Autonomous Haulage Dispatch Map (2D)

Truck longitudinal dynamics

Wheel-drive force is limited by power and by traction: F_t = min(ηP / v, μ W_d) with ηP the power at the wheels and W_d the weight on the driven axle. Resistance is F_r = m g (C_rr + G) + ½ ρ C_d A v² (rolling + grade + air). Acceleration is a = (F_t − F_r) / m. On a long climb the speed settles where ηP / v = m g (C_rr + G) + ½ ρ C_d A v²; for a 385 t loaded truck on an 8 % ramp with 2.35 MW this is about 18.7 km/h, which the first model check reproduces from the integrated simulation.

Downhill the retarder absorbs the excess: loaded trucks are limited to 9 m/s on grades steeper than 3 %. In curves the speed is capped by lateral acceleration, v ≤ √(a_lat R), with R from the road centre-line curvature.

Following other trucks

Each truck uses an intelligent-driver style law a = a_free (1 − (v/v0)&sup4;) − a_c (s*/s)² with desired gap s* = s0 + vT + vΔv / (2√(a_c b)). Autonomous trucks use a shorter time gap T and no driver-to-driver speed spread; manual mode lengthens T, lowers the speed factor and adds spotting delay and variability.

Queues and production

A shovel, the crusher and the fuel bay are FIFO servers. Utilisation is ρ = λ / (c μ); for a single exponential server the mean time in system is W = 1 / (μ − λ) and Kingman's approximation W_q ≈ ρ/(1−ρ) · (c_a² + c_s²)/2 · E[S] shows why variability alone creates queues. Little's law L = λ W holds for any stable system and is the third check. The cycle is T = T_load + T_haul(loaded) + T_dump + T_return(empty) + T_queue; the production bound drawn on the graphs is min( N · payload / T_0 , n_shovel · payload / T_load , n_pocket · payload / T_dump ).

Dispatch and routing

Fixed sends every truck to a fixed shovel, nearest to the shovel with the shortest free-flow trip, shortest queue to the shovel with the fewest assigned trucks, and min. expected cycle to the shovel that minimises max(travel time, assigned trucks × service time), a simplified version of the greedy rules used in commercial dispatch systems. Routes are Dijkstra shortest paths on expected travel time; with congestion-aware routing each truck already on a lane adds a penalty, so the bypass road is used when the crusher road is busy.

Fuel

Engine power is the traction power divided by η (floor: idle 110 kW); fuel is kWh × 0.215 kg/kWh / 0.84 kg/L. Trucks below 22 % of a 4000 L tank visit the fuel bay.

Honest simplifications

One lane per direction, no overtaking and no junction conflict logic; each road segment has a constant grade; payload is fixed per cycle; no blasting, weather, tyre-heat (TKPH) or operator shifts; the retarder is a speed cap rather than a power model; truck and shovel dimensions follow a 220 t class but the numbers are illustrative, not a vendor specification. Autonomy parameters (gap, spotting time, variability) are plausible assumptions, not measured fleet data. This page is an educational model and is not affiliated with any mining-equipment manufacturer.

• Press Too few trucks, then Too many trucks, then Sweep fleet size: production rises, then saturates at the shovel/crusher bound while cycle time and diesel per tonne keep growing.
• Press Nearest-shovel dispatch and watch the queue at shovel A while shovel B sits idle; then compare all four policies with Compare dispatch policies.
• Press Stop a truck on the road and follow the queue wave on the ramp graph and in the queue plot.
• Toggle Autonomous / Manual and compare cycle time, queues and fuel; then raise the time gap to 3 s.
• Set rolling resistance to 6 %: the loaded climb speed drops, cycle time grows and diesel per tonne rises.
• Click a truck to follow its speed and elevation profile over the last ten minutes.