MADDPG and P2P-VFRL for minimizing AoI in NTN network under CSI uncertainty

Citation Author(s):
Tarbiat Modares University(TMU)
Submitted by:
Maryam Ansarifard
Last updated:
Sat, 05/27/2023 - 02:11
Data Format:
0 ratings - Please login to submit your rating.


In this paper, we develop a hierarchical aerial computing framework composed of high altitude platform (HAP) and unmanned aerial vehicles (UAVs) to compute the fully offloaded tasks of terrestrial mobile users which are connected through an uplink non-orthogonal multiple access (UL-NOMA).
To better assess the freshness of information in computation-intensive applications the criterion of age of information (AoI) is considered. In particular, the problem is formulated to minimize the average AoI of users with elastic tasks, by adjusting UAVs’ trajectory and resource allocation on both UAVs and HAP, which is restricted by the channel state information (CSI) uncertainty and multiple resource constraints of UAVs and HAP. In order to solve this non-convex optimization problem, two methods of multi-agent deep deterministic policy gradient (MADDPG) and federated reinforcement learning (FRL) are proposed to design the UAVs’ trajectory and obtain channel, power, and CPU allocations. It is shown that task scheduling significantly reduces the average AoI. This improvement is more pronounced for larger task sizes. On one hand, it is shown that power allocation has a marginal effect on the average AoI compared to using full transmission power for all users. Compared with traditional transmission schemes, the simulation results show our scheduling scheme results in a substantial improvement in average AoI.


To run the code you need to install the following packages;

Numpy, torch, random, Scipy, os



Submitted by Abdelhamed ABDE... on Tue, 08/29/2023 - 06:01