.png){kind=icon}
Return
PLUTO: Platooning through Uncertainty-aware Task Offloading
.png){kind=icon}
PRE
AI
DOI：10.1016/j.future.2026.108613.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The growing adoption of connected and autonomous vehicles creates significant challenges in managing tasks efficiently in vehicular platooning systems under uncertain conditions, such as changing traffic, road obstacles, communication instability, and adverse weather. Although platooning improves fuel efficiency, traffic flow, and road safety, it also requires reliable real-time task management in resource-constrained computing environments. Existing studies mainly focus on delay reduction or resource allocation, while limited attention has been given to uncertainty-aware task offloading under dynamic platooning conditions. To address this problem, we propose PLatooning through Uncertainty-aware Task Offloading (PLUTO), a hybrid framework that combines a three-tier edge–fog–cloud architecture with adaptive task management. The framework integrates Long Short-Term Memory (LSTM) for workload prediction, Hybrid Bayesian Neural Networks (HBNN) for uncertainty estimation, and Double Deep Q-Networks (DDQN) for adaptive offloading decisions. Tasks are categorized into latency-intolerant, storage-critical, and compute-intensive classes to support efficient workload distribution and resource utilization across distributed servers. The proposed framework is evaluated through simulation and real-world testbed experiments under varying workloads. Experimental results show 97% offloading efficiency and an 82% real-world task success rate. The framework improves workload balancing, resource utilization, and fuel efficiency while maintaining stable task execution under uncertain conditions. These results demonstrate the suitability of PLUTO for real-time task management in resource-constrained vehicular platooning environments.
Keywords:
PLUTO
vehicular platooning
task offloading
uncertainty-aware
edge–fog–cloud architecture
Journal
F
IF:
0
Papers:
642
Citations:
0
Organization
Cited Papers
No cited papers available