Skip to main content

Crop Transport Routing

CAPACITATED VEHICLE ROUTING

Agriculture · Distribution · Operational Canonical method: CVRP

During the ten-day corn harvest window on a 1,200 ha Châteauguay-valley grain-oilseed farm, a fleet of grain trucks runs continuously between the combine-filled field edges and the central silo. Each truck has a finite capacity; each field has a known tonnage ready to move. The operational question — which truck visits which fields, in what order — is the Capacitated Vehicle Routing Problem (CVRP), an NP-hard generalization of TSP that admits strong heuristics (Clarke-Wright savings) and exact methods (branch-cut-price) for the scales typical of a single farm.

Why this decision matters

Three verified facts on agricultural transport

~14%
of global food produced is lost between post-harvest and retail — much of it on the farm-to-silo and silo-to-processor legs.
FAO, State of Food and Agriculture 2019 · fao.org
3.4 Mt
of grain corn produced in Québec in the 2022 crop year — every tonne requires a short-haul movement from field to on-farm silo or commercial elevator.
Statistics Canada, Table 32-10-0359-01 · StatCan field crops table
5–15%
of total distance travelled is savings typically delivered by good CVRP heuristics (Clarke-Wright, sweep) versus naive one-field-per-truck routing.
Toth & Vigo (2014), Vehicle Routing · SIAM

Where this decision fits

Agriculture sector decision chain · operational, in-season horizon

Strategiccrops & fields
Workforceseasonal labour
Plantingseed & inputs
Growingirrigation & pests
Harvestscheduling
Transportthis page

Problem & formulation

Classical CVRP with Miller-Tucker-Zemlin subtour elimination

OR family
CVRP
Complexity
NP-hard
Solver realism
★★☆ Heuristic (Clarke-Wright)
Reference
Dantzig & Ramser (1959)

Sets & indices

SymbolMeaningDomain
\(V = \{0, 1, \ldots, n\}\)Nodes: 0 = central silo (depot), \(1..n\) = field pickup points\(|V| = n+1\)
\(A = V \times V\)Directed arcs between every pair of nodes\(|A| = (n+1)^2\)
\(K\)Homogeneous truck fleet available during the harvest dayfinite

Parameters

SymbolMeaningUnit
\(d_{ij}\)Road distance from node \(i\) to node \(j\)km
\(q_i\)Tonnes ready to collect at field \(i\) (\(q_0 = 0\))t
\(Q\)Payload capacity per truckt

Decision variables

SymbolMeaningDomain
\(x_{ij}\)1 if some truck traverses arc \((i,j)\), 0 otherwise\(\{0,1\}\)
\(u_i\)Cumulative load when a truck arrives at node \(i\) (MTZ)\([q_i, Q]\)

Objective

Minimize total distance travelled by the fleet:

$$\min \; \sum_{(i,j) \in A} d_{ij} \, x_{ij}$$

Constraints

Each field is visited exactly once; the depot is entered and left by each vehicle; capacity is respected; subtours are eliminated:

$$\sum_{j \in V} x_{ij} \;=\; 1 \qquad \forall\, i \in V \setminus \{0\} \qquad \text{(leave each field once)}$$ $$\sum_{i \in V} x_{ij} \;=\; 1 \qquad \forall\, j \in V \setminus \{0\} \qquad \text{(enter each field once)}$$ $$\sum_{j \in V \setminus \{0\}} x_{0j} \;\leq\; |K| \qquad \text{(trucks leaving the depot)}$$ $$u_j \;\geq\; u_i + q_j - Q\,(1 - x_{ij}) \qquad \forall\, i,j \in V \setminus \{0\},\ i \neq j \qquad \text{(MTZ: subtour + capacity)}$$ $$q_i \;\leq\; u_i \;\leq\; Q \qquad \forall\, i \in V \setminus \{0\}$$ $$x_{ij} \;\in\; \{0, 1\}$$

MTZ combines capacity and subtour elimination in \(\mathcal{O}(n^2)\) constraints. Stronger DFJ cuts or set-partitioning + column generation scale further.

Solution methods at a glance

Up to ~25 fields, MTZ solved by a MIP solver is practical. Beyond that, the Clarke-Wright savings heuristic (1964) is the classical first line: merge routes whenever the savings \(s_{ij} = d_{0i} + d_{0j} - d_{ij}\) of joining them is positive and capacity allows. State-of-the-art metaheuristics (LKH-3, HGS-CVRP) close the gap to optimum to under 1% on benchmark instances.

Real-world → OR mapping

Harvest-day vocabulary translated to the CVRP

On the farmIn the CVRP model
Central silo / grain elevator at the farmyardNode \(0\) (depot)
Each harvestable field waiting for pickupNode \(i \in V \setminus \{0\}\)
Road distance along gravel / country roads\(d_{ij}\)
Tonnes of corn, soy, or wheat ready in the field\(q_i\)
Grain-trailer payload (20–25 t typical)\(Q\)
Available truck-drivers that day\(|K|\)
Does truck \(k\) drive from \(i\) to \(j\)?\(x_{ij}\)
Total km driven by the fleetObjective

Interactive solver

Clarke-Wright savings on a representative Québec farm instance

CVRP Clarke-Wright solver
Fields around a central silo · homogeneous truck fleet · O(n² log n)
★★☆ Heuristic
Fields laid out in a 10km radius
Tandem grain trailer ≈ 22 t
Fixes field positions & yields
Total distance (km)
Routes (trucks used)
Total tonnes moved
vs naive baseline

How Clarke-Wright works

Start with a “one field per truck” plan: \(n\) trivial routes \(0 \to i \to 0\). For each pair of fields, compute the savings \(s_{ij} = d_{0i} + d_{0j} - d_{ij}\): distance saved if a truck visited \(i\) and \(j\) in sequence instead of two separate round trips. Sort the savings descending; greedily merge route-endpoints whenever the merged route still fits within capacity. Output routes are typically 5–15% shorter than the baseline and within a few percent of the proven optimum.

Reading the solution

What the route map tells the farm manager

Three patterns to watch for

  • Cluster routes vs. radial routes. When field yields are small, Clarke-Wright merges many adjacent fields into long “cluster” routes. When yields approach capacity \(Q\), routes become near-radial (one trip per field).
  • Truck count vs. makespan. Total distance is weakly sensitive to an extra truck — the real cost driver is per-km fuel and driver-hour. But more trucks in parallel shorten the harvest-day makespan.
  • Close-to-depot fillers. Fields near the silo become opportunistic “short-hop” fillers when a truck has capacity left after longer legs.

Sensitivity questions the solver answers instantly

  • Capacity drops (older trucks, wet corn adding weight)? Reduce \(Q\) — watch route count grow.
  • Add a far-off rented field? Use a larger seed / re-spatialize to see how the extra 5–10 km reshapes routes.
  • One field much bigger than the others? Raise max yield — a dedicated near-full truck is likely.

Model extensions

Natural agricultural generalizations of the base CVRP

Time windows (VRPTW)

If trucks can only arrive during a combine's unload window (say, every 90 minutes), the problem becomes VRPTW. Sharper constraints, tighter solutions, but materially harder to solve.

See the fertilizer-routing VRPTW →
Stochastic demand

Actual tonnage depends on local yield variance (wet spots, uneven emergence). Stochastic CVRP with chance-constrained capacity or two-stage recourse prevents routes from running short by one or two tonnes.

See the operational cell →
Heterogeneous fleet

Mixing a 22 t tandem with a 15 t single-axle trailer makes vehicle assignment part of the decision. The LP relaxation + branch-and-price approach (SDVRP, VRPHF) is standard.

Related applications →

Key references

Cited above · DOIs and permanent URLs

Dantzig, G. B., & Ramser, J. H. (1959).
“The truck dispatching problem.”
Management Science, 6(1), 80–91. doi:10.1287/mnsc.6.1.80
Clarke, G., & Wright, J. W. (1964).
“Scheduling of vehicles from a central depot to a number of delivery points.”
Operations Research, 12(4), 568–581. doi:10.1287/opre.12.4.568
Miller, C. E., Tucker, A. W., & Zemlin, R. A. (1960).
“Integer programming formulation of traveling salesman problems.”
Journal of the ACM, 7(4), 326–329. doi:10.1145/321043.321046
Toth, P., & Vigo, D. (Eds.) (2014).
Vehicle Routing: Problems, Methods, and Applications (2nd ed.).
Ahumada, O., & Villalobos, J. R. (2009).
“Application of planning models in the agri-food supply chain: A review.”
European Journal of Operational Research, 196(1), 1–20. doi:10.1016/j.ejor.2008.02.014
FAO. (2019).
The State of Food and Agriculture 2019: Moving forward on food loss and waste reduction.
Statistics Canada.
Table 32-10-0359-01 — Estimated areas, yield, production, average farm price and total farm value of principal field crops.

Need to optimize harvest transport
on your farm or cooperative?

Get in Touch
Data and numerical examples are illustrative (representative Quebec grain-oilseed farm). This is an educational tool, not production software.