Toward Scalable and Agent-Agnostic Human–Machine Collaboration in Dynamic Workspaces - Position Statement

Abstract

Human–machine collaboration (HMC) in complex, dy namic environments, such as construction sites, remains fundamentally constrained by scalability, safety guarantees and model dependence. These environments typically involve heterogeneous agents (humans, robots and heavy machinery), static obstacles and rapidly evolving task conditions. Current approaches often rely on computationally intensive perception pipelines, human-specific behavioural models or centralised planning architectures that do not scale efficiently as the number of agents in creases. We argue for a lightweight, decentralised and agent agnostic framework for collaboration that guarantees safety whilst remaining adaptable to changing team compositions. Rather than redesigning control structures when additional agents are introduced, collaborative behaviour should emerge from shared physical principles embedded within the workspace.