Accountability and Explainability in Robotics: A Proof of Concept for ROS 2- And Nav2-Based Mobile Robots.

Using mobile robots in environments with humans has become increasingly common, raising concerns over security and safety. The lack of capabilities to justify their decisions and behaviours to non-expert users poses a significant challenge in gaining their trust, especially in safety-critical situations. Therefore, explaining why a robot carries out a specific and unexpected action is crucial to comprehend the cause of a failure when attempting to achieve a goal or complete a task. Besides, it is essential to make such explanations understandable to people, providing appropriate, effective, and natural responses to their queries. This work depicts a Proof of Concept of an accountability and explicability engine on ROS-based mobile robots. Our solution has two components: first, a black box-like component to provide accountability, and another for generating natural language explanations from data in the black box. Initial results show that it is possible to get accountable data from robot actions to obtain understandable explanations.