Optimized Human-Robot Co-Dispatch Planning for Petro-Site Surveillance under Varying Criticalities

Securing petroleum infrastructure requires balancing autonomous system efficiency with human judgment for threat escalation, a challenge unaddressed by classical facility location models assuming homogeneous resources. This paper formulates the Human-Robot Co-Dispatch Facility Location Problem (HRCD-FLP), a capacitated facility location variant incorporating tiered infrastructure criticality, human-robot supervision ratio constraints, and minimum utilization requirements. We evaluate command center selection across three technology maturity scenarios. Results show transitioning from conservative (1:3 human-robot supervision) to future autonomous operations (1:10) yields significant cost reduction while maintaining complete critical infrastructure coverage. For small problems, exact methods dominate in both cost and computation time; for larger problems, the proposed heuristic achieves feasible solutions in under 3 minutes with approximately 14% optimality gap where comparison is possible. From systems perspective, our work demonstrate that optimized planning for human-robot teaming is key to achieve both cost-effective and mission-reliable deployments.

💡 Research Summary

The paper addresses the emerging need to protect petroleum infrastructure by integrating autonomous robotic surveillance with human oversight. Classical facility location models treat resources as homogeneous and ignore the hierarchical criticality of assets, making them unsuitable for modern security operations that require both continuous monitoring and ethical decision‑making. To fill this gap, the authors formulate the Human‑Robot Co‑Dispatch Facility Location Problem (HRCD‑FLP), a capacitated facility location model that simultaneously decides (i) which candidate sites should host command centers and at which operational level (Low, Medium, High), (ii) how each demand site (refineries, processing plants, pipelines, etc.) is assigned to a command center, and (iii) the mix of human operators and autonomous robots deployed at each opened facility.

Key model ingredients:

• Tiered asset criticality – each demand point belongs to a criticality tier that defines a surveillance coverage requirement (SCU) and a service‑level agreement (maximum allowable response time).
• Supervision ratio (α) – a policy parameter that enforces a minimum human‑to‑robot ratio (e.g., 1:3, 1:5, 1:10) reflecting technology maturity and regulatory constraints.
• Facility levels – three discrete levels with distinct fixed construction/operational costs, capacity limits for humans and robots, and response‑time capabilities.
• Coverage split (α_j) – site‑specific proportion of human effort required due to task complexity or escalation likelihood.

The MILP objective minimizes total system cost, i.e., the sum of fixed facility costs and variable deployment costs for humans and robots. Constraints enforce (1) at most one level per candidate site, (2) every demand point is assigned to at least one active facility, (3) assignments respect response‑time SLAs, (4) resource quantities stay within level‑specific upper and lower bounds, (5) the deployed robots and humans collectively satisfy the SCU demand of assigned sites, and (6) the global supervision ratio is respected across all facilities.

Because the problem size grows quickly (binary variables for facility‑level selection and assignments, integer variables for resource quantities), the authors develop a two‑phase heuristic for larger instances. Phase 1 uses a Lagrangian‑relaxation‑guided greedy algorithm to select facilities and assign demand points. Phase 2 solves a linear program to fine‑tune the human‑robot mix given the fixed assignments. Computational experiments on a realistic case study in the Dhahran district and on synthetically generated instances show:

• For small instances (≤30 demand points) commercial MILP solvers (CPLEX, Gurobi) find optimal solutions in sub‑second time.
• For larger instances (up to 200 demand points) the heuristic produces feasible solutions within 3 minutes, with an average optimality gap of ≤14 % compared to the best known MILP bound.
• Varying the supervision ratio from a conservative 1:3 to a future‑oriented 1:10 reduces total cost by roughly 38 % while still meeting 100 % coverage of all critical assets. High‑criticality Tier‑1 facilities continue to be served by high‑level command centers to satisfy stringent SLAs, whereas lower‑criticality assets are covered by more distributed, lower‑level facilities.

The study demonstrates that incorporating human‑robot teaming constraints into facility location decisions yields substantial cost savings without compromising security performance. It also provides a quantitative tool for policymakers to evaluate the trade‑off between autonomy level (supervision ratio) and investment in command‑center infrastructure.

Limitations include the static nature of the model (no consideration of robot failures, maintenance schedules, human fatigue, or dynamic threat evolution) and the assumption of a single‑sourcing assignment (each demand point served by exactly one facility). Future work is suggested to extend the framework to multi‑period planning, stochastic demand, and richer human‑robot interaction cost models.

Overall, the paper makes a novel contribution by bridging critical infrastructure protection, facility location theory, and human‑robot collaboration, offering both a rigorous mathematical formulation and a practical solution approach for real‑world petroleum‑site surveillance.