Interactive CSP 43 Interactive Arc - Consistency 74 A Case Study on Shape Recognition 115 Conclusions and Future Works

We present an Interactive Constraint Satisfaction model for problems where knowledge is not completely deened at the beginning of the computation , but can be interactively acquired during the computational process. In this case, domains can be partially deened when the constraint satisfaction process starts. Thus, domains are dynamically deened in contrast with classical CSPs where domains are statically deened at the beginning of the constraint satisfaction process. The interactive CSP model copes with incomplete domain deenition by propagating constraints on already deened domain values and by adding new constraints on undeened parts of domains. These new constraints can be used to incrementally process new information without restarting a constraint propagation process from scratch each time new information is available. In addition, these constraints can be eventually used in order to guide the knowledge acquisition process. Examples are given in order to show the eeectiveness of the approach. In particular, a case study in the eld of visual object recognition is considered.