A toolkit of script generator and data analyzer for generating virtual actions

Programming by demonstration (PbD) enables the direct delivery of the knowledge of a predecessor to virtual characters. Further, a PbD-based action generation framework automatically generates sequences of motions to be executed by a virtual character, on the basis of the collected data. However, this framework depends on the platform of the virtual environment. Moreover a variety of parameters required for generating actions depends on the collected data. Thus, it is required to realize a method to analyze the collected data and precisely set the parameters. This paper proposes an action generation toolkit that includes a script generator and a data analyzer. The proposed method was verified by a car simulation. The proposed method successfully increased the generality of the action generation framework, used for generating actions, and accurately controlled the action generation process.