AI has enormous potential in infrastructure. Upwards of US$153 trillion will be spent on infrastructure in the next ten years globally and such projects need innovative thinking from asset managers, engineers and computer scientists. The US is spending trillions of dollars on infrastructure and China likewise is spending $2 trillion just on infrastructure associated with highways and railways. Another $1 trillion from China is slated for infrastructure in Africa, particularly Nigeria, as Africa’s largest country.
The secondary positive effects of infrastructure spending is the corresponding consumer spending that boosts investment in other sectors. The approach of the US and China of a comprehensive national infrastructure strategy gives businesses opportunities to expand in supportive areas, such as malls, high speed rail stations, restaurants and housing. In the innovation space, because both the US and China have national infrastructure policies and are tech leaders, both will use AI to improve efficiencies, capture big data, streamline design and build processes, as well as improve design and safety in respect of infrastructure before other countries.
While many countries have infrastructure investment projects planned, many will need to be re-engineered for big data capture for AI and that opportunity, namely connecting big data and AI in infrastructure design, build and asset management, is a critical piece in infrastructure where China and Japan have already made inroads. The conversation in Canada of innovation, infrastructure and AI hasn’t progressed as it should because of a shortage of young people in the infrastructure and the lack of grants to innovative startups.
In infrastructure, predictive AI can detect when the probability of major infrastructure disruptions will increase significantly, whether triggered by natural or man-made causes This anticipatory capability is particularly vital; it can help predict where problems will arise, to adapt infrastructure operations to forestall disruption as, or before they happen.
Predictive AI will be significant for airport and flight management. Flight delays have a negative impact on airlines, airports and passengers and thus being able to predict events and manage the impacts is critical to the decision-making process in aviation. Airports using air traffic systems with AI and machine learning for airspace efficiently could save the US economy US$50 billion per year in saved costs from airport delays.
Flight disruptions are estimated to cost airlines US$25 billion per year. Using AI, machine learning, predictive AI and cognitive computing to predict various types of flight disruptions and make operational changes to accommodate the disruptions would put a dent in flight disruption costs. It is estimated that predictive AI will be adopted by most airlines and airports within the next five years.
Predictive AI is being used to improve airport management by leveraging big data analytics and machine learning to predict airport operational conditions five times sooner than previously possible. To predict airport operational conditions, analytics use machine learning models and numberical analysis based on historical patterns to predict future conditions. For example, if an airport always changes its runway configuration when the wind blows in the opposite direction, and based on the weather forecasts, the wind at the airport will change directions at 2 pm, then the model can predict a runway configuration change after 2 pm. Since runway configuration changes cause approach patterns and taxi times to change, airlines that know operational conditions ahead of time can optimize planning flights, routes and fuel consumption.
AI is also used to improve public safety in airports. For example, video analytics software can detect primary hazards in airports and be used to alert airport administrators about suspicious objects or intrusions into restricted areas. Advancements in video analytics make it possible to reconstruct 2D video images into 4D understanding by adding depth and time, thereby improving accuracy and reducing the rate of false-positives. Facial recognition, long used at airports in China, is being embraced in the West as well for airport security and immigration. The use of AI in a review of historical video surveillance footage helps match a suspect’s face for law enforcement purposes at airports.
As well, queue detection functionality in video analytics software can help eliminate long lines, whether they occur in drop-off and pick-up areas or in security and passport control queues. AI can help alert airport administrators when crowds begin forming in real time, so that they can deploy staff to open more ticket counters, security gates or passport control lanes, thereby decreasing wait times and increasing passenger satisfaction. Data captured by video analytics can also help airports predict and optimize staffing plans and scheduling to ensure that airports better support patterns of activity.
The Kaggle data science community, based in the US, is hosting an online competition to build machine learning tools that can augment airport security agents to make airport management systems more accurate and efficient. Competitors will develop ways to improve threat recognition algorithms to accurately predict the location of threat objects on a person’s body. The TSA is providing a data set of images so that competitors can train their algorithms on images of people carrying weapons.
The Gatwick Airport in London is creating a seamless passenger experience from curb to gate and, to that end, is tracking different areas of airport activity to predict activity and passenger volumes to improve airport management and service delivery. A Switzerland-based company is developing disruption warning and prediction capabilities using industry-specific and public data feeds, such as Twitter, to monitor imminent disruptions and allow for disruption management to reduce costs of aviation management.
PASSUR Aerospace, a US company, uses predictive AI to improve performance of airline and airport logistics using multiple data sources from disparate systems, together with machine learning, to create an integrated solution for airport logistics.
Researchers from State University of New York and Binghampton University developed an AI program to predict how long passengers will be stuck waiting for flights using multiple data sources, including weather and air traffic data to allow airlines to see relationships between input variables (such as weather) and output variables (such as flight delays). The program is designed to help airlines inform passengers more quickly and accurately about travel conditions.
In the US, the TSA is testing AI technology to increase the speed and accuracy of airport security screening using a new type of full-body scanning that scans multiple passengers at a time, incorporating facial recognition to identify people who may be in a database and to detect facial expressions and postures that may pose a security risk, although the latter raises legal and ethical concerns in respect of AI law.
The San Francisco International Airport is undergoing a US$2.4 billion renovation that will include, inter alia, a new AI software tool that automates cost-benefit analysis for buildings and sites. The idea is to use AI to manage where construction funds are allocated, i.e., to evaluate each proposed design element through a comprehensive business case analysis of costs and benefits for financial, social and environmental bottom lines. For example, the AI software predicted that building a green roof over Terminal 1 of the Airport would yield a US$5 million saving over a 50-year timespan, making a compelling case for its inclusion in the airport’s redesign.
JetBlue, a budget US airline, is launching a trial of biometric-enabled flight self-boarding at the Logan International Airport to use facial recognition technology to optimize the process.
Airports in Dubai are beginning to use automation software to improve airport planning, particularly with respect to coordinating resources such as check-in gates and aircraft stands. AI can perform calculations for 40 flights per second and integrate information from multiple sources to improve the automation process.
The operational, communicative and collaborative abilities of ports, particularly with respect to cargo owners, agents, shipping companies and other port units, can be vastly improved using AI. AI in port settings can closely track cargo, equipment, personnel and logistics flows to intelligently detect risks, maintain operations and ensure the safety and security of local cities. It is predicted that, in the future, advances in AI will allow ports to connect seamlessly with their machine control functions to allocate and distribute resources as efficiently as possible under different operational conditions and environments.
The US Coast Guard uses predictive AI for port security to schedule patrols to deter and disrupt potential threats, including terrorist threats. For example, the system uses data on passenger load numbers and traffic changes to create schedules and make it challenging for a bad actor to predict opportune times to attack infrastructure.
In Japan, Mitsui OSK Lines and the Yokohama National University are using predictive AI to analyze data related to ocean transportation with a view to predicting the ocean shipping market and bunker prices more accurately for a competitive edge in the shipping industry with respect to transport and logistics.
ClearMetal, a US company, uses predictive AI and machine learning to manage shipping logistics as freight moves from one port to another. It aggregates data about carrier operations and weather forecasts, labour issues and economic changes. The data is used to create a predictive model for 75 points of uncertainty, which updates in real time to respond to changes. The platform uses AI and machine learning to analyze historical data for patterns pointing to future trends. For example, ClearMetal can tell a company that 35 containers will be ready to ship from Vancouver to China in 2 weeks; when it is expected to arrive in Vancouver; where in the port it should be delivered; as well as what the traffic and weather are like at the port of delivery. With weather conditions affecting shiping or cancelled shipments, the program adjusts its projections so that the company can allocate more resources or recalculate a shipping route and delivery time.
Cities around the world are beginning to use AI to optimize the delivery of transportation and for highway management, such as by using sensors and cameras to optimize traffic light timing for the optimal traffic flow. The rising availability of big data makes transportation an ideal sector for AI disruption. AI will increasingly impact city infrastructure as increased availability of data from residents will be used in predictive models to deliver services. AI in smart traffic light systems can save the US economy US$121 billion per year and eliminate 56 billion pounds of CO2annually.
AI can be used for vehicular traffic management in highways to improve safety by increasing situational awareness, and to provide drivers with real-time route information. AI applications can also be employed to improve network-level mobility and reduce overall system energy use and transportation-related emissions. With the adoption of intelligent highway technology, AI can increase highway capacity by up to 273%. For example, in the UK, traffic lights are being fitted with technology for safety and roadway load management. The AI-powered traffic lights monitor speed and congestion, prioritize cyclists, buses and ambulances with green lights and use heatmaps to analyze how pedestrians and motorists are using the road. The monitors detect different road uses and are able to regulate traffic in real-time. China has partnered with Vancouver’s quantum computer company, D-Wave, to bring smarter traffic management systems with AI to China’s largest cities.
In China, Alibaba’s Aliyun is being used to predict real-time traffic in collaboration with the Hangzhou government to alleviate road congestion. Aliyun provides real-time traffic recommendations and travel routes based on its video and image recognition AI technology. By employing Aliyun, the Hangzhou government has seen an 11% increase in traffic efficiency on Hangzhou roads.
In India, the Tata Group is creating a series of roads to develop an autonomous traffic system that will account for pedestrians that walk through traffic at will, lanes that merge without warning, poor signage and stray cattle. Indian roads present a deep learning challenge because they involve modern highways and dirt tracks, with erratic street signage and the risk of stray elephants or camels, as well as three-wheeled rickshaws that ride the roads with cars.
In the US, the Regional Transportation Commission of Southern Nevada and Israel-based startup WayCare, a predictive analytics platform for smart cities, launched a pilot program to help prevent crashes and reduce traffic backups. A traffic management centre will record and analyze traffic flow on two highways in Las Vegas, cross-referencing historical data with real-time data, including, inter alia, traffic light timing, major traffic events, weather conditions, vehicle location and speeds to optimize traffic management. The pilot is premised on the idea that resources can be better deployed to problem areas to reduce traffic fatalities and congestion, saving Las Vegas millions of dollars in law enforcement and emergency services, and municipal insurance expenses.
German automaker Volkswagen is researching ways in which deep learning and quantum computing can be combined to optimize traffic flow in densely-populated cities, analyzing many different inputs, including stop lights, traffic flow, pedestrians and road construction, to predict which routes are better for drivers by studying past data about traffic patterns and recurring traffic situations in urban environments to predict future traffic scenarios.
Oil & Gas
AI is playing a growing role in infrastructure in the oil and gas industry, particularly with respect to monitoring pipeline integrity and predicting volume demands. Pipeline integrity is a significant critical infrastructure component in oil producing countries. In Canada, for example, the pipeline sector accounts for more jobs than agriculture or financial services. Pipeline robotics are already in development – they access pipeline integrity to ensure spills don’t occur – a particularly vital step with environmentally sensitive loads.
The Southwest Research Institute in the US is developing a machine learning system to improve pipeline safety which will be able to capture and process image data from various thermal and infrared cameras to detect leaks in pipelines to mitigate leakage and spills early. In addition to saving costs, early detection allows for more effective and responsive environmental management.
Wind Farms/Wind Turbines
AI is also helping to monitor and improve the efficiency of wind farms and wind turbines in remote locations. For example, wind turbines off Denmark near the North Sea are so geographically isolated that they cannot be serviced for many months of the year. As a result, companies operate remote diagnostic centres nearby using AI to monitor turbines, analyze the data to detect irregularities potentially indicative of impending failures and to make remote operational decisions.
In the US, the National Center of Atmospheric Research is using AI to help understand how much wind power will be produced the next day so that energy utilities and grid operators can use wind turbines at peak efficiency. The AI-based software considers wind speed, together with data from weather satellites, stations and other wind farms, to determine how much wind power is available to the grid from the turbines at a given time and when there are changes in wind capacity to make predictive decisions on energy.
The University of Manchester has a project to deploy AI and robotics for the remote operation and maintenance of offshore wind farms. Over 80% of the cost of offshore wind farm operation and maintenance costs arise from remote site access difficulties. The project will look at the use of AI and robotics to reduce maintenance costs by creating predictive and diagnostic techniques to allow problems to be detected early, and implement cost-effective maintenance solutions. As well, robots and advanced sensors will be used to reduce the need for human intervention in hazardous and remote environments.
The idea of using robots to attend to remote wind farms is not new. In 2013, Helical Robotics, a US-based robotics solutions company launched robotic platforms to improve safety and reduce liability and personnel costs for wind turbine maintenance, and created a lightweight magnetic climbing robotic that weighs less than 40 lbs but can lift 20 lbs, deployable in less than one minute, making it useful on a myriad of wind turbines. The robot can scale wind towers for blade inspection, traditionally performed by a human inspector, and can examine turbine blades from the ground about 100 meters away using a high-powered telescope.
AI is perfectly designed for hydroelectricity. Each of the components of a hydroelectric dam from intake, turbine, alternators, transformers, dam status and drain system lend themselves naturally to improvements from predictive AI, and in addition, predicting consumption of the resulting electricity is also AI relevant, which could allow for more tailored capacity and performancy of hydropower.
Dams produce 17% of the world’s electricity and digitizing big data from hydroelectric dams is crucial to adopt AI to continue to meet demands with efficiency. The opportunity lies in the fact that mathematical models operating power production are 30 years old and new methods are needed to optomize energy globally. Appications of machine learning to automate dams and predictive AI for resource management would improve hydropower efficiency and save costs. With climate change affecting water availability and supply, the role of AI in hydroelectricity is even more pressing.
At the University of Guelph, engineers are using machine learning to help turn small dams into cost-effective renewable energy resources for rural users based on the Archimedes screw developed in 250 BC; with the Archimedes screw, water flowing over a dam enters a channel containing a screw, which turns from the weight of the falling water, thereby converting the resulting mechanical energy into electricity. The University of Guelph researchers are using neural networks and machine learning to determine how to optimize customization of the Archimedes screw to enable users in rural parts of Ontario to generate power for several hundred households.
The Digital Finance Institute wrote and published a Report on Commercial AI (available here) that canvassed the pulse of AI from media stories and academic reports, covering various sectors of the economy. This article covers the Infrastructure portion of the Report.
[1 ] B. Genge, Dependable Systems and Networks Workshop, “Data fusion-based anomaly detection in networked critical infrastructures,” 2013.
 Christine Duhaime, Artificial Intelligence Insights, Report on the future of AI, October 15, 2016.
 BTN News, “Aviation now using artificial intelligence to predict flight disruptions and more,” January 26, 2017.
 Mark Miller, Aviation Pros, “The Next Wave In Airport Efficiency: Predictive Analytics,” April 6, 2016.
 Airport Business, “Travelers Can Make Themselves at Home in Smarter, Safer Airports,” June 27, 2017.
 Airport Business, Aviation Pros, “Travelers Can Make Themselves at Home in Smarter, Safer Airports,” June 27, 2017.
 John Mannes, TechCrunch, “The Kaggle data science community is competing to improve airport security with AI,” June 22, 2017.
 BTN News, “Aviation now using artificial intelligence (AI) to predict flight disruptions and more,” January 26, 2017.
 Sys-Con Media, “PASSUR and GE Partner to Advance Airline and Airport Operations,” June 20, 2017.
 Mark Prigg, Daily Mail Online, “The AI that could tell you how long your flight is delayed for,” November 14, 2016.
Jim Ogden, The Jetsetter’s Homestead, “Artificial Intelligence Could Speed Airport Security,” March 24, 2017.
Sys-Con Media, “San Francisco International Airport uses insights from Autocase to convert “triple bottom line” from aspiration to reality on $2.4B Renovation,” June 19, 2017.
 Robotics & AI, Future Travel Experience, “Artificial intelligence, robotics and biometrics on JetBlue’s technology watch list,” June 2017.
 Karl Flinders, Computer Weekly, “Dubai Airports automates resource planning,” June 8, 2017.
 Accenture and SIPG, “Connected Ports Driving Future Trade,” 2016.
 Yasmin Tadjeh, National Defense, “Company Using Artificial Intelligence to Improve Security,” June 1, 2016.
Port Technology, “MOL to Tap Artificial Intelligence for Shipping,” September 1, 2016.
Sue Walsh, RT Insights, “ClearMetal Announces Predictive Analytics Platform for Shipping,” February 29, 2016.
Transportation Planning, Stanford University, “One Hundred Year Study on Artificial Intelligence Transportation Planning,” 2015.
 Hussein Dia, Emerging Technologies, “An agent-based approach to modelling driver route choice behaviour under the influence of real time information,” 2002.
 Evan Ackerman, IEEE Spectrum, “Study: Intelligent Cars Could Boost Highway Capacity by 273%,” September 4, 2012.
 Jamie Rigg, Engadget, “AI will take control of Milton Keynes traffic lights next year,” May 17, 2017.
 AI Toronto Conference, Digital Finance Institute, Toronto, March 29, 2017.
 Saritha Rai, Bloomberg Technology, “Anarchy on India’s Roads Has Driverless Car in a Jam,” June 29, 2017.
 Mick Akers, Las Vegas Sun, “RTC, tech firm tean uo on pilot program to better anticipate accidents, traffic jams,” July 12, 2017.
 Sara Castellanos, Wall Street Journal, “Volkswagen to Use Deep Learning to Optimize Traffic Flow,” June 28, 2017.
 Matt Slowikowski, Oil Price, “How Artificial Intelligence Could Help Transform the Oil Industry,” August 21, 2016.
 David Holley, XConomy, “Pipe Leak Detection System Employs Machine Learning to Limit Errors,” April 7, 2017.
 Mary Gorges, Cisco, “Oceans of data from world’s offshore wind farms,” February 2017.
 University of Manchester, Phys.org, “Artificial intelligence and robots to make offshore windfarms safer and cheaper,” March 9, 2017.
 Joshua S. Hill, Clean Technica, “Robotics to Help Maintain Offshore and Onshore Wind Turbines,” April 3, 2013.