According to a paper published on arxiv.org, researchers have introduced EI2ROC, a three-tier multi-agent architecture designed to enable autonomous intelligence in mining Integrated Remote Operation Centres (IROCs).
The paper addresses what it describes as two gaps in mining automation: the lack of an integrated agentic architecture that orchestrates Supervisory Control and Data Acquisition systems, digital twins, predictive analytics and human-machine interfaces under safety-critical constraints, and the absence of an evaluation framework that addresses real-time, Health-Safety-Environment, regulatory-auditability and cost-per-decision requirements for industrial agentic deployments.
According to arxiv.org, the EI2ROC architecture combines a Large Language Model orchestrator with specialized monitoring and production-capacity-prediction agents that communicate through a Model Context Protocol-style tool interface. The paper also proposes a multi-dimensional benchmarking framework for industrial agentic AI comprising a scenario library, harness, metrics and human-calibration protocol.
The paper explicitly states which components are implemented in the current prototype versus those belonging to the proposed reference design or future work, and reports preliminary cost and latency observations from an operating site. The research was presented at Strategic Mine Planning 2026 in Perth, Australia in November 2026.