Interference analysis in multiple rule firing systems

Many multiple rule firing models are being formulated to improve the performance of forward chaining production systems. Multiple rule firing can possibly compromise working memory consistency, due to interference between actions of one rule and conditions of another. Techniques based on data dependency graphs of production systems have been proposed for compile time and run time interference detection. These techniques are highly memory intensive requiring large storage space and impose a high computational overhead. We formulate interference as a non-empty join of condition and action elements in a production system. The RETE net is used to compute the joins and detect interference, thus obviating the need for data dependency graphs. This technique is less memory intensive requiring negligible storage space when compared to data dependency graph techniques.