Showing papers presented at "International Conference on Event-based Control, Communication, and Signal Processing in 2018"

Internal Server State Estimation Using Event-based Particle Filtering

Abstract: Closed-loop control of cloud resources requires there to be measurements readily available from the process in order to use the feedback mechanism to form a control law. If utilizing state-feedback control, sought states might be unfeasible or impossible to measure in real applications; instead they must be estimated. However, running the estimators in real time for all measurements will require a lot of computational overhead. Further, if the observer and process are disjoint, sending all measurements will put extra strain on the network.In this work-in-progress paper, we propose an event-based particle filter approach to capture the internal dynamics of a server with CPU-intensive workload whilst minimizing the required computation or inter-system network strain. Preliminary results show some promise as it outperforms estimators derived from analytic expression for stationary systems in service rate estimation over number of samples used for a simulation experiment. Further we show that for the same simulation, an event-based sampling strategy outperforms periodic sampling.

PRED18: Dataset and Further Experiments with DAVIS Event Camera in Predator-Prey Robot Chasing

Abstract: Machine vision systems using convolutional neural networks (CNNs) for robotic applications are increasingly being developed. Conventional vision CNNs are driven by camera frames at constant sample rate, thus achieving a fixed latency and power consumption tradeoff. This paper describes further work on the first experiments of a closed-loop robotic system integrating a CNN together with a Dynamic and Active Pixel Vision Sensor (DAVIS) in a predator/prey scenario. The DAVIS, mounted on the predator Summit XL robot, produces frames at a fixed 15 Hz frame-rate and Dynamic Vision Sensor (DVS) histograms containing 5k ON and OFF events at a variable frame-rate ranging from 15-500 Hz depending on the robot speeds. In contrast to conventional frame-based systems, the latency and processing cost depends on the rate of change of the image. The CNN is trained offline on the 1.25h labeled dataset to recognize the position and size of the prey robot, in the field of view of the predator. During inference, combining the ten output classes of the CNN allows extracting the analog position vector of the prey relative to the predator with a mean 8.7% error in angular estimation. The system is compatible with conventional deep learning technology, but achieves a variable latency-power tradeoff that adapts automatically to the dynamics. Finally, investigations on the robustness of the algorithm, a human performance comparison and a deconvolution analysis are also explored.

Event-Based Control and Estimation of Server Systems

Abstract: Feedback control is increasingly being applied in server systems to make them more robust and efficient. This includes managing quality of service, minimizing power consumption, and adapting to varying workloads. Successful adaptation and control in turn relies on accurate tracking of workload variations and timely detection of changes in the computing infrastructure. Given that server systems are inherently event based, it is natural to consider event-based control and estimation schemes for them. As a prototypical problem, we study a single server system with time-varying arrival rate and derive optimal switching rules for the service rate. The goal is to keep the response time within bounds while minimizing the energy consumption of the server. We also design an event-based estimator of the server states using a particle filter approach. Finally, we outline some research challenges related to event-based control and information fusion in server systems. (Less)