Explicit World Models for Reliable Human-Robot Collaboration

TL;DR

The paper reframes reliable embodied AI as the construction and continual updating of an explicit world model that encodes the common ground between humans and robots, rather than relying on opaque end-to-end systems or formal verification alone. It surveys human-inspired grounding, explicit world modeling, and neuro-symbolic architectures to argue that reliable collaboration in social, multimodal settings emerges from shared interpretations of states and intentions. The authors highlight TCQA, joint attention, and NeSyC-style approaches as concrete building blocks toward lightweight, real-time world models that can adapt to ambiguous human goals. The work aims to catalyze multidisciplinary efforts to develop practical, explainable HRC capable of aligning robotic behavior with human expectations across diverse contexts.

Abstract

This paper addresses the topic of robustness under sensing noise, ambiguous instructions, and human-robot interaction. We take a radically different tack to the issue of reliable embodied AI: instead of focusing on formal verification methods aimed at achieving model predictability and robustness, we emphasise the dynamic, ambiguous and subjective nature of human-robot interactions that requires embodied AI systems to perceive, interpret, and respond to human intentions in a manner that is consistent, comprehensible and aligned with human expectations. We argue that when embodied agents operate in human environments that are inherently social, multimodal, and fluid, reliability is contextually determined and only has meaning in relation to the goals and expectations of humans involved in the interaction. This calls for a fundamentally different approach to achieving reliable embodied AI that is centred on building and updating an accessible "explicit world model" representing the common ground between human and AI, that is used to align robot behaviours with human expectations.