Synergistic Multi-Robot Knowledge Aggregation for Energy-Sustainable Embodied Autonomy

Abstract

This paper explores the critical energy demands arising from the integration of artificial intelligence and robotics, forming embodied AI (EAI) systems. It presents an examination of current learning paradigms’ energy footprints and introduces ”collective learning” as a novel approach to significantly reduce the energy consumption associated with EAI skill acquisition. The discussion within this paper models the dynamics of skill knowledge transfer and compares various learning paradigms, highlighting how collective knowledge sharing across EAI agents minimizes energy expenditure and accelerates learning efficiency. The presented analysis emphasizes the potential of this paradigm shift for achieving more energy-sustainable and scalable robotic intelligence.