Collection and unloading routes

Collection routes were modeled with plant visits, weighing, unloading and restart, so vehicle load and work time could be reviewed over complete runs.

Industry
Resource collection
Decision family
Sequence optimization
Evaluation scope
Run reproduction study

Customer identity and identifying operational details are withheld.

The operating decision

Nearby pickups can still create a long or overloaded run once collection work, plant visits, weighing and unloading are included.

Inputs
Pickup locations · Collection workload · Vehicle capacity · Plant and weighing locations · Measured course and work times · Navigation travel times
Decision
Pickup assignment · Visit order · Plant visits · Unloading and restart points
Constraints
Vehicle load limit · Working time · Required weighing and unloading sequence

How it is approached

  1. Model a complete collection cycle

    Represent collection, weighing, unloading and the next trip as one sequence instead of treating unloading as an afterthought.

  2. Check load and working time

    Evaluate each pickup run with its plant visits, vehicle capacity and service work included.

  3. Validate the timing model first

    Replay measured courses while fixing visit order, departure and work time, and recalculate travel time from navigation sources to inspect timing error.

Check that the model reproduces the run.

Actual and replayed total run time for three anonymous collection courses.
Historical route-replay validation: actual and replayed total run time for three anonymous courses. This checks reproduction accuracy; the differences are not optimization savings. The source also notes leg-level mismatch and excludes unexplained return legs.

Run reproduction study

Validation & scope

The validation examines reproduction of measured run times, not time or cost savings from optimized routes. Unexplained return legs were excluded, and per-leg checks are necessary because errors can cancel in a total. The delivered demonstration and timing evidence have distinct scopes.

Bring the decision, its data and its constraints.

Back to applications Discuss your operation