From Assumptions to Actions: Turning LLM Reasoning into Uncertainty-Aware Planning for Embodied Agents

Embodied agents operating in complex, dynamic environments often struggle with uncertainty in their perceptions and actions. This paper proposes a novel framework that bridges the gap between large language model (LLM) reasoning and real-world execution by enabling uncertainty-aware planning. Our approach allows embodied agents to translate high-level LLM generated goals into robust, executable action sequences by explicitly modeling and accounting for perceptual and environmental uncertainties, leading to safer and more reliable robot deployments.