Safety-Governed Service Orchestration for Vehicular Micro-Edge Computing

The growing availability of sensing, computing, and communication capabilities in connected vehicles has enabled cooperative vehicular services that extend beyond traditional driving functions. Vehicular Edge Computing (VEC) has emerged as a promising paradigm for distributing perception and computation tasks across vehicles and edge infrastructure. However, existing orchestration mechanisms mostly optimize quality-of-service metrics and treat vehicles as generic resource providers, which largely neglects real-time driving safety. To address this issue, we propose in this paper VoXaaS (Vehicle-oriented eXecution as a Service), a safety-governed vehicular micro-edge service framework in which vehicles expose lightweight perception and computation micro-functions only when explicitly permitted by safety constraints. VoXaaS relies on Safety-Aware Volunteering (SAV), a continuous, risk-driven qualification mechanism that regulates service eligibility, contribution type, and contribution capacity. To do so, SAV relies on dynamic driving risk, resource headroom, and service providers’ reliability. In VoXaaS, service orchestration is performed by a Service Meta-Data Node using trust- and QoS-aware utility ranking. It is also performed by a local Gate Agent (GA) that enforces safety-first admission, resource arbitration, and execution control on each vehicle, with the aim to ensure that service optimization never overrides safety. Extensive simulation under realistic mobility and increasing task arrival rates shows that VoXaaS reduces selected-provider risk by more than 80% compared to QoS-only and trust-based baselines. It also eliminates unsafe volunteering, while preserving full admission and avoiding load shedding. Despite enforcing strict safety constraints, VoXaaS maintains the highest task success rate and up to 14% higher throughput than QoS-driven orchestration, with only moderate latency overhead. These results confirm that safety gains stem from risk-aware provider qualification rather than task rejection. They also confirm that explicit safety governance can coexist with scalable service performance. We argue that by tightly integrating orchestration, elastic trust regulation, and real-time driving risk, VoXaaS advances the state of the art toward dependable, safety-governed vehicular micro-edge ecosystems.