Loading Hall Column Generation Solution Quality Improvements

This feature, for the users of OTM's Loading Hall Column Generation Algorithm, provides a set of solution quality improvements. 

The improvements include the following:

ENABLE LOAD POINT PENALTIES WITH OPTIMIZE COST OBJECTIVE IN COLUMN GENERATION

All 3 objectives account for packing items based also on load point penalties along with the objective selected, when load point consideration is turned on by the logic config parameter USE LOAD POINT PENALTIES IN PACKING.

INTRODUCE A PENALTY SCALING FACTOR

When the user wants to control the impact of actual cost (or any other selected objective which translates to cost) vs penalty added as cost, a scaling factor can be used to increase or decrease the value of load point penalty for all the combinations entered by the user in Load Point Penalty power data screen. The parameter in the Conopt logic config ID for this value is LOAD POINT PENALTY SCALING FACTOR. Enter a higher value like 50 to multiply the load point penalty by 50 to generate a load-point optimal solution. Enter a smaller value like 0.001 to generate a cost-optimal solution. This is useful when there are multiple optimal solutions with respect to cost or load point penalty. Then this can be used to favour a solution with either objective. The default for the parameter is 1. Setting it to 0 will make the solution ignore load point penalties altogether.

ENABLE DYNAMIC PROGRAMMING TO GENERATE COLUMNS

Initially, the packing algorithm supported for load points was Quick Packing and Column Generation. With this enhancement Enumerative Algorithm (Dynamic Programming) can also be used to evaluate all possible packings for a container considering load point penalties. When used with Column Generation this will provide better solution overall for all containers when compared to quick packing. There can be a few cases of degeneracy, but turning on Quick Packing as well will solve the issue. The recommended setup to use would be Column Generation with both Enumerative and Quick Packing algorithm

Configuration for Loading Hall Solution Improvement

Loading Hall Tuning Parameters

This graphic shows the results before the application of the Enumerative Algorithm.   The idea is to minimize the loading halls visited while maximizing the utilization of the truck.

Loading Hall Example Before Improvement

Penalty for this case: 75 (QR-VF) + 50 (VF-VH) = 125 for 2nd shipment (3 loading hall combinations). And 50 for QR-VW for 3rd shipment. 175 overall.

After the use of the new algorithm with Quick Packing, Enumerative and Column Generation turned on and the objective is Optimize Cost.   .

With default scaling factor (1), the resulting shipments have fewer loading points while being loaded.  This results in better service.

Loading Hall Example After Improvement

Penalty for this case: 75 (QR-VH) for 1st shipment, 50 (QR-VW) for 2nd shipment, 50 (VF-VH) for 3rd shipment. 175 overall. Avoided 3 hall combination for same shipment.

The benefit is to maximize the loaded weight on each truck while minimizing the number of stops that need to be made in a large plant to pick up the items being shipped.

Steps to Enable

The planner needs to configure the Enumerative Algorithm in the Logic Configuration for CONOPT. (Container Optimization)