Indirect human-robot task communication using affordances

Abstract

One problem in current human-robot task communication is the laborious need to define action and target object parameters for each task request. This paper's solution to the problem is to enable indirect task communication by mimicking the human cognitive ability to understand affordances, i.e. action possibilities in the environment with respect to different actors. This enables humans to communicate tasks using only the task-related action or target object names, and thus avoid the need to remember explicit task request utterances. The proposed task communication is integrated as a subsystem into an existing service robot, and its functionality is evaluated through a set of user experiments in an astronaut-robot task communication context. Affordance-based indirect task communication is shown to successfully reduce the workload experienced by the human and to decrease task communication times, while also being the preferred way to communicate tasks.