Minimal sparse observability of complex networks: Application to MPSoC sensor placement and run-time thermal estimation & tracking

A Survey of Prediction and Classification Techniques in Multicore Processor Systems

“…We found that the Kalman filtering based predictions are very accurate and allow the proposed energy reduction heuristic to provide consistent energy savings under a given performance constraint for all benchmarks that we investigated. Also in the context of high performance processors, the authors of [18] proposed a sparse Kalman filter to estimate the states of a dynamical network system. They then applied their solution to the thermal model network of many-core processors to solve the problem of finding the minimum number of in-situ sensors that can be used for both thermal profile estimation and tracking of hotspots in dynamic thermal management solutions.…”

FPGA emulation and prototyping of a cyberphysical-system-on-chip (CPSoC)

“…Equipped with advances in machine learning, distributed control system design, and sensors (that have wide application in CPS), CPSoC brings many such attributes directly on-chip for an intelligent, self-aware, autonomic computing fabric. Specialized features such as increasing customization, heterogeneity, adaptation and agility, virtual sensing [18], [19], synergistic crosslayer cooperation and actuator fusion [8] are some key attributes for CPSoC computing platforms.…”

“…The CPSoC simulation platform also introduces a new component VAR (Vulnerability, Aging, and Reliability) by integrating Hotspot[34],[35]. VAR provides a prediction model for system vulnerability ( e.g., hotspot tracker, power bug tracker[19], malicious attack tracker etc. ), aging and threshold voltage shift due to NBTI, PBTI, HCI, EM and other failure modes[36], and component and system level reliably, availability, and mean time to failure (MTTF) of each component and cores.…”

On Distributed Pinning Observers for a Network of Dynamical Systems

“…This is useful in certain engineering applications where sensing is 'costly'. For example in multi-core processors, the build up of temperature has an adverse affect on power and reliability [10]. Monitoring the temperature distribution is important (and a precursor to developing control systems to dissipate the unwanted heat).…”

“…Incorporating temperature sensors encroaches on the available silicon area on the chips, and therefore comes at a high 'cost' in terms of space utilization. Consequently in such systems there is a clear trade-off between the requirement for establishing an accurate estimate of the temperature distribution while maintaining a minimal footprint in terms of the 'realestate' costs of deploying physical sensors [10]. Of course this paradigm is not applicable to all large scale problem formulations: for example in certain engineering systems such as wireless networks the most significant 'cost' is associated with communication rather than sensing per se.…”