In the face of increasingly computing-intensive and delay-sensitive vehicular applications, vehicular edge computing (VEC) has become a promising computing paradigm by deploying computing resources at the edge. This paper investigates an age of information (AoI)-aware vehicular edge offloading problem by dynamically adjusting the edge offloading ratio and selecting the VEC server, taking into account the computing energy efficiency (CEE). To adapt to the time-varying network topology of VEC, we propose a multi-agent cooperative edge offloading solution relying on actor-attention-critic framework, where each vehicular user equipment (VUE) employs an attention mechanism to regulate its attention to other VUEs, facilitating selective focus on important information to enhance policy learning. The simulation results show that the proposed solution can achieve a more compelling trade-off between AoI and CEE compared with the baseline solutions.
AoI-Aware Energy-Efficient Vehicular Edge Computing Using Multi-Agent Reinforcement Learning with Actor-Attention-Critic
24.06.2024
983823 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Multi-agent deep reinforcement learning with actor-attention-critic for traffic light control
SAGE Publications | 2024
|Europäisches Patentamt | 2020
|Actor-Critic Reinforcement Learning for Control With Stability Guarantee
BASE | 2020
|