Skip to content
This repository has been archived by the owner on Jul 25, 2023. It is now read-only.

Latest commit

 

History

History
93 lines (81 loc) · 3.01 KB

pps-latest-result.md

File metadata and controls

93 lines (81 loc) · 3.01 KB
Raw

Parallel Push & Swap (PPS) results with v1.3 implementation

  • Because the PPS implementation (v1.0-v1.2) includes bugs, I retried PPS experiments using v1.3. This file shows the results.
  • As a brief conclusion, the results are not so different from the PIBT paper but the runtime is much faster due to the change of the underlying A* search. Note that PIBT runtime is also much improved.
  • If you are interested in the PIBT algorithm, please check a new implementation instead of this repo.
  • I still do not confirm the bug at least in the following instances, but if you find something wrong, let me know. I will address the problem depending on the situation. But remember, I have not used this repo for a long time.

8x8

  • timesteplimit=500
  • random starts/goals, 100 instances, seed: 0-99
agents success rate (%) path (ave.) runtime (ms)
10 100 8.0 0
15 100 10.5 0
20 100 13.7 1
25 100 17.5 2
30 98 21.7 3
40 98 36.4 12
50 92 67.2 43

lak105d

  • timesteplimit=1000
  • random starts/goals, 100 instances, seed: 0-99
agents success rate (%) path (ave.) runtime (ms)
10 100 21.4 0
25 100 27.2 3
50 100 39.8 16
75 100 56.8 57
100 98 79.6 131

arena

  • timesteplimit=1000
  • random starts/goals, 100 instances, seed: 0-99
agents success rate (%) path (ave.) runtime (ms)
10 100 32.6 1
25 100 33.9 3
50 100 36.5 8
100 100 41.3 24
200 100 52.7 92
300 100 66.3 243
400 99 81.8 548
500 98 100.9 1100

ost003d

  • timesteplimit=1500
  • random starts/goals, 100 instances, seed: 0-99
agents success rate (%) path (ave.) runtime (ms)
10 100 164.2 35
25 100 172.4 106
50 100 179.0 268
100 100 193.9 758
200 100 223.0 2512
300 100 248.3 5074
400 99 279.5 10159
500 100 312.1 16108

random-32-32-20

  • timesteplimit=1000
  • MAPF benchmark, random-scenario: 1-25
agents success rate (%) path (ave.) runtime (ms)
10 100 24.2 1
25 100 25.9 2
50 100 31.4 9
75 100 37.7 18
100 100 44.0 33
200 92 88.3 271
300 100 183.6 1819
400 76 388.7 7943

warehouse-10-20-10-2-1

  • timesteplimit=1000
  • MAPF benchmark, random-scenario: 1-25
agents success rate (%) path (ave.) runtime (ms)
10 100 68.1 1
25 100 86.3 8
50 100 99.5 24
75 100 107.4 40
100 100 111.7 65
200 100 123.0 277
300 100 140.9 752
400 100 169.5 1774
500 100 199.2 3619