CFCO: A Social-Welfare-Maximizing Collaborative Coalition Formation Method for Task Offloading in Vehicular Edge Computing
Year: 2026
Type: preprint
Abstract: Vehicular edge computing (VEC) provides computation resources close to data sources for autonomous driving, cooperative perception, and intelligent vehicular services. However, limited on-board computing resources, fluctuating V2I reliabil ity, and shared communication and computation resources at edge servers make it difficult for a single local-computing or MEC-offloading policy to achieve consis tently high system utility. To address this problem, this paper proposes CFCO, a Collaborative Coalition Formation for Offloading method that maximizes social welfare. CFCO unifies local computing, D2D collaborative computing, and MEC offloading as three execution modes. It models cooperation among non-MEC vehicles as a coalition formation process and alternates coalition merge/split operations, MEC-mode refinement, and local vehicle-migration actions under a strict social-welfare improvement rule. Unlike policies that optimize only delay or energy, CFCO jointly captures early-completion benefit, energy cost, and delay penalty, thereby reflecting heterogeneous task preferences across vehicles. In this revised study, we evaluate CFCO on real-world vehicle trajectories from the Next Generation Simulation (NGSIM) dataset and report statistically robust results averaged over ten independent random seeds. In a representative one-dimensional NGSIM snapshot setting with N = 60 vehicles and V2I availability probabil ity 0.7, CFCO achieves a social welfare of 481.277, an average task delay of 0.225 s, a task success ratio of 0.998, and a Jain fairness index of 0.716. Com pared with the second-best BR-LS baseline, CFCO improves social welfare by 0.2% while reducing average delay by 1.9%, and runs more than twelve times 1 faster. Compared with MEC-only, CFCO improves social welfare by 71.0% and reduces average delay by 50.4%. The final CFCO decision keeps 20.5% of vehi cles in local computing, offloads 22.5% to MEC, and assigns 57.0% to D2D collaborative coalitions. A small-scale exact optimality-gap study further shows that CFCO is within 0.03% of the optimal ILP solution. Parameter sweeps fur ther show that CFCO remains superior under varying local CPU capacity, D2D bandwidth, MEC capacity, vehicle density, and V2I reliability.
Source: Research Square
Institution: Aba Teachers University
Language: English
Topic: IoT and Edge/Fog Computing
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204183793