Improving Adaptive Human-Robot Cooperation through Work Agreements

Abstract

Human-robot teams for disaster response need to dynamically adapt their task allocation and coordination to the momentary context, based on adequate trust stances and taking account of the relevant values and norms (such as safety, health, and privacy). This paper presents a Work Agreement framework that supports this capability. The research question is: Which minimal set of concepts, relations, and associated formalization can be used to model Work Agreements with adequate expressiveness and flexibility? An ontology has been developed that defines core concepts with their relations (creditor, debtor, antecedent, consequent, lifespan, acceptance), encompassing the knowledge to specify, activate, monitor, and reason about work agreements. The framework was implemented and tested as part of the TRADR project. The TRADR system brought forward the desired adaptive team behavior of the concerning robot. The tests led to further refinements of the work agreements framework.