Read HFVRP instances #550

leonlan · 2024-03-26T07:27:33Z

This PR adds support for reading HFVRP instances and is part of #543.

Notes:

It is essential that you add a test when making code changes.
This keeps the code coverage level up, and helps ensure the changes work as intended.
When fixing a bug, you must add a test that would produce the bug in the master branch, and then show that it is fixed with the new code.
New code additions must be well formatted. Changes should pass the pre-commit workflow, which you can set up locally using pre-commit.
Docstring additions must render correctly, including escapes and LaTeX.

codecov · 2024-03-26T07:31:07Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 95.70%. Comparing base (02a7c03) to head (8e1058b).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #550      +/-   ##
==========================================
+ Coverage   95.68%   95.70%   +0.02%     
==========================================
  Files          73       73              
  Lines        3243     3260      +17     
  Branches      888      888              
==========================================
+ Hits         3103     3120      +17     
  Misses         48       48              
  Partials       92       92

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

pyvrp/read.py

leonlan · 2024-03-26T08:36:37Z

@N-Wouda Also check out https://github.com/PyVRP/Sandbox/pull/6 for the new instances and README. I'll bump that to PyVRP/Instances and update the submodule after review.

docs/source/dev/supported_vrplib_fields.rst

tests/data/MoreVehiclesThanSectionData.txt

tests/test_read.py

docs/source/dev/supported_vrplib_fields.rst

leonlan · 2024-03-27T08:49:09Z

I need to do a bit more testing on this, probably tomorrow. I have tried to solve all HFVRP instances with 10 second runtime, but there are some instances that don't result in feasible solutions, even for longer runtimes.

This instance is pretty annoying: even though we have an unlimited fleet, the initialization procedure uses vehicle type 0 by default. But some clients have demand that exceeds this vehicle type capacity.

❯ poetry run pyvrp HFVRP/X670-FSMF.vrp --max_runtime 100 --seed 0 --num_procs 6 --round_func exact

Details

Instance    OK  Obj.          Iters. (#)  Time (s)
----------  --  ------------  ----------  --------
X1001-FSMF   Y   105683793.0          16     10.28
X101-FSMFD   Y    34309886.0         912    10.014
 X106-FSMD   Y    14862233.0         615    10.007
   X110-HD   Y    13390527.0         752    10.007
 X115-HVRP   N           inf         731    10.008
 X120-FSMF   Y   286903060.0         218     10.06
 X125-HVRP   Y    91738535.0         491    10.011
X129-FSMFD   Y    59349184.0         285    10.019
 X134-FSMD   Y     9112301.0         576    10.013
   X139-HD   Y    12764782.0         614    10.018
 X143-FSMF   Y    39587282.0         581    10.033
 X148-HVRP   Y    73253691.0         311    10.006
X153-FSMFD   Y    26729273.0         392     10.01
   X157-HD   Y    15840920.0         318    10.025
 X162-FSMD   Y    28205268.0         608     10.02
 X167-FSMF   Y    32109154.0         414    10.004
 X172-HVRP   Y    96711271.0         387    10.013
X176-FSMFD   Y    78112981.0         373    10.016
   X181-HD   N           inf         182    10.006
 X186-FSMD   Y    30054000.0         444    10.011
 X190-FSMF   Y    19224320.0         251    10.012
 X195-FSMF   Y    83040781.0         318    10.034
   X200-HD   Y    57475675.0         257     10.02
 X204-FSMD   Y    38434092.0         400    10.019
X209-FSMFD   Y    35083366.0         308     10.04
 X214-HVRP   Y    15934261.0         291    10.005
   X219-HD   N           inf         135    10.012
 X223-HVRP   Y    75033659.0         138    10.018
X228-FSMFD   Y    49156151.0         200    10.017
 X233-FSMD   Y    14867156.0         328    10.007
 X237-FSMF   Y   456129168.0          82     10.03
X242-FSMFD   Y   528472202.0          66    10.103
 X247-HVRP   Y    47150559.0         212    10.025
 X251-FSMD   Y    26221058.0         175    10.032
 X256-FSMF   Y    53428709.0         298    10.019
   X261-HD   Y    25343986.0         225    10.022
   X266-HD   Y    70619921.0          88    10.038
 X270-FSMD   Y    57598095.0         212    10.026
 X275-HVRP   N           inf          84    10.039
 X280-FSMF   Y    71173297.0         194    10.066
X284-FSMFD   Y    32062900.0         145    10.016
 X289-HVRP   Y   123064090.0         171    10.022
   X294-HD   Y    39721752.0         230    10.013
 X298-FSMD   Y    22917325.0         194    10.021
X303-FSMFD   Y    42791517.0         239    10.018
 X308-FSMF   Y    64953953.0         153    10.016
 X313-FSMD   Y    80739410.0         123    10.058
 X317-HVRP   N           inf          55    10.074
   X322-HD   Y    37199894.0         177    10.002
X327-FSMFD   Y    36860077.0         142     10.05
 X331-FSMF   Y   756165652.0          33     10.41
 X336-FSMF   Y   223694906.0         129    10.056
 X344-FSMD   Y    30435673.0         131    10.061
 X351-HVRP   Y    53593212.0         115     10.02
   X359-HD   Y    42151766.0         100    10.009
X367-FSMFD   Y    61584467.0         125     10.01
   X376-HD   N           inf          24    10.062
 X384-FSMF   Y   137965665.0         155    10.048
 X393-HVRP   Y    70000248.0          84    10.021
X401-FSMFD   Y    92500933.0          90    10.043
 X411-FSMD   Y    15848489.0          88    10.066
 X420-FSMD   Y    67711087.0          58    10.067
 X429-HVRP   Y    87635944.0          99    10.021
 X439-FSMF   Y   515039832.0          21    10.548
X449-FSMFD   Y   124300055.0          65    10.135
   X459-HD   Y    24263674.0          80    10.059
   X469-HD   Y   218545454.0          85    10.062
 X480-FSMD   Y    62098193.0          60    10.058
 X491-FSMF   Y   139887598.0          71    10.135
X502-FSMFD   Y   920420326.0          25    10.629
 X513-HVRP   Y    42623383.0          89    10.014
   X524-HD   Y   113242674.0          56    10.069
X536-FSMFD   Y   262387157.0          77    10.158
 X548-FSMF   Y  1114776234.0           9    10.192
 X561-FSMD   Y    33195750.0          62     10.14
 X573-HVRP   Y   107445194.0          53    10.148
 X586-FSMF   Y   960804618.0          13    10.783
 X599-FSMD   Y   140683919.0          70    10.212
   X613-HD   Y    62564170.0          56     10.13
 X627-HVRP   Y   108189727.0          25    10.125
X641-FSMFD   Y   194056042.0          53    10.125
   X655-HD   N           inf          15     10.69
 X670-FSMF   N           inf          27    10.037
 X685-FSMD   Y    93823628.0          54    10.103
 X701-HVRP   N           inf           7    10.896
X716-FSMFD   Y    75580358.0          44    10.041
X733-FSMFD   Y   849465258.0           8    10.287
 X749-FSMF   Y   222501932.0          45    10.304
 X766-FSMD   Y   115930661.0          29    10.117
   X783-HD   Y    86371809.0          34    10.218
 X801-HVRP   N           inf           6    10.642
 X819-FSMD   Y   162791596.0          30    10.195
   X837-HD   Y   213310410.0          14    10.704
 X856-HVRP   N           inf           4    11.646
 X876-FSMF   Y   340451028.0          30    10.438
X895-FSMFD   Y    94797404.0          44    10.075
X916-FSMFD   Y  2024469076.0           5    10.091
 X936-FSMD   Y   130411439.0          35    10.345
   X957-HD   N           inf          17    10.303
 X979-HVRP   Y   208099335.0          17    10.093

leonlan · 2024-03-28T14:33:42Z

OK, it turns out that there is at least one instance (X115-HVRP) that cannot even be feasibly solved. This is not a parsing issue, I also checked that the data of the original raw instance.

There may be more, but I don't want to do that manually. It should be very easy to check with an exact bin packing program.

I have messaged the original author of the paper Pessoa et al. (2018). Until then, this PR is on-hold.

Details about X115-HVRP

Instance: X115-HVRP.txt

These are the vehicle capacities:

... (small capacity vehicles)
131
131
131
131
131
322

These are the client demands:

0 (depot)
86
98
89
91
74
87
65
62
60
99
96
79
... (remaining clients all have small demand)

The presented 12 clients must be assigned to these vehicles with large capacities. There are more vehicles, but the capacity is too small. Summing up these values:

Total of vehicle capacity: 131+131+131+131+131+322 = 977
Total of client demand: 86+98+89+91+74+87+65+62+60+99+96+79 = 986

This can't even feasibly fit!

leonlan · 2024-03-28T19:49:32Z

OK, I manually checked all other instances and turns out X115-HVRP is the only infeasible instance. That's good news; I'll just run a benchmark without this instance.

There were still some issues with the weights. I have fixed this now and am running a new benchmark.

tests/test_PenaltyManager.py

N-Wouda

I need to think about the scaling issue a bit more. The increase in e.g. penalty term value and overall ballooning size of all values scares me a bit.

pyvrp/read.py

tests/data/MoreVehiclesThanSectionData.txt

pyvrp/read.py

pyvrp/PenaltyManager.py

leonlan · 2024-04-29T06:09:32Z

This PR is ready to be reviewed again. The open issue with HFVRP instances was how to deal with the scaling. The current solution is not to scale unit distance costs, instead we assume that the instance data is integral. Accordingly, the original HFVRP instances are modified to match this assumption.

N-Wouda · 2024-05-23T16:05:44Z

Oh, sorry, I still need to look at this. I'll try to do so later today!

N-Wouda

Looks OK to me, I only have one question about the rounding (and the earlier concern about the huge penalty terms up to 1 million, but that might be OK now that we have 64-bit integers).

N-Wouda · 2024-05-24T08:02:14Z

pyvrp/read.py

+        # Unit distance costs are not rounded: we assume here that the instance
+        # presents the costs in the desired precision!
+        unit_distance_costs = instance["vehicles_unit_distance_cost"]


That seems a bit odd. Why would we not round these? What goes wrong if we do?

leonlan added 2 commits March 25, 2024 19:38

Read HFVRP instances

53a6647

Simplify read

e8295c0

leonlan added 3 commits March 26, 2024 09:02

Check that HF data is complete

8254bb5

Do not correct vehicle name index by one

a66cdb8

Modify test a bit more

7a5717f

leonlan commented Mar 26, 2024

View reviewed changes

pyvrp/read.py Outdated Show resolved Hide resolved

Rename correct

5eb6185

leonlan marked this pull request as ready for review March 26, 2024 08:18

N-Wouda reviewed Mar 26, 2024

View reviewed changes

leonlan added 2 commits March 26, 2024 09:55

Address typos and test comment

ab0293e

Remark about CAPACITY and CAPACITY_SECTION

3e4231a

leonlan commented Mar 26, 2024

View reviewed changes

docs/source/dev/supported_vrplib_fields.rst Show resolved Hide resolved

leonlan added 5 commits March 29, 2024 08:17

Fix variable cost and unit distance cost renaming changes

069731d

Add HFVRP specific rounding

7d046c5

Increase penalty upper bound to 1M

f242373

Fix PM test

ae5b74e

Rephrase scaling comment

faa6217

N-Wouda self-requested a review March 29, 2024 16:14

leonlan commented Mar 29, 2024

View reviewed changes

tests/test_PenaltyManager.py Outdated Show resolved Hide resolved

Update text

e91c344

N-Wouda reviewed Mar 30, 2024

View reviewed changes

pyvrp/read.py Outdated Show resolved Hide resolved

tests/data/MoreVehiclesThanSectionData.txt Outdated Show resolved Hide resolved

pyvrp/read.py Outdated Show resolved Hide resolved

pyvrp/read.py Outdated Show resolved Hide resolved

pyvrp/read.py Outdated Show resolved Hide resolved

N-Wouda reviewed Mar 30, 2024

View reviewed changes

pyvrp/PenaltyManager.py Show resolved Hide resolved

leonlan added 3 commits March 30, 2024 14:14

Fix wrong unit_distance_cost name

4e32152

Address some review comments

a376f44

Fix unpacking

e38058d

leonlan added 5 commits March 31, 2024 17:36

Merge remote-tracking branch 'origin/main' into hfvrp-read

ffa1a6a

Do not round unit distance costs

2d964a5

Merge branch 'main' into hfvrp-read

970b25c

Merge branch 'main' into hfvrp-read

9313160

Rephrase

8e1058b

leonlan requested a review from N-Wouda April 29, 2024 06:06

N-Wouda reviewed May 24, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Read HFVRP instances #550

Read HFVRP instances #550

leonlan commented Mar 26, 2024 •

edited

codecov bot commented Mar 26, 2024 •

edited

leonlan commented Mar 26, 2024

leonlan commented Mar 27, 2024 •

edited

leonlan commented Mar 28, 2024 •

edited

leonlan commented Mar 28, 2024 •

edited

N-Wouda left a comment

leonlan commented Apr 29, 2024

N-Wouda commented May 23, 2024

N-Wouda left a comment •

edited

N-Wouda May 24, 2024

Read HFVRP instances #550

Are you sure you want to change the base?

Read HFVRP instances #550

Conversation

leonlan commented Mar 26, 2024 • edited

codecov bot commented Mar 26, 2024 • edited

Codecov Report

leonlan commented Mar 26, 2024

leonlan commented Mar 27, 2024 • edited

leonlan commented Mar 28, 2024 • edited

leonlan commented Mar 28, 2024 • edited

N-Wouda left a comment

Choose a reason for hiding this comment

leonlan commented Apr 29, 2024

N-Wouda commented May 23, 2024

N-Wouda left a comment • edited

Choose a reason for hiding this comment

N-Wouda May 24, 2024

Choose a reason for hiding this comment

leonlan commented Mar 26, 2024 •

edited

codecov bot commented Mar 26, 2024 •

edited

leonlan commented Mar 27, 2024 •

edited

leonlan commented Mar 28, 2024 •

edited

leonlan commented Mar 28, 2024 •

edited

N-Wouda left a comment •

edited