Cyber-Physical Systems (CPS) have emerged as a unifying concept for systems whose computational aspects are tightly integrated with the physical world. CPSs have often a strong focus on resource-efficiency, i.e., power efficiency and thermal constraints are important. The close interaction with the physical environment and humans also lead to a high level of uncertainty, i.e. it is difficult to specify how the system will be used, what the requirements are, and what the load will be. This has consequences for the computational parts of CPS. Static designs are unrealistic. Instead, the systems must dynamically react to changing conditions. They need to take action based on on-line measurements of performance, resource consumption etc., i.e., they must be based on feedback.
The objective is to develop a unified approach to feedback computing that covers: embedded systems, massively parallel manycores, desktop systems, and distributed systems. The aim of the control is either to improve performance, e.g., reduce latency or increase throughput, or to reduce resource consumption. The project emphasizes the development of generic techniques that can be employed within several or all of the four areas above.
The project is divided into three parts:
- Feedback computing for the distributed cloud
(Johan Eker, Victor Millnert, Enrico Bini)
- Feedback computing for distributed camera systems
(Martina Maggio, Gautham Nayak Seetanadi, Karl-Erik Årzén)
- Feedback Computing for manycores
- Gautham Nayak Seetanadi, Martina Maggio, Karl-Erik Årzén, Luis Almeida, and Javier Camara
Event-Driven Bandwidth Allocation with Formal Guarantees for Camera Networks
IEEE Real-Time Systems Symposium (RTSS), December 2017 [doi]
- Gautham Nayak Seetanadi, Luis Oliveira, Luis Almeida, Karl-Erik Arzén, and Martina Maggio
Game-theoretic network bandwidth distribution for self-adaptive cameras
SIGBED Rev. 15, 3, August 2018, 31-36 [doi]