Simulation code of our paper in IEEE Transactions on Mobile Computing: ''Towards Decentralized Task Offloading and Resource Allocation in User-Centric MEC''. This is the extended version of our previous paper ``When the User-Centric Network Meets Mobile Edge Computing: Challenges and Optimization''.
Abstract: In the traditional cellular-based mobile edge computing (MEC), users at the edge of the cell are prone to suffer severe inter-cell interference and signal attenuation, leading to low throughput even transmission interruptions. Such edge effect severely obstructs offloading of tasks to MEC servers. To address this issue, we propose user-centric mobile edge computing (UCMEC), a novel MEC architecture integrating user-centric transmission, which can ensure high throughput and reliable communication for task offloading. Then, we formulate an long-term delay minimization problem by jointly optimizing task offloading, power allocation, and computing resource allocation in UCMEC. To solve the intractable problem, we propose two decentralized joint optimization schemes based on multi-agent deep reinforcement learning (MADRL) and convex optimization, which consider both cooperation and non-cooperation among network nodes. Simulation results demonstrate that the proposed schemes in UCMEC can significantly improve the uplink transmission rate by at least 176.99% and reduce the long-term average total delay by at least 16.36% compared to traditional cellular-based MEC.
We have designed a large number of portable multi-agent deep reinforcement learning (MADRL) environments of UCMEC to verify different algorithms. Specifically, IPPO and MAPPO algorithms are modified based on light-mappo. Other MADRL algorithms like IQL and MADDPG are simulated based on epymarl. Thanks very much for the contributions of the authors of these repositories. Please visit the homepages of these repositories to view specific usage tutorials.
Simply download the code, create a Conda environment, and then run the code, adding packages as needed.
You can quickly design your multi-agent environment using the following template. Please refer to /envs/env_core.py file.
import numpy as np
class EnvCore(object):
"""
# Environment Agent
"""
def __init__(self):
self.agent_num = 2 # set the number of agents(aircrafts), here set to two
self.obs_dim = 14 # set the observation dimension of agents
self.action_dim = 5 # set the action dimension of agents, here set to a five-dimensional
def reset(self):
"""
# When self.agent_num is set to 2 agents, the return value is a list, and each list contains observation data of shape = (self.obs_dim,)
"""
sub_agent_obs = []
for i in range(self.agent_num):
sub_obs = np.random.random(size=(14, ))
sub_agent_obs.append(sub_obs)
return sub_agent_obs
def step(self, actions):
"""
# When self.agent_num is set to 2 agents, the input of actions is a two-dimensional list, and each list contains action data of shape = (self.action_dim,).
# By default, the input is a list containing two elements, because the action dimension is 5, so each element has a shape of (5,)
"""
sub_agent_obs = []
sub_agent_reward = []
sub_agent_done = []
sub_agent_info = []
for i in range(self.agent_num):
sub_agent_obs.append(np.random.random(size=(14,)))
sub_agent_reward.append([np.random.rand()])
sub_agent_done.append(False)
sub_agent_info.append({})
return [sub_agent_obs, sub_agent_reward, sub_agent_done, sub_agent_info]Just write this part of the code, and you can seamlessly connect with MAPPO. After env_core.py, two files, env_discrete.py and env_continuous.py, were separately extracted to encapsulate the action space and discrete action space. In elif self.continuous_action: in algorithms/utils/act.py, this judgment logic is also used to handle continuous action spaces. The # TODO here in runner/shared/env_runner.py is also used to handle continuous action spaces.
In the train.py file, choose to comment out continuous environment or discrete environment to switch the demo environment.