Is cloud control is possible when facing real-time challenges in a safety-critical environment? This project provides first insights and data into cloud-based control by means of a feasibility study in the automotive domain. We present the Control as a Service (CaaS) concept, which investigates cloud-based control scheme for a model car with WLAN / CAN gateway. CaaS delivers successfully an architectural design and a proof-of-concept implementation for a simple cloud-based throttle limitation scenario, in which the throttle values requested by driver (remote control user) is dynamically regulated by a controller implemented as a virtual ECU (Electronic Control Unit) in the cloud. We found that the correct handling of time-varying network delay is one of the most relevant challenges. Therefore we started a simulation study in parallel to design controllers that are capable of coping with network imperfections. Preliminary simulation results indicate the potential of model predictive control to solve such control problem.
This project was carried out in collaboration with DENSO AUTOMOTIVE Deutschland GmbH. Some of the results have been published in the paper Control as a Service (CaaS): Cloud-based Software Architecture for Automotive Control Applications, presented at the Second International Workshop on the Swarm at the Edge of the Cloud, held in Seattle, Washington, in conjunction with CPSWeek 2015.
Download the presentation.