Analyzing Fuzzy Surface Modeling Using Load-Balanced Computation

In applications such as digital terrain modelling it is desirable to provide an explicit visualization of the model accuracy, for supporting users in spatial analysis activities. Coupling fuzzy arithmetic with traditional geometric models is an interesting possibility, but it requires the solution of computational problems related to the use of fuzzy-based models. In this context, we present a load balanced algorithm for the efficient use of parallel computing for typical queries of fuzzy models in Geographic Information Systems. We shortly introduce fuzzy arithmetic and its applications to terrain modelling. Then, the algorithm, which represents the computational kernel of querying fuzzy models, is described and its cost is evaluated. A parallelisation strategy for this algorithm is outlined. To obtain an efficient parallel code a great attention has to be paid to load balancing. We adopt an approach to load balancing, which introduces a low communication and computational overhead, it is flexible, portable, and it yields to good performances on different computing architectures. Experimental results obtained on a massively parallel machine, and on a network of heterogeneous workstations are discussed. Despite we refer to Geographic Information Systems as the present application field, the presented methodology could be applied to any scientific area in which surface representation and data uncertainity are important issues.