Skip to content

Framework Example 5: VRPTW Branch-and-Price — Overview ​

中文

1. Overview ​

Demo5 is the runnable VRPTW integration. It builds a route-based restricted master, prices elementary resource-constrained routes, and applies branch-and-price through the network-scheduling framework.

2. Contexts and Dependencies ​

ContextResponsibilityDependency
VRPCustomers, vehicle types, routes, resources, and validationNormalized input and calculation policies
Route generationResource-constrained elementary-route pricingVRP, dual prices, branch rules
Route compilationCustomer/fleet rows, artificial coverage, and phase objectivesVRP, generated routes

The application coordinates branch nodes and phase transitions. Generation and compilation exchange route columns and dual prices.

3. Concepts, Sets, and Predicates ​

C is the customer set, V the vehicle-type set, and Rb,t the inserted compatible routes at branch node b, iteration t. Predicates express customer coverage, elementarity, resource feasibility, and branch compatibility.

4. Variables and Intermediate Values ​

xr is route usage and ui is phase-I artificial coverage. The master derives customer coverage Yi and fleet usage Fv. Generation computes label load, service-start time, and reduced cost. The child pages distinguish input attributes, algorithm state, linear expressions, and solver variables.

5. Assertions, Constraints, and Objectives ​

Customer coverage satisfies Yi+ui=1 and fleet usage cannot exceed availability. Phase I minimizes artificial coverage; phase II fixes ui=0 and minimizes route cost. This is not a mixed objective using an arbitrary large constant. Pricing and validation enforce route capacity and time windows.

6. Algorithms and Lifecycle ​

The application seeds routes, solves node LPs, searches improving columns with ESPPRC, and handles fractional solutions and branch nodes. Pricing includes customer and fleet duals. A return constrained by a column limit is not automatically complete pricing.

7. Register → Construct → Solve → Analyze ​

VRP validates and normalizes input. Generation builds the pricing graph. Compilation registers the restricted master. Branch-and-price coordinates master solves, pricing, and phase transitions. Analysis turns integer route usage into a validated solution.

8. Source Entry Points ​

9. Kotlin/Rust Comparison and Design Decisions ​

Both framework examples use route columns. The equations distinguish node LPs from integer solutions and phase-I feasibility restoration from phase-II cost optimization. A direct arc-model test oracle does not replace the route-column production master.

10. Context Model Pages ​

11. Change Log ​

VersionChangeReason
1.1Aligned bilingual overviews, notation, and source entry pointsKeep the overview consistent with its context models