<![CDATA[Off-highway equipment is undergoing a transition toward autonomous and AI-enabled operation under demanding duty cycles, limited connectivity, environmental variability, and persistent labor shortages. This review examines the engineering requirements for deploying physical AI in heavy equipment, with particular emphasis on perception, functional safety, validation, and autonomous operation. The review synthesizes recent literature published from 2020 onward across five interconnected domains: multi-sensor perception and scene understanding; functional safety and AI-related standards; simulation and digital twin infrastructure; autonomous haulage in surface and underground mining; and autonomous navigation in agricultural machinery. The analysis identifies a growing gap between the capabilities of field-oriented perception and control architectures and the validation provisions required to demonstrate safe and reliable AI behavior. Across the reviewed domains, simulation, digital twins, model-in-the-loop, software-in-the-loop, and hardware-in-the-loop testing increasingly constitute essential infrastructure for evaluating safety-relevant AI behavior before field deployment. The review further indicates that validation capacity is becoming a critical engineering constraint for scaling physical AI, particularly when perception uncertainty, environmental variability, sensor degradation, communication limitations, and hardware–software interactions must be assessed under representative operating conditions. These findings show that autonomous capability development alone is insufficient to support operational deployment without a corresponding validation architecture. The review therefore proposes a validation-centric sequencing framework that connects perception development, simulation, model- and software-in-the-loop testing, hardware-in-the-loop verification, safety assurance, and field validation. The framework provides an engineering basis for organizing progressive validation activities, establishing traceability between development and verification, and reducing the gap between AI capability development and reliable operational deployment of autonomous off-highway equipment]]>
Copyrights © 2026
