An automated testing methodology based on self-checking software

Abstract

New techniques are required to speed the software testing process and to enhance the ultimate reliability of mission-critical embedded information systems. Here we employ result-checking, a technique in which software assures the correctness of its own computations through efficient: run-time checks. Building on previous work at U.C. Berkeley and Raytheon Systems Company, the Air Force Research Laboratory is funding the Automated Generation of Avionics Software Tests (AGAST) program to demonstrate that result-checking can be the basis for a flexible new methodology of software testing. This methodology aims at the production of software which has an enhanced sensitivity to its own errors (complemented with appropriate instrumentation), and which therefore may be tested with extremely high efficiency and accuracy.