BACO: A Backoff-Based Coordinated Task Offloading Scheme for Infrastructure-Free MEC

Abstract

This paper studies task offloading in infrastructure-free mobile edge computing (MEC), where end devices must autonomously discover candidate servers and make offloading decisions under dynamic connectivity and time-varying computing resources. However, the lack of centralized coordination incurs high overhead to acquire complete and timely server status information in device-dense environments, while independent user decisions conflict over high-performing servers. To address these challenges, we propose BACO, a backoff-based coordinated offloading scheme with three innovative designs. First, the proposed BACO framework coordinates user requests and server responses via corresponding distributed backoff processes over the shared broadcast channel. Specifically, long-waiting tasks and resource-rich servers are prioritized in backoff timer assignment, while user decision conflicts are avoided. A candidate server table at each user supports further multi-user coordination by reusing revealed server status information. Second, a salient-driven mapping algorithm is proposed to allow contending users and candidate servers to independently set backoff timers, mitigating channel collisions caused by excessive server responses by super-linearly amplifying priority differences. Third, the BACO backoff process is theoretically analyzed in terms of optimal-server revelation probability, collision probability, and backoff timer overhead, which supports environment-adaptive optimization of key backoff-related parameters under time-varying MEC environments. Extensive simulations and testbed experiments demonstrate that BACO consistently outperforms state-of-the-art baselines, reducing multi-user average offloading latency by up to 52.77% while maintaining favorable scalability and robustness in dynamic environments.