1995a. Approximation and Abstraction in Solid Object Kinematics. NYU Computer Science Tech. Report 706, September 1995

B. Faltings. 1987. Qualitative Kinematics in Mechanisms. In Proceedings of the Tenth International Joint Conference on Artificial Intelligence, 436-442. Menlo Park, Calif: International Joint Conferences on Artificial Intelligence.

K. Forbus. 1980. Spatial and Qualitative Aspects of Reasoning about Motion. In Proceedings of the First National Conference on Artificial Intelligence. Menlo Park, Calif: American Association for Artificial Intelligence.

A. Gelsey. 1995. Automated Reasoning about Machines. Artificial Intelligence, vol 74, pp. 1-53.

L. Joskowicz and E. Sacks. 1991. Computational Kinematics. Artificial Intelligence, 51: 381-416.

L. Joskowicz, E. Sacks, and V. Srinivasan. 1997. ``Kinematic Tolerance Analysis,'' Computer-Aided Design, Vol. 29, No. 2.

P. Nielsen. 1988. A Qualitative Approach to Mechanical Constraint. Proc. AAAI-88, pp. 270-274.

E. Sandewall. 1989. Combining Logic and Differential Equations for Describing Real-World Systems. In Proceedings of the First International Conference on Knowledge Representation and Reasoning, 412-420. Menlo Park, Calif: Morgan Kaufmann.

T. Stahovich, R. Davis, and H. Shrobe. 2000. Qualitative rigid-body mechanics. Artificial Intelligence Vol. 119, pp. 19-06.