Oranits: Mission Assignment and Task Offloading in Open RAN-based ITS using Metaheuristic and Deep Reinforcement Learning

Researchers introduce Oranits, a system for optimizing mission assignment and task offloading in Open RAN-based autonomous vehicle networks using metaheuristic algorithms and deep reinforcement learning. The proposed MA-DDQN framework achieves 11% improvement in mission completions and 12.5% improvement in overall benefit compared to baseline methods, advancing edge computing efficiency in intelligent transportation systems.

Oranits addresses a critical optimization challenge in autonomous vehicle networks by tackling mission assignment and task offloading decisions that traditional approaches handle inefficiently. The research recognizes that real-world mission dependencies and offloading costs create complex interdependencies requiring sophisticated algorithmic solutions rather than oversimplified models. This work emerges from the broader convergence of edge computing, vehicular networks, and AI optimization—domains increasingly vital as autonomous vehicle deployments scale globally.

The paper presents two complementary approaches: CGG-ARO provides a metaheuristic baseline achieving 7.1-7.7% performance improvements, while MA-DDQN demonstrates superior results through multi-agent coordination and adaptive learning mechanisms. The 11% jump in mission completions using MA-DDQN suggests deep reinforcement learning's capacity to handle dynamic, time-sensitive environments where static optimization proves inadequate. This methodological advancement matters because transportation systems require real-time decision-making under uncertainty—precisely where DRL excels over traditional approaches.

For the broader industry, Oranits contributes technical infrastructure supporting autonomous vehicle deployment at scale. Improved task offloading efficiency directly translates to reduced latency, lower computational costs, and enhanced vehicle safety—critical factors determining commercial viability of autonomous fleets. As 5G and Open RAN networks expand globally, systems that efficiently distribute computing loads between vehicles and edge servers become competitive differentiators for telecom operators and vehicle manufacturers.

The research trajectory points toward increasingly sophisticated multi-agent coordination frameworks. Future developments may incorporate federated learning, network slicing optimization, and heterogeneous compute resource management. This work establishes benchmarks against which next-generation systems will be measured.

• →MA-DDQN framework achieves 11% improvement in mission completion rates through multi-agent deep reinforcement learning coordination
• →Oranits explicitly models mission dependencies and offloading costs, addressing limitations in prior optimization approaches for autonomous vehicle networks
• →Metaheuristic CGG-ARO baseline delivers 7.1% mission completion improvements, establishing performance benchmarks for edge computing systems
• →Real-time adaptive learning capabilities position DRL-based solutions as superior to static algorithms in dynamic transportation environments
• →Efficient edge task offloading optimization directly reduces latency and computational costs critical for autonomous vehicle safety and scalability