2019 Smart City Summit & Expo Mar. 26~ Mar. 29

AIoT Invigorates Smart City

As artificial intelligence technology develops, the applications begin to play a key role in government delivery of citizen services. Smart cities are entering to a new phase of Smart City 2.0 by integrating AI as well as IoT technologies into services.

Smart City Summit & Expo Official Website

Program March 25–29, 2019
Exhibits  March 26–29, 2019  (March 26th 9:00-18:00; March 27th-29th 10:00-18:00)

VENUE : Nangang Exhibition Center, Hall 1 (No.1, Jingmao 2nd Rd., Nangang District, Taipei City)


Floor Plans

NCHC booths 

Welcome to Smart City Expo at NCHC booths J405 & I514 during March 26–29, 2019 at Nangang Exhibition Center, Hall 1.

We look forward to seeing you at Smart City Expo!

Application of AIoT on structures/wind turbine health monitoring

Contacts: Dr. Su 蘇威智
Email: 1203017@narlabs.org.tw  

  • Integration of sensor nodes for vibration measurement (IoT technology) with AI-based diagnosis approach of dynamical system.
  • Implementation of AI-based diagnosis approach of dynamical system via personal and movable AI/BD integration computing platform.  
  • Application on bridge and buildings health monitoring.  
  • Application on wind turbine condition monitoring.

Integration of sensor nodes for vibration measurement (IoT technology) with AI-based diagnosis approach of dynamical system

Implementation of AI-based diagnosis approach of dynamical system via personal and movable AI/BD integration computing platform

Application on bridge and buildings health monitoring
Application on wind turbine condition monitoring.

Deep Learning for Flooding Water level Image Recognition- Water Resources Agency, MOEA Cooperation Project

Contacts: Dr. Wu
Email: c00jhw00@narlabs.org.tw 

In cooperation with the Water Resources Agency, Ministry of Economic Affairs, combined with deep learning and intersection surveillance images, vehicle identification is used to identify the image of road flood depth estimation. This method can use the existing monitoring system in the city to quickly improve the flood warning and notification flood detect system.


Flooding Water level Image (1)


Flooding Water level Image (2)


Image Recognition example


Deep Learning Database

Take the Hsinchu City Government as an example - Government departments how to promote the study of artificial intelligence and smart city applications  

Contacts: Ms. Vivian Cheng  
Email: T0013@ems.hccg.gov.tw 

The Hsinchu City Government uphold the opening data, create innovative ideas and mutual prosperity. At the same time, we pay attention to the impact of artificial intelligence technology on smart cities. It is well known that artificial intelligence technology development materials are often indispensable, so the high-quality fisheye camera image data is actively released and cooperate with National Chiao Tung University and National Center for High-Performance Computing. Based on the fisheye camera image traffic survey technology, it is expected to promote the research and application of advanced artificial intelligence technology in urban traffic management.

Cooperation results 
  • Crossroad fisheye monitor image traffic detection application
  1. Vehicle identification
  2. Analysis of traffic hot zone
  3. Track tracking and counting
  4. Image correction / vehicle image segmentation and tracking 

  • Future direction

  1. Analysis of the relationship between city-level traffic information and people flow information  
  2. Importing artificial intelligence technology to optimize traffic sign control and traffic route planning  
  3. Intelligent instant traffic message broadcast    


Analysis of traffic hot zone


Analysis of traffic hot zone

The Analysis System of People-flow Behavior and Traffic-Flow Behavior

Contacts: Dr. Huang
Email: schumi@narlabs.org.tw  

The People-flow Behavior and counting system based on object tracking is used for visitors calculation, people flow density, security in smart city. The Automatic Vehicle Counting and Classification System (AVCCS) base on image recognition and deep learning is proposed to count and classify vehicle number and behavior in the city, and provide valuable information, such as prediction of the traffic flow behavior model and optimize the traffic lights.


The Huludun Park of the Taichung World Flora Exposition covers an area of about 18 hectares and is one of the important venues in the 2018 World Flora Exposition. Since the Huludun Park is a free open park, there is no ticket gate, in order to effectively understand and count the statistics of visitors. 

This application achieves monitoring and control of the number of people in the park through image recognition of human flow technology and analysis technology. This application builds an automated visitor counting analysis system based on image processing and computer vision identification technology.  

It can conduct statistical analysis on the entrances and exits of visitors to the World Flora Exposition, and instantly return the data to the Taichung City Government Information Center via the Internet. 

Automated flow analysis counts dramatically reduce the cost of manual calculations at traditional sites.24-hour automated safety monitoring, based on data and image follow-up to judge the abnormal behavior of the masses. 

This application adopts image processing and computer vision recognition technology to design the visitor behavior analysis system. Through image recognition, each person's independent image features are identified, and a set of numbers belonging to the visitor is given. The system identification process, if there is crowd overlap, The system can also identify and re-track through each person's image features, greatly reducing the problem of confusion after overlap.  

Through the high recognition capability of the system, it can be identified only by the standard shooting angle. It can be widely used in various occasions and can be extended to various applications, such as pedestrian hotspots and urban traffic. Effectively solve the problem of counting angles of overhead cameras, and infrared counting sets the disadvantages of traditional counting methods such as electronic fences.  

The system can analyze historical people flow, time flow and other important indicators, and has a proactive state return system, which can notify relevant personnel at the

Integrating Autonomy into Urban Systems

Contacts: Dr. Cathy Wu
Email: cathywu@mit.edu

Self-driving cars are widely projected to have significant positive impact on urban systems — from accidents to congestion to greenhouse gas emissions to mobility access.
At the same time, when they do enter our urban systems, recent research shows that, in the US, they may improve our transportation energy consumption by 40% or worsen it by 100%, due to complex factors including induced travel demand.
In order to understand and shape the adoption of such automation, to guide towards a desirable outcome, we need tools with which to study the impact of automation on such large-scale complex urban systems.

This demo shows the potential for deep reinforcement learning (RL) to provide insights into this complex challenge.
We start by studying a series of "traffic LEGO blocks," pieces of complex urban networks, and each of which are challenging control problems.
Employing techniques from deep RL, we found that even 4-10% of automated vehicles can greatly affect low-level traffic dynamics, producing a variety of interesting emergent behaviors, and improving the average system velocity by 30%-140% in diverse traffic scenarios.
This is just the beginning of our research journey into the study of the impact of automation on complex societal systems, and these experiments demonstrate the potential for deep RL as a key tool for future studies.

Innovative applications of the big data of offshore wind energy in Taiwan

Contacts: Mr. Chang
Email: c00wyc00@nchc.narl.org.tw 

Green energy research and development plan is one of the Taiwan government’s five major innovation industries of science and technology. It is planned the green energy in 2025 will be 20% of total energy used in Taiwan. While the offshore wind power is targeted for 5.5GW in 2025, that is required to install 600 offshore wind turbines, equivalent to an investment amount of 480 billion NTD. In cooperation with Academia Sinica, National Taiwan Ocean University, Taiwan Ocean Research Institute, Taiwan Generations Corp. energy company, this project targets on the Fuhai wind farm as study site, which is located in Changhua offshore area. In this study, the high-resolution marine and atmospheric models, with capability of simulating typhoon scenario, are used for forecast in the wind farm to establish the weather and marine big data (BD) database. Then, the deep learning analyses provide the high-precision offshore wind, wave and current fields for the next five days, and the reinforcement learning analyses provide the optimized path planning for offshore shipping companies. Currently, the high-precision wind forecast technique has been extended to TPC’s offshore wind field. It is believed that the above developed AI technologies can provide valuable assessment and applications for other offshore wind power plant manufacturers in the future. 

To make it easier for researchers/engineers in offshore wind energy industries or other related engineering fields to apply AI/BD analysis, this project has developed a personal and movable AI/BD integration computing platform. The 1+3 VM (Virtual Machine) system can be used for education and training, and the 1+3 physical cluster system allows users to quickly deploy applications. Thus, it will be very beneficial for academic research teams or SMEs, with sensitivity/security data. Presently, the product prototype has been completed. We’ll keep cooperated with several teams for application development, and more and more demonstrative use cases will be established.  

  • Taiwan offshore wind energy (weather and marine) BD database establishment and display
  • Deep learning implementation for wind/wave/current field prediction and forecast service 
  • Reinforcement learning implementation for optimized path planning for shipping 
  • Development of personal and movable AI/BD integration computing platform