Integrated Commonsense Reasoning and Deep Learning for Transparent Decision Making in Robotics

Abstract A robot’s ability to provide explanatory descriptions of its decisions and beliefs promotes effective collaboration with humans. Providing the desired transparency in decision making is challenging integrated robot systems that include knowledge-based reasoning methods data-driven learning methods. As a step towards addressing this challenge, our architecture combines complementary strengths non-monotonic logical incomplete commonsense domain knowledge, deep learning, inductive learning. During enables on-demand explanations decisions, evolution associated beliefs, outcomes hypothetical actions, form relational relevant objects, attributes, actions. The architecture’s capabilities are illustrated evaluated context scene understanding tasks planning performed using simulated images from physical manipulating tabletop objects. Experimental results indicate reliably acquire merge new information about constraints, preconditions, effects accurate presence noisy sensing actuation.