The Berth Allocation Problem with Channel Restrictions - Datasets
These datatasets relate to the computational study presented in the paper "The Berth Allocation Problem with Channel Restrictions", authored by Paul Corry and Christian Bierwirth. They consist of all the randomly generated problem instances along with the computational results presented in the paper.
Results across all problem instances assume ship separation parameters of [delta_1, delta_2, delta_3] = [0.25, 0, 0.5].
Excel Workbook Organisation:
The data is organised into separate Excel files for each table in the paper, as indicated by the file description. Within each file, each row of data presented (aggregating 10 replications) in the corrsponding table is captured in two worksheets, one with the problem instance data, and the other with generated solution data obtained from several solution methods (described in the paper). For example, row 3 of Tab. 2, will have data for 10 problem instances on worksheet T2R3, and corresponding solution data on T2R3X.
Problem Instance Data Format:
On each problem instance worksheet (e.g. T2R3), each row of data corresponds to a different problem instance, and there are 10 replications on each worksheet.
The first column provides a replication identifier which is referenced on the corresponding solution worksheet (e.g. T2R3X).
Following this, there are n*(2c+1) columns (n = number of ships, c = number of channel segmenets) with headers p(i)_(j).(k)., where i references the operation (channel transit/berth visit) id, j references the ship id, and k references the index of the operation within the ship. All indexing starts at 0. These columns define the transit or dwell times on each segment. A value of -1 indicates a segment on which a berth allocation must be applied, and hence the dwell time is unkown.
There are then a further n columns with headers r(j), defining the release times of each ship.
For ChSP problems, there are a final n colums with headers b(j), defining the berth to be visited by each ship. ChSP problems with fixed berth sequencing enforced have an additional n columns with headers toa(j), indicating the order in which ship j sits within its berth sequence. For BAP-CR problems, these columnns are not present, but replaced by n*m columns (m = number of berths) with headers p(j).(b) defining the berth processing time of ship j if allocated to berth b.
Solution Data Format:
Each row of data corresponds to a different solution.
Column A references the replication identifier (from the corresponding instance worksheet) that the soluion refers to.
Column B defines the algorithm that was used to generate the solution.
Column C shows the objective function value (total waiting and excess handling time) obtained.
Column D shows the CPU time consumed in generating the solution, rounded to the nearest second.
Column E shows the optimality gap as a proportion. A value of -1 or an empty value indicates that optimality gap is unknown.
From column F onwards, there are are n*(2c+1) columns with the previously described p(i)_(j).(k). headers. The values in these columns define the entry times at each segment.
For BAP-CR problems only, following this there are a further 2n columns. For each ship j, there will be columns titled b(j) and p.b(j) defining the berth that was allocated to ship j, and the processing time on that berth respectively.
Cite this collection
Creative Commons Attribution 4.0 (CC-BY)