In Singapore, more than 40 percent of a building’s total energy consumption can be attributed to its chiller plant, according to ADSC Senior Research Scientist Zhenjie Zhang. By using new technology developed as a collaboration between ADSC and a Singapore air-conditioning company, Kaer, companies could save thousands of dollars per month on their energy bills.

A Kaer chiller plant

Poor efficiency in an HVAC system is often due to the excessive overhead and technical difficulty of manually tuning a chiller plant, which is the part of an air conditioning system that is responsible for keeping the air cool. ADSC and Illinois researchers are working with Kaer, who owns and operates numerous chiller plants across the island, to develop an advanced optimization and data management platform that can be easily used in any building without expert knowledge. By using the researchers’ machine learning algorithms, some plants have already realized savings of five to 20 percent.

The collaborative project brings together experts in the industry in computer science, including ADSC’s Zhang and University of Illinois Computer Science Professor Emeritus Marianne Winslett, Nanyang Technology University Computer Science Assistant Professor, and ADSC affiliate Xiaokui Xiao, and in mechanical engineering, including University of Illinois Mechanical Science and Engineering Professor Andrew Alleyne and Illinois graduate student Bryan Keating. The project is funded by a two-year, $1.2 million grant through Singapore’s Building Construction Authority, as part of the Singapore Green Buildings Innovation Cluster (GBIC).

“One of the things we’re excited about is that at Kaer we design and operate the chiller plants, so we know the problems and how expensive these systems are,” said Chai Kok Soon, director of Research and Development at Kaer. “By bringing together all these people with different backgrounds, we can hopefully use the data we’ve collected to help solve the problems.”

The goal is to use the chiller plant in a way that uses the least amount of energy possible while keeping the building at the proper temperature. Developing a solution is difficult because of the complexity of the plant system and is compounded by external factors such as varying weather patterns, building usage, and the fact that different buildings may have completely different characteristics, which doesn’t allow for easy transfer of technology.

Machine learning provides the theory to predict independent variables, such as power and cooling load, from dependent variables for chiller plants. However, the major challenge is to apply the theory to implement a real-time, autonomous learning system. The learning-based energy management system needs to perform automatic optimization and baselining to evaluate energy saving. Kaer developed a software platform to implement the theory with data visualization, optimization, real-time control and Internet of Things modules to do continuous training, testing, evaluation and learning of actual chiller plants.

By partnering with Kaer, the researchers can test their optimization algorithms on a live system, as well as use years’ worth of data that Kaer has collected from its plants — which are in places like shopping malls, schools, and businesses — to develop their methods.

“Working with ADSC and Kaer has been very valuable to us in obtaining real world data on complex systems operating in challenging practical environments,” Alleyne said. “It helped us refine our modeling and simulation tools. It also gave us good insight into the effectiveness of our algorithms and ways we could modify them to meet the real-world situations.”

According to Zhang, the chiller plant industry currently requires experts to manually alter chiller plants for optimal performance and most companies don’t have full-time engineering support.

“Unfortunately, even though these engineers can control the plants in an efficient way to save more energy, it’s very difficult to train these engineers,” Zhang said. “That’s why machine learning can be a good addition to the current technology.”

Zhang added that this is also a great time for this research because sensors have recently become an essential part of chiller plants, which allows researchers to collect and analyze the plant’s data, allowing them to understand the plant’s behaviors.

The researchers have found that the platform consistently outpeforms expert engineers when it comes to cost savings.

“Since September, we have achieved savings of four to nine percent or even more, compared to some of our most well-run chiller plants,” Kok Soon said.

Kok Soon added that they have seen five to 20 percent improvement on many chiller plants, depending on the expertise of the energy experts, control strategy, equipment conditions, and frequency of adjustments. This leads to round-the-clock savings for the companies, as well as freeing up engineers for more high-value work, such as diagnostic problem solving.

The researchers are working on a platform that will enable their work to use on a large scale, as well as identify, diagnose, and send alerts when something goes wrong with a sensor.

“We have a few plants that are using this technology now and we are hoping to make this technology available to the public soon, so many companies will be able to do optimization without human expertise,” Kok Soon said.

ADSC researchers are developing technology that can build virtual 3D replicas out of moving 2D images, even when the images contain similar textures and sizes.

A building reconstructed using ADSC-developed technology

ADSC Research Scientist Daniel Lin is building upon work conducted at the University of Washington, which leveraged images from Flickr.com to construct a virtual 3D replica of Rome using Structure-from-Motion (SfM), which is the process of inferring a 3D scene structure from moving 2D images. They used common points between the thousands of images to reconstruct the city’s popular sites and landmarks.

Their parallel distributed matching system worked great when recreating cities such as Venice and Dubrovnik, Croatia, because those cities have many buildings that are distinctive and are of different textures and sizes. But bundler-style systems have difficulty distinguishing the makeup of modern cities, which contain repetitive strutures and similar textured regions.

Lin, who is collaborating with Singapore University of Technology and Design Assistant Professor Sai-Kit Yeung, is working to build a system that can accommodate all images and scenes, even where there is little texture and lots of repetition, such as a row of skyscrapers in Singapore or the flat walls of a home.

“We are trying to stabilize the system, so that it can works most of the time, avoiding potentially frustrating failures,” Lin said.

The basis of Lin’s research is feature matching, where a computer can look at an image or scene and recreate it using features, or distinctive image regions, of that image that can be described with a unique descriptor. By comparing one image to another, the algorithm developed by Lin, called RepMatch, can match identical points in each image.

“By definition, there is only one correct match, but many ways to match incorrectly,” he said. “Correct matches tend to be clustered together and incorrect matches tend to be scattered. This helps us get the most consistent set of matches from each scene. Once we do that, we can use a geometrical growing tool to get even more matches.”

Some of the benefits to Lin’s algorithm is it allows computers to estimate the camera positions in the images more accurately, in addition to holding images together and avoiding breakage when an image turns corners or go outdoors.

While Lin’s algorithm has made it possible to recreate these 3D images of a variety of spaces, the feature matching is extremely slow when using large data sets. Lin has begun developing a newer and faster algorithm that still guarantees the same amount of accuracy.

It’s Lin’s hope that this technology could be used in many different areas, from furniture sales and real estate to a possible alternative to laser scanners.

“The main problem is that our system is a bit too slow, but if our results improve, we could see this being used in many different fields,” he said.

Urban noise is sometimes seen as merely a harmless annoyance for people who live near busy streets or train stations, but it can cause health problems such as sleep disturbance, hearing loss, hypertension, and heart-related diseases.

Arrays are mounted to the top of an electric vehicle to measure noise levels.

Researchers have therefore been seeking ways to understand and combat the rising levels of urban noise. A team at the Advanced Digital Sciences Center (ADSC) has developed a signal processing technique that measures urban noise through portable microphones secured to the top of a moving vehicle, enabling the creation of a wide-ranging map of noise pollution.

“With a comprehensive understanding of the levels and types of noise pollution in urban areas, we can then analyze this information to create well-designed soundscapes that can alleviate the bad effects of environmental noise on physical and mental human health,” said Cagdas Tuna, a postdoctoral researcher at ADSC, a University of Illinois research center in Singapore.

Current noise-monitoring techniques are built into microphones fixed to the ground—they only measure sound from that vantage point, making a city-wide noise map an incredibly expensive idea. However, with the team’s portable solution, sound can be measured in as many locations as possible in the neighborhoods travelled by the vehicle.

To gather acoustic signals, they mount a microphone arrangement on an electric vehicle—the quiet engine keeps it from interfering with other external sounds. While driving along, the sensors identify a variety of noises and can pinpoint the location of sounds in relation to the vehicle.

Advanced signal processing tools recover and generate the noise-sources into an acoustic map at multiple frequencies.

“We have developed several different acoustic imaging algorithms, based on the multiple-location measurement scheme, to generate 2D acoustic maps,” said Tuna, a University of Illinois alumnus in electrical and computer engineering. “The maps show the noise-levels and locations of dominant noises.”

The team has been testing the new technique on Singapore streets over the past year. Tuna, who presented this work at the 23rd International Congress on Sound & Vibration (ICSV23) in July 2016, will continue to develop this technique by collecting more measurements around noisy areas such as construction sites.

This acoustics team at ADSC working on the project includes CSL and ECE Professor and ADSC Director Doug Jones, Tuna, Shengkui Zhao, ADSC research scientist, and Thi Ngoc Tho Nguyen, ADSC Senior Software Engineer.

This material is based on research supported in part by the Singapore Ministry of National Development and National Research Foundation under L2 NIC Award No.: L2NICCFP1-2013-7 and in part by the research grant for the Human-Centered Cyber-Physical Systems Programme at the Advanced Digital Sciences Center from Singapore’s Agency for Science, Technology and Research (A*STAR). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Singapore Ministry of National Development, National Research Foundation and Singapore’s Agency for Science, Technology and Research (A*STAR).

Researchers at the Advanced Digital Sciences Center (ADSC) have developed a technology that allows facilities to monitor their electricity usage in finer granularity without making major modifications to a space’s electrical distribution panels. As a result, building owners could avoid the cost and disruption that would occur from rewiring or replacing components, particularly in older buildings.

Based on the technology, Senior Research Scientist Binbin Chen and Senior Research Engineer William Temple formed a Singapore-based spinoff company, Ampotech, last year, along with co-founder, Ziling Zhou, who was with the National University of Singapore before joining the team.

Ampotech sensors are installed above circuit breaker switches and monitor energy usage.

Ampotech is an energy monitoring and data analytics company that provides hardware and software enabling building owners to see where their energy is being used and to analyze that data. The basis of the product, AmpoSense, is a patent-pending, circuit-level energy monitoring system where users simply stick an array of sensors above circuit breaker switches to measure the magnetic field and determine energy usage of each individual branch. It is based on a technology Chen and his ADSC colleague, Sreejaya Viswanathan, designed and prototyped about three years ago. Since it licensed out the technology, Ampotech has significantly improved the readiness of the technology and also developed complimentary software—AmpoCloud—to complete their solution by allowing users to manage and track a building’s electricity usage in real-time via the cloud.

All buildings have a number of fuse boxes or electrical distribution panels where the incoming power supply is split into different circuits or branches. Each branch powers a different piece of equipment or space, such as lighting, wall outlets, a washing machine, or a refrigerator.

“Our solution gives people a way to monitor the energy use on each of those branches without having to replace any components or make any major changes to the panel,” Chen said. “It turns circuit breakers into smart meters.”

While there are existing products, such as CT-based solutions or DIN rail-mounted meters, that monitor energy being used in a building or floor, Ampotech provides users a way to see the end use. Additionally, current products are challenging and costly to install, while Ampotech’s technology provides the same data using a stick-on device.

“Older buildings don’t have a lot of energy monitoring infrastructure and the owners often don’t want to invest in making major changes,” Temple said. “Our system helps solve that problem since it’s so easy to install and receive data.”

Ampotech has been awarded the 2-Stage Innovation Grant (iGrant) from Singapore’s Building and Construction Authority. The grant allows the team to further develop the technology and bring it closer to a commercially-ready state. The team has completed the first phase of the project and is now working on late-stage development, running multiple pilot projects as they test the technology in field environments.

“This is a totally new product, so it’s not something people are familiar with or know already,” Chen said. “We spent a lot of time trying to find the right target market. When we first developed it, we knew the technology was innovative, but users won’t buy it just because it’s cool. They’ll buy it because it’s useful.”

Last summer, the researcher installed their technology on two panels in a Singapore-based electronics manufacturing company with two production lines. They focused on analyzing equipment energy usage and operating patterns and were able to identify opportunities to achieve 10 percent in energy savings during the two-month project.

“One advantage we were able to offer them is help in enforcing energy saving rules and policies,” Chen said. “In a factory, they have policies where a machine should only be used when another is in operation, for example. These policies can’t easily be enforced, but with our system in place, we can provide real-time monitoring of these rules and send notifications to a manager when a policy is being violated.”

The team is beginning another pilot project with a property developer in Singapore that will focus on office space energy usage and is geared toward educating the employees on ways they can help the office become more energy efficient.

“The scope of the project is a lot larger. We’re going from tens to hundreds of branches being monitored,” Temple said. “We’ll be monitoring energy usage of their lights, computers, microwave, refrigerator, and the like and communicating with the workers to help encourage energy efficiency.”

Ampotech plans to have a version commercially available by the end of 2016, as they’ve already had significant interest from potential partners and companies.

“It’s an exciting technology because electricity is such a basic need,” Temple said. “Homes, offices, factories, schools—they all use similar panels and wiring, so the opportunity is big. This system can change the way organizations and homeowners monitor their energy use to deliver substantial savings.”

We laugh, we cry, we grimace, we smile—our emotions are communicated through our expressions, and up until recently, humans were the only ones who could accurately interpret and analyze them. Now, through a newly developed technology by ADSC researchers, these expressions can be tracked, in real time, and measured on a spectrum of emotions.

The expressions of the individual are rated on a continuum based on arousal, valence, and intensity. The cartoon characters in the arousal-valence space are courtesy of Delft University of Technology. studiolab.ide.tudelft.nl/studiolab/pmri/

The team—composed of researchers from the Advanced Digital Sciences Center (ADSC), a University of Illinois research center in Singapore—built software that uses 49 points on the human face to put facial expressions on a spectrum of emotion, measured by arousal, positivity (valence), and intensity.

The team has already licensed this technology to Panasonic, an electronics company in Japan, and are finding other applications in areas as diverse as advertising, education, human resources, politics, and more.

Unlike other emotion-tracking software, this system doesn’t place expressions into predetermined categories, like happy, sad, disgusted, surprised, or neutral. Instead, they’re placed on a spectrum that more accurately encompasses the wide variety of human emotion.

“We found that other models try to classify emotions into predetermined categories, usually seven. But people in real life are more complicated than that,” said Stefan Winkler, a distinguished scientist at ADSC. “People could be happy, but they could be slightly, moderately, or very happy—variations in intensity. Or sometimes people exhibit compound emotions. They could be positively surprised or negatively surprised.”

The researchers used a model from the psychology domain that places emotions on a continuum, rather than predefined categories. The horizontal axis measures valence, which estimates pleasure and displeasure; the vertical axis tracks arousal, which calculates activation or deactivation. The intensity of the expressions—how raised the eyebrows are, how much the smile curves upward—determines the magnitude of arousal and valence.

By tracking 49 points, the software had enough measurements to capture facial expression, but ran into another challenge: how to interpret those points.

“We used a validated database of photos that had been previously rated in regards to perceived emotions and intensity of expression,” said Vassilios Vonikakis, an ADSC research scientist. “We then used this big data set—300,000 images—as training examples for machine learning. The computer taught itself what facial displacements were associated with certain emotions, learning to associate values of arousal and valence with the positions of particular points on the face.”

The software, which can be used with a standard video webcam and installed on any laptop, can also detect engagement.

“We can use the head pose to judge if the person is paying attention to the camera or not. If the person turns away, we can conclude they are less engaged with the screen,” said Siddhanta Chakrabarty, a software engineer at ADSC. “This could have applications in online education—determining if students are engaged with an online lecture.”

The team is looking at a host of other applications. Advertising agencies could measure the reactions of people when they see a billboard or watch a commercial to determine ad effectiveness. Human resources could record the expressions of job candidates during an interview and quantify the expressions of the potential employee. The ability to analyze the expressions of a crowd could prove useful during political speeches, or anywhere with a large crowd.

“Our approach of analyzing emotions on a spectrum—rather than using predefined categories of expressions—makes analyzing a crowd much easier,” said Vonikakis. “Our software gives more fine-grained results individually, and can also estimate the aggregate emotion of a crowd.”  

The group will continue to improve the accuracy of the system, and have already sold several licenses to commercial ventures, including Panasonic, to utilize the technology.

“ADSC’s emotion-tracking system yields good correlations in terms of arousal, valence, and intensity as it leverages big datasets for machine learning. That enables us to build more reliable systems,” said Kim Koon Chan, general manager of Singapore Technology Center of Panasonic Industrial Devices Singapore. “We expect the market to increase rapidly and welcome new customers to this field.”

The team was awarded a GAP grant from A*STAR to develop this from a research prototype to a commercial product.

“We’re excited to be using this technology in many different areas—it has potential to provide a much more robust way of describing and quantifying facial expressions than has previously been possible,” said Winkler.

Advanced image recognition research from the Advanced Digital Sciences Center (ADSC) has found an unusual testbed: a sushi buffet.

Software developed by ADSC recognizes and counts each piece of sushi as it rushes by.

As pieces of sushi rush along a conveyor belt in Sushi Express in Singapore, an image recognition system quickly identifies each sushi roll and tallies the number of pieces per category. It helps employees know when they are running low on smoked salmon rolls, for example, and need to add more to the conveyor.

This system makes the sushi identification process seamless and simple, thanks to complex algorithms that were created by a team of researchers at ADSC, a University of Illinois research center in Singapore. The technology uses object detection and recognition techniques such as machine learning and classification algorithms.

Yuduo Zheng “We were working with object detection and classification in the lab, and found that there was a need in industry that could also provide us with data to test our methods,” said Yuduo Zheng, a senior software engineer at ADSC. “We discovered that it can actually work well with sushi and provide useful statistics to our industry partner, Xjera Labs.”

Their system consists of two parts: detection and classification. To detect the sushi piece, the algorithm extracts what is called the local binary patterns (LBP) from the sushi—a type of visual descriptor in computer vision. The system then uses AdaBoost, short for adaptive boosting, which increases the algorithm strength and learning capabilities to better identify the correct piece of sushi.

They further improve their classification component with deep learning, which uses machine learning algorithms to extract data and make accurate predictions.

ADSC and Xjera Labs worked together to develop the system, and Xjera Labs has since licensed the technology and are involved in commercializing it for several chain sushi companies.

The ADSC team is continuing to improve the algorithms so the system will recognize new types of sushi, as well as seamlessly adapt to novel conditions and products of different brands and outlets.

“This technique could also be used for other conveyor-belt related companies and industries, such as logistics, by using it to check and count different types of objects on the conveyor belt,” said Zheng.

The ADSC team involved in this work included Zheng; Yongchao Wei, senior software engineer; Yue Xu, former engineer; and Lu Ding, software engineer.