Presenter Information

Yuewen ZhengFollow

Degree Name

Master of Computer Science and Systems (MCSS)

Department

Institute of Technology

Streaming Media

Location

Carwein Auditorium (KEY 102), UW Tacoma

Event Website

http://guides.lib.uw.edu/tactalks

Start Date

19-5-2016 6:15 PM

End Date

19-5-2016 6:20 PM

Abstract

Our world is being inundated with sensors and sensing devices. We have devices that can monitor environmental conditions, traffic flows, building control systems, and even record and transmit live video feeds over the internet to large data stores that can be analyzed using traditional Big Data processing frameworks. However, in many cases we could potentially use that streaming sensor data immediately if we had the capability to aggregate, analyze, and use those streams in real time.

We are currently building a Real-­Time Streaming Analytical System. We are leveraging several open source platforms, including Apache Kafka, Flink, Avro, and Nifi, to build a complete end-­to-­end system that use hierarchical, online machine learning models to predict and control systems at multiple levels of granularity. Our system is completely distributed and capable of scaling to handle thousands of sensors for different use cases with fault tolerance features to ensure 24-­7 operation.

We are currently deploying and evaluating using data streams from embedded platforms and sensors, such as Intel Edisons, Raspberry Pis, and Particle Photons in the Cherry Parkes building of UW-­Tacoma to collect temperature, humidity, CO2 levels, photoelectric sensor values, and video streams. These sensors provide data that drives our server prediction system with simulated control that demonstrates how we could improve the efficiency of lighting and environmental control systems to improve energy efficiency as well as provide improved visibility into building use for security.

Our ultimate goal is to enable UWT buildings to become smart, self-­learning buildings, capable of controlling themselves by adjusting light and temperature to save energy, balancing fresh air proportion to avoid disease and improve study environment, and providing alerts during accidents and emergencies such as fire, break-­ins, etc. In the future we intend to expand the scope of our system to different use cases as agriculture, industry and field-­work research.

Share

COinS
 
May 19th, 6:15 PM May 19th, 6:20 PM

Building and Evaluating a Real-Time Streaming Analytical System for the Internet of Things (IoT)

Carwein Auditorium (KEY 102), UW Tacoma

Our world is being inundated with sensors and sensing devices. We have devices that can monitor environmental conditions, traffic flows, building control systems, and even record and transmit live video feeds over the internet to large data stores that can be analyzed using traditional Big Data processing frameworks. However, in many cases we could potentially use that streaming sensor data immediately if we had the capability to aggregate, analyze, and use those streams in real time.

We are currently building a Real-­Time Streaming Analytical System. We are leveraging several open source platforms, including Apache Kafka, Flink, Avro, and Nifi, to build a complete end-­to-­end system that use hierarchical, online machine learning models to predict and control systems at multiple levels of granularity. Our system is completely distributed and capable of scaling to handle thousands of sensors for different use cases with fault tolerance features to ensure 24-­7 operation.

We are currently deploying and evaluating using data streams from embedded platforms and sensors, such as Intel Edisons, Raspberry Pis, and Particle Photons in the Cherry Parkes building of UW-­Tacoma to collect temperature, humidity, CO2 levels, photoelectric sensor values, and video streams. These sensors provide data that drives our server prediction system with simulated control that demonstrates how we could improve the efficiency of lighting and environmental control systems to improve energy efficiency as well as provide improved visibility into building use for security.

Our ultimate goal is to enable UWT buildings to become smart, self-­learning buildings, capable of controlling themselves by adjusting light and temperature to save energy, balancing fresh air proportion to avoid disease and improve study environment, and providing alerts during accidents and emergencies such as fire, break-­ins, etc. In the future we intend to expand the scope of our system to different use cases as agriculture, industry and field-­work research.

http://digitalcommons.tacoma.uw.edu/tactalks/2016/spring/7