🎓 Lesson 7
D5
Dispatch Algorithms: Fixed Interval vs. Real-Time Adaptive
Fixed interval dispatching sends trucks to loading points at regular time intervals, while real-time adaptive dispatching adjusts truck assignments instantly based on live data like queue length, shovel status, and travel time.
🎯 Learning Objectives
- ✓ Compare fleet productivity (tonnes/hour/truck) under fixed-interval vs. real-time adaptive dispatch using simulated cycle time data
- ✓ Design a basic real-time adaptive dispatch rule set for a 12-truck, 2-shovel open-pit operation
- ✓ Analyze dispatch algorithm impact on key performance indicators (KPIs) including wait time standard deviation, shovel utilization, and fuel consumption per tonne
- ✓ Apply queuing theory fundamentals to calculate expected queue length under fixed-interval dispatch with Poisson-distributed truck arrivals
📖 Why This Matters
In modern high-tonnage open-pit mines, even a 5% improvement in haul truck utilization translates to millions in annual cost savings—and reduces carbon intensity per tonne. Fixed-interval dispatch, once the industry default, often leads to 'shovel starvation' or 'truck queuing', both eroding productivity. Real-time adaptive systems—now deployed at Rio Tinto’s Pilbara operations and BHP’s South Flank—demonstrate 8–12% higher fleet throughput and 30% lower average wait times. Understanding when and how to deploy each algorithm is foundational to designing resilient, efficient, and sustainable haulage systems.
📘 Core Principles
Dispatch algorithms govern how haul trucks are allocated to loading and dumping points. Fixed-interval dispatch assumes steady-state conditions: trucks are cycled at fixed clock-based intervals (e.g., every 90 seconds), making it simple to implement but brittle under variability. Real-time adaptive dispatch treats the haul cycle as a dynamic optimization problem—using feedback from telematics (GPS position, engine load, payload weight) and machine status (shovel ready/idle/down) to reassign trucks continuously. It leverages concepts from discrete-event simulation, queuing theory (M/G/c models), and constrained optimization (e.g., minimizing total weighted tardiness). Critical enablers include reliable communication infrastructure (LTE/5G private networks), low-latency edge computing, and accurate digital twins of haul roads and equipment.
📐 Expected Queue Length under Fixed-Interval Dispatch
When truck arrivals approximate a Poisson process (common in large fleets), fixed-interval dispatch creates artificial periodicity that violates Poisson assumptions—leading to queue buildup. Using Little’s Law and M/D/c approximation, we estimate average queue length Lq to benchmark inefficiency.
💡 Worked Example
Problem: A 2-shovel, 16-truck system operates with fixed dispatch every 110 s. Observed average shovel cycle time = 72 s (σ = 8 s), average haul + dump time = 142 s (σ = 18 s). Assume Poisson truck arrivals (λ = 1/110 s⁻¹ per shovel) and deterministic service (μ = 1/72 s⁻¹). Calculate expected queue length Lq.
1.
Step 1: Convert rates to per-hour: λ = 3600/110 ≈ 32.73 trucks/hour/shovel → total λ = 65.45 trucks/h; μ = 3600/72 = 50 trucks/h/shovel → total service rate = 100 trucks/h
2.
Step 2: Compute traffic intensity ρ = λ/(c·μ) = 65.45/(2×50) = 0.6545
3.
Step 3: Apply M/D/c formula: Lq ≈ (ρ² / (1 − ρ)) × (1 + CV²ₛ/2) / (2c), where CVₛ = σ_service / mean_service = 8/72 ≈ 0.111 → CV²ₛ ≈ 0.0123 → Lq ≈ (0.6545² / (1−0.6545)) × (1+0.0123)/4 ≈ (0.428 / 0.3455) × 0.253 ≈ 1.239 × 0.253 ≈ 0.313
4.
Step 4: Interpret: Avg. queue length ≈ 0.31 trucks — seemingly low, but field data shows actual queues exceed 1.2 due to unmodeled variability (e.g., road grade changes, radio delays); this highlights model limitation and need for adaptive response.
Answer:
The theoretical Lq is 0.31 trucks, but empirical validation at Newmont’s Boddington mine showed median observed queue = 1.4 trucks — confirming fixed-interval dispatch underestimates queuing under real-world stochasticity.
🏗️ Real-World Application
At Vale’s S11D iron ore mine (Carajás, Brazil), implementation of a real-time adaptive dispatch system (using MineSite™ Fleet Management integrated with Cat Command for Hauling) reduced average truck wait time at shovels from 221 s to 154 s—a 30% reduction—and increased shovel utilization from 68% to 81%. The system uses shovel cycle telemetry, real-time GPS positioning, and predictive travel time models (calibrated hourly using onboard accelerometer and inclinometer data) to assign the nearest available truck with sufficient payload capacity and compatible dump destination. Crucially, it overrides pre-scheduled intervals during shovel maintenance events—rerouting trucks to alternate shovels within 8 seconds of fault detection.
🔧 Interactive Calculator
🔧 Open Fleet & Dispatch📋 Case Connection
📋 Chilean Copper Mine: Autonomous Haul Fleet Deployment
Achieving safe, reliable, and productive autonomous haulage under extreme environmental conditions (high altitude: 3,200...
📋 Australian Iron Ore Open Pit: Conveyors vs. Trucking Economic Threshold
Determine the economic breakeven point (tonnes per annum and distance) at which fixed-cost conveyor systems become more...
📋 Canadian Gold Mine: Steep Ramp Optimization in Narrow Vein Underground
Excessive truck cycle times and premature tire/brake wear due to suboptimal ramp gradient (15%) combined with tight hori...
📋 South African Platinum Mine: Waste Dump Reclaim Optimization
Inefficient haulage routing and underutilized fleet capacity during waste dump reclamation, resulting in excessive diese...