Virtual-Experiment-Driven Process Model (VEDPM)

Abstract

Computer simulations are the last resort for many complex problems such as swarm applications. However, to the best of the author's knowledge, there is no convincing work in proving ''What You Simulate ls What You See'' (WYSIWYS). Many models are built on long, subjective code that is prone to abnormalities, which are about corrupted virtual scientific laws rather than software bugs. Thus, the task of validating scientific simulations is very difficult, if not impossible. This dissertation provides a new process methodology for solving the problems above: Virtual-Experiment-Driven Process Model (VEDPM). VEDPM employs simple yet sound virtual experiments for verifying simple, short virtual laws. The proven laws, in turn, are utilized for developing valid models that can achieve real goals. The resulted simulations (or data) from proven models arc WYS1WYS. Two complex swarm applications have been developed rigorously and successfully via VEDPM--proving that VEDPM is workable. In addition, the author also provides innovative constructs for developing autonomous unmanned vehicles--swarm software architecture and a modified subsumption control scheme, and their design philosophies. The constructs are used repeatedly to enable unmanned vehicles to switch behaviors autonomously via a simple control signal.