Summary
A client’s Vehicle Routing Problem (VRP) experienced significant solver slowdown when modeling large package deliveries exceeding vehicle capacity. The design split oversized demands into separate Visit entities, each with two planning variables: a boolean active flag and an integer demand allocation (8-16 units). Server performance degraded exponentially as visit counts increased due to combinatorial explosion from the dual-variable approach. The root cause was modeling inefficiency in how demand partitioning was represented.
Root Cause
The critical performance bottleneck originated from:
- Independent planning variables (
activeanddemand) creating multiplicative state combinations perVisit