A SOFTWARE RELIABILITY MODEL FOR ARTIFICIAL INTELLIGENCE PROGRAMS

In this paper we develop a software reliability model for Artificial Intelligence (AI) programs. We show that conventional software reliability models must be modified to incorporate certain special characteristics of AI programs, such as (1) failures due to intrinsic faults, e.g., limitations due to heuristics and other basic AI techniques, (2) fuzzy correctness criterion, i.e., difficulty in accurately classifying the output of some AI programs as correct or incorrect, (3) planning-time versus execution-time tradeoffs, and (4) reliability growth due to an evolving knowledge base. We illustrate the approach by modifying the Musa-Okumoto software reliability growth model to incorporate failures due to intrinsic faults and to accept fuzzy failure data. The utility of the model is exemplified with a robot path-planning problem.