Skip to content

Latest commit

 

History

History

data

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
action-steering
action-steering
 
 
motivation_main-results
motivation_main-results
 
 
offline-training
offline-training
 
 
README.md
README.md
 
 

README.md

Experiments

Organization

The folder contains 3 sub-folders:

  • offline-training/ contains the offline datasets and agents' weights (both agents eMBB - HT in the paper - and URLLC - LL in the paper) that were kindly provided by the authors of

    [PBO+22] M. Polese, L. Bonati, S. D’Oro, S. Basagni, and T. Melodia, “ColO-RAN: Developing Machine Learning-based xApps for Open RAN Closed-loop Control on Programmable Experimental Platforms,” IEEE Transactions on Mobile Computing, July 2022.

  • motivation_main-results/ contains the experiments for Section 3, Section 6.2 and Appendix C of our paper.
  • action-steering/ contains the experiments for the Section 6.3 and Appendix D of our paper.

Agents, Slices and Traffic Patterns

We run experiments for three different slices similarly to [PBO+22], namely eMBB, MTC, URLLC with two different agents, denoted as eMBB - HT in the paper - and URLLC - LL. At the time of resource allocation, the agent embb gives preference to the slice eMBB while the agent urllc gives preference to the slice URLLC. We also use two different traffic patterns, hereafter we summarize their properties, that are used to configure MGEN.

Traffic pattern 1 (trf1):

  • eMBB: 5,595,UDP,PERIODIC,[357.15 1400]
  • MTC: 5,595,UDP,POISSON,[44.64 125]
  • URLLC: 5,595,UDP,POISSON,[89.29 125]

Traffic pattern 2 (trf2):

  • eMBB: 5,3595,UDP,PERIODIC,[357.15 700]
  • MTC: 5,3595,UDP,POISSON,[133.92 125]
  • URLLC: 5,3595,UDP,POISSON,[178.58 125]

Detailed Configuration

Experiments for the motivation_main-results/

All the experiments run on Colosseum/SCOPE for 20 minutes.

embb-trf1:

UE Exp Number eMBB UE MTC UE URLLC UE
6 1 2 2 2
5 2 2 1 2
4 3 1 1 2
3 4 1 1 1
2 5 1 0 1
1 6 0 0 1
1 7 1 0 0
1 8 0 1 0

embb-trf2:

UE Exp Number eMBB UE MTC UE URLLC UE
6 9 2 2 2
5 10 2 1 2
4 11 1 1 2
3 12 1 1 1
2 13 1 0 1
1 14 0 0 1
1 15 1 0 0
1 16 0 1 0

urllc-trf1

UE Exp Number eMBB UE MTC UE URLLC UE
6 27 2 2 2
5 28 2 1 2
4 29 1 1 2
3 30 1 1 1
2 31 1 0 1
1 32 0 0 1
1 33 1 0 0
1 34 0 1 0

urllc-trf2

UE Exp Number eMBB UE MTC UE URLLC UE
6 35 2 2 2
5 36 2 1 2
4 37 1 1 2
3 38 1 1 1
2 39 1 0 1
1 40 0 0 1
1 41 1 0 0
1 42 0 1 0

Experiments for the action-steering/ results

These results are obtained processing experiments that are slightly different than the above ones. First, the number of users changes during the course of the experiment with a drop from 6 (2 users per slice) to 5 to emulate changing conditions. Then, the experiments follow the next workflow

  1. network runs for a specific amount of time (10 minutes) - phase-a
  2. 1 eMBB user drops at minute 5 of phase-a
  3. an online training phase takes place (10-15 minutes) to make the corresponding agent aware of the changes; in this phase the agent explores states, hence could take non-optimal decisions - phase-b
  4. the network resumes usual operations (5-10 minutes) - phase-c; during this phase we implement the action replacement (AR in the paper, see Section 5.2) strategies.

We test two differnt observation windows (O in the paper):

  • 10, with experiment configuration (Note: the results in the paper are derived using this configuration):
    • phase-a: 10 minutes,
    • phase-b 15 minutes,
    • phase-c 5 minutes.
  • 20: with experiment configuration:
    • phase-a: 10 minutes,
    • phase-b 10 minutes,
    • phase-c 10 minutes.