Programmable Data Plane as a Real-Time Feedback Controller for Cyber-Physical Systems

Nayak et al. explored programmable data planes in automotive networks, but only scratched the surface of using switches for control logic. This idea aims to embed feedback controllers (e.g., PID, state machines) directly into the data plane, allowing the network itself to perform real-time actuation decisions for robotic, UAV, or automotive applications (see also Cesen & Rothenberg for in-network robotics offloading). Unlike prior work that pushes sensor data to the edge/cloud, here the network can intercept, analyze, and respond to sensor/actuator events at line rate. For example, a programmable switch could mediate emergency stop signals, enforce real-time constraints, or aggregate sensor data for distributed control. This could yield significant reductions in control loop latency and jitter, transforming networked CPS reliability and safety. The research would explore programming abstractions, formal verification, and hardware feasibility for such in-network controllers.

References:

1. Offloading Robotic and UAV Applications to the Network Using Programmable Data Planes. F. R. Cesen, Christian Esteve Rothenberg (2023). Conference on Network Function Virtualization and Software Defined Network.
2. Reimagining Automotive Service-Oriented Communication: A Case Study on Programmable Data Planes. N. Nayak, Uthra Ambalavanan, J. M. Thampan, Dennis Grewe, Marco Wagner, S. Schildt, J. Ott (2023). IEEE Vehicular Technology Magazine.