🎓 Lesson 15
D5
Fleet Dispatch Algorithm Selection Criteria
A fleet dispatch algorithm is a set of rules that tells autonomous haul trucks which loader to go to, when to leave, and which dump point to use — all to keep the whole mining operation running as smoothly and efficiently as possible.
🎯 Learning Objectives
- ✓ Analyze trade-offs between FCFS, shortest-queue, and priority-based dispatch algorithms using cycle time and utilization metrics
- ✓ Design a weighted priority dispatch rule for mixed-fleet operations (e.g., 130t and 240t trucks) given payload capacity, travel time, and maintenance status
- ✓ Apply queuing theory fundamentals to calculate expected waiting time at loaders under Poisson arrival assumptions
- ✓ Explain how real-time traffic conflict resolution impacts dispatch algorithm selection in high-density AHS environments
- ✓ Evaluate algorithm performance using KPIs (e.g., % idle time, average load wait, payload variance) against ISEE/SAE J2980 benchmarks
📖 Why This Matters
In modern autonomous mines, up to 70% of haul truck downtime stems not from mechanical failure—but from inefficient dispatch decisions: trucks idling at loaders, unnecessary detours, or unbalanced workloads across shovels. Selecting the right dispatch algorithm directly impacts productivity (up to ±18% fleet throughput), fuel use (±12% per tonne), and even tyre life (via reduced queuing abrasion). With over $2B invested globally in AHS fleets (McKinsey, 2023), choosing the wrong algorithm can delay ROI by 14–22 months — making this not just a software question, but a core mine planning and operational discipline.
📘 Core Principles
Dispatch algorithms operate at three conceptual layers: (1) Assignment logic (which truck serves which shovel/dump), (2) Sequencing logic (order of service when multiple trucks compete), and (3) Adaptation logic (how the algorithm responds to disruptions—e.g., shovel delay, GPS outage, or battery thermal throttling). Rule-based systems (e.g., First-Come-First-Served) are simple and stable but ignore payload mismatch and dynamic congestion. Optimization-based systems (e.g., Mixed-Integer Linear Programming) maximize global objectives but require <500ms solve times and robust constraint handling. Hybrid approaches—like the Rio Tinto ‘Dynamic Priority’ model—combine real-time weightings (distance, payload ratio, predicted shovel fill time, battery SOC) with hard safety constraints (minimum 35m separation, max 3 trucks per shovel queue). Critically, all viable algorithms must comply with ISO 13849-1 PLd functional safety requirements for automated vehicle coordination.
📐 Weighted Priority Score (WPS)
The Weighted Priority Score quantifies suitability of assigning Truck i to Loader j at time t. Higher scores indicate better matches. Used in adaptive dispatch systems to balance efficiency and fairness across heterogeneous fleets.
Weighted Priority Score (WPS)
WPSᵢⱼ(t) = w₁·Dᵢⱼ(t) + w₂·Pᵢⱼ(t) + w₃·Sᵢ(t)Composite score evaluating suitability of assigning truck i to loader j at time t; used for dynamic dispatch ranking.
Variables:
| Symbol | Name | Unit | Description |
|---|---|---|---|
| WPSᵢⱼ(t) | Weighted Priority Score | dimensionless | Normalized suitability index (0.0–1.0) for truck i → loader j assignment |
| Dᵢⱼ(t) | Normalized Distance Factor | dimensionless | 1 − (distanceᵢⱼ / Dₘₐₓ); Dₘₐₓ = system-defined max dispatch range (e.g., 800 m) |
| Pᵢⱼ(t) | Payload Match Ratio | dimensionless | min(payloadⱼ / capacityᵢ, 1.0); capped at 1.0 to prevent penalty for oversized trucks |
| Sᵢ(t) | State-of-Charge Normalization | dimensionless | (SOCᵢ(t) − SOCₘᵢₙ) / (SOCₘₐₓ − SOCₘᵢₙ); typical SOCₘᵢₙ = 20%, SOCₘₐₓ = 95% |
| w₁, w₂, w₃ | Weight Coefficients | dimensionless | User-configurable, sum to 1.0; adjusted via KPI feedback or mine plan phase (e.g., ramp-up vs. steady-state) |
Typical Ranges:
Baseline deployment: w₁ = 0.25–0.45, w₂ = 0.3–0.5, w₃ = 0.15–0.3
Battery-constrained operation: w₃ ≥ 0.35
💡 Worked Example
Problem: Truck T7 (240t capacity, SOC=82%, current location 420m from Loader L3) competes with Truck T12 (130t capacity, SOC=65%, 310m from L3). Loader L3’s next cycle is estimated at 48s; its last payload was 225t. Assign weights: distance (w₁=0.4), payload match (w₂=0.35), SOC (w₃=0.25). Normalize each metric to [0,1] where 1 = best.
1.
Step 1: Normalize distance — min distance = 310m → T7 score = 1 − (420−310)/420 = 0.74; T12 = 1.0
2.
Step 2: Normalize payload match — ideal = 225t → T7 match = 225/240 = 0.94; T12 = 225/130 = 1.0 (capped) → normalized = 1.0
3.
Step 3: Normalize SOC — T7 = 0.82, T12 = 0.65 → scale linearly to [0,1] assuming min acceptable = 20% → T7 = (0.82−0.2)/(0.9−0.2) = 0.89; T12 = (0.65−0.2)/0.7 = 0.64
4.
Step 4: Compute WPS — T7 = 0.4×0.74 + 0.35×0.94 + 0.25×0.89 = 0.296 + 0.329 + 0.222 = 0.847; T12 = 0.4×1.0 + 0.35×1.0 + 0.25×0.64 = 0.4 + 0.35 + 0.16 = 0.91
Answer:
T12 receives higher priority (WPS = 0.91 > 0.847), despite lower SOC, due to superior distance and perfect payload match. This reflects real-world practice: underfilling a 240t truck wastes capacity, while a 130t truck achieves near-full utilization.
🏗️ Real-World Application
At BHP’s South Flank iron ore mine (WA), the transition from FCFS to the ‘Adaptive Load Priority’ algorithm (developed with Hexagon Mining) reduced average shovel wait time from 112 s to 68 s and increased fleet utilization from 71% to 83%. Crucially, the new algorithm incorporated real-time shovel bucket fill estimation (via LiDAR + AI) and dynamic reweighting: when shovel availability dropped below 85%, payload-matching weight increased from 0.35 to 0.55, preventing chronic underloading of 240t trucks. This case is documented in the 2022 SME Annual Meeting Paper No. 22-117 and validated against SAE J2980 Annex D verification protocols.
✏️ Dispatch Algorithm Comparison Exercise
Given: Three trucks (T1: 130t, 78% SOC, 510m from shovel; T2: 240t, 91% SOC, 630m; T3: 130t, 44% SOC, 390m). Shovel next cycle = 52s; last payload = 128t. Using WPS with weights w₁=0.3 (distance), w₂=0.4 (payload match), w₃=0.3 (SOC), and normalization bounds (distance: 0–800m; SOC: 20–95%), calculate WPS for each truck and recommend assignment. Then explain why your top choice is operationally preferable—even if not numerically highest—if shovel cycle time increases to 72s.
🔧 Interactive Calculator
🔧 Open Autonomous Haulage Systems Integration Calculator📋 Case Connection
📋 Open Pit Iron Ore Mine Fleet Harmonization at Rio Tinto Pilbara
Heterogeneous OEM protocols causing dispatch system latency >12s and payload mismatch errors