Towards Incremental Development of Human-Agent-Robot Applications using Mixed-Reality Testbeds

Abstract

Testing and evaluation is an essential part of developing applications involving teams of humans, agents and robots. In such applications, individual algorithms cannot be reliably tested in isolation because their performance depends on complex interactions with the environment and with other humans, agents and robots involved. At the same time, testing with real robots and human individuals is costly and potentially dangerous. We propose an incremental development framework employing mixed-reality testbeds. It aims at reduction of development costs and risks by fully or partially substituting parts of the application and the surrounding reality with computational models. The framework parameterizes testbeds in terms of their fidelity and size, and comes with guidelines for managing the two parameters through a sequence of iterations so as to maximize the effectiveness of system development. The framework is illustrated on an example application in the domain of multi-UAV tracking of mobile targets.