The 28th International Workshop on Qualitative Reasoning (QR-15) presented advances toward reasoning tractably with massive qualitative and quantitative models, automatically learning and reasoning about continuous processes, and representing knowledge about space, causation, and uncertainty. The technical track included two invited talks, 11 oral presentations, and 5 poster presentations.

Srivastava, Siddharth (University of Massachusetts, Amherst) | Zilberstein, Shlomo (University of Massachusetts, Amherst) | Immerman, Neil (University of Massachusetts, Amherst) | Geffner, Hector (ICREA and Universitat Pompeu Fabra)

We consider a new class of planning problems involving a set of non-negative real variables, and a set of non-deterministic actions that increase or decrease the values of these variables by some arbitrary amount. The formulas specifying the initial state, goal state, or action preconditions can only assert whether certain variables are equal to zero or not. Assuming that the state of the variables is fully observable, we obtain two results. First, the solution to the problem can be expressed as a policy mapping qualitative states into actions, where a qualitative state includes a Boolean variable for each original variable, indicating whether its value is zero or not. Second, testing whether any such policy, that may express nested loops of actions, is a solution to the problem, can be determined in time that is polynomial in the qualitative state space, which is much smaller than the original infinite state space. We also report experimental results using a simple generate-and-test planner to illustrate these findings.

If sufficient information about a system is not available, such that no quantitative model can be established to give a real-valued behavioral description about it over time, one has to turn to alternative way of modeling, making use of the available incomplete information to build qualitative models, on which an analysis and reasoning to it can be carried out. Qualitative simulation as one of the main techniques in qualitative reasoning has the great potential to play a very important role to solve engineering problems. However, due to the ambiguity of qualitative representation and calculus, a great number of the behaviors may be produced and seems to intractable, it often obscures the resulting real behavior of systems. The existing filtering techniques can not handle it in a complete satisfactory way. We believe that a further development will be still necessary before it can be extensively applied in engineering.

In this framework, the faulty part of a system can be identified by comparing the behavior derived by stochastic qualitative reasoning with the actual measured behavior. The latter is represented as the series of qualitative values that are obtained by classifying quantitative measurements into several qualitative categories based on a definition of the qualitative regions. The fault detection is often ineffective under the inappropriate definitions. This paper proposes a method that can automatically define the qualitative regions from the measured data. In this system, data are controlled using a certain value and follow a normal distribution. Measurement data must be transformed into stable qualitative values so that its behavior can be distinguished from fault conditions: therefore, the middle of the qualitative region which has the most stable qualitative value is determined as the average value of the data. The width of the most stable qualitative value is determined based on the standard deviation. This method is applied to an actual air conditioning system. According to the definition of qualitative regions that is determined from the field data, the faults can be identified.