Smartphones to the Rescue? Recovering Autonomous Vehicles Without Wheels or Pedals

This paper explores a fallback control system that enables nearby humans, such as on-site responders, to manually operate to recover an autonomous vehicle (AV) using only a smartphone. As AV manufacturers increasingly remove traditional driver interfaces, manual intervention will soon become infeasible, leaving remote driving as the primary recovery mechanism. However, remote driving suffers from high latency, bandwidth demands, and degraded situational awareness, limiting its effectiveness in time-critical or failure scenarios. As an alternative, this paper proposes a smartphone-based and edge-assisted scheme called SAVRE (Smartphone to the Autonomous Vehicle’s REscue). It addresses the manual intervention challenges by assisting the human operator inside the vehicle with edge computing to minimize control loop latency. Inspired by fly-by-wire systems, SAVRE integrates lane-keeping assistance and input filtering to ensure safe and intuitive operation. The system is implemented and evaluated using the CARLA simulator under varying network conditions. Experimental results show that SAVRE significantly outperforms remote driving in responsiveness, lane-keeping stability, and tolerance to delay and jitter, demonstrating its viability as a potential fallback solution for AV failure recovery.