Abstract
Constraint satisfaction problems, commonly found in time/location referencing, database maintainence and air-traffic control applications, can be efficiently handled by a recently proposed binary relation inference network. There are some intrinsic weaknesses underlying this type of discrete-time inference network; namely, network instability and oscillation and the slow convergence rate commonly observed in large networks. Previous studies in network behaviour relied on ad hoc simulation; this paper describes some analytical methods to define the conditions which characterize asymptotic network behaviours. State-space techniques are used to describe dynamic behaviours, and provide the necessary mathematical tools to understand and explain network performance.