Open astarforbae opened 3 weeks ago
The actual value of pds_a
requires a knowledge of the intervals between the arrival times of different tasks.
For example, if the current AoI is 30
and the interval between the HOL task and the next task is 10
slots, then pds_a=20
if b=1
. However, we do not record such interval information in our system state so we cannot obtain the actual values of pds_a
in this way.
As a workaround, we minus the expected reduction of AoI instead of the actual reduction of AoI. In our original simulation setting, the task arrival rate is 0.2
so the expected interval between tasks is 5
, and this is how the update formula of pds_a
comes. However, we modified the arrival rate to 0.3
at a later time and forgot to modify the update formula accordingly. But based on our experience, the DDPG can learn and adapt to this discrepancy so the influence on the final results should be limited.
Got it! I really appreciate your help!
Hi, i would like to know why the AoI of PDS is calculated by a - 5 * b. I'm really confused with it.