Feeding data-hungry Artificial Intelligence: Insights from the DOT AI Industry Day

Posted December 31, 2018 by Daniel Baldini

In November, I attended The Department of Transportation (DOT) Artificial Intelligence (AI) Technology Day. The one-day symposium focused on the use of AI. Representatives from different industries presented the work they are doing with AI, including IBM’s Watson, the Smithsonian’s Pepper, and the Federal Highway Administration’s Cooperative Automation Research Mobility Applications, to name a few. Many of these groups provided examples of their own AI creations throughout the event, and how they feed these data-hungry tools to compute vast amounts of data to make better decisions. Below are my impressions of the symposium’s presentations along with how these efforts link to Volanno’s research project awarded by DOT’s Small Business Innovation Research Program (SBIR).

Watson

Representatives from IBM gave insights into IBM’s Watson. This computer system has the capability of answering questions posed to it in a natural language. IBM’s Watson team noted that the top three hot topics in AI at the moment are information accessibility, unstructured data, and expertise scaling through training and predictive analytics. Essentially, AI and machine learning principles are making unstructured data more accessible to humans. A human brain cannot compute this vast amount of data. But an AI tool has the potential to analyze data in many more forms and create more accessible, understandable representations of that data.

Watson has performed a large test pilot with the US Army and its fleet vehicles. Watson took unstructured data from sensors on vehicles, maintenance manuals, safety materials, vehicle history information, and other reports to interpret and analyze data, also known as condition based maintenance. Watson was able to detect and pinpoint specific issues for a full diagnosis reducing cost and time. IBM intends to eventually integrate Watson into the cockpits of airplanes, to help pilots and airlines diagnose mechanical issues quicker. Fueling Watson with past flight data and real time data will result in further accessibility for airlines to previously inaccessible data. IBM foresees a number of benefits for this integration, especially with increased commercial flight safety. IBM’s Watson Team stated, “Artificial Intelligence will help us find better, more relevant information quicker, and help us consume information at a faster pace.”

Pepper

A team from Smithsonian presented on the humanoid, Pepper, which is available at four Smithsonian locations currently. Pepper is a humanoid robot developed by Softbank Robotics, designed to communicate with humans. Softbank donated 25 humanoid robots, which will be programmed specifically for the locations they will be placed at.  For example, the Pepper at the National Museum of African Art, has been taught to translate Kiswahili phrases for the World on the Horizon exhibit. Recently Softbank agreed to provide 100 more humanoids to the Smithsonian Institute. The Smithsonian hopes to eventually have Pepper at all 19 Smithsonian locations in the near future, depending on the results of the pilot program.

Pepper can sense when people are close by and engage them in conversation. Pepper also has the ability to interpret museum-goers interactions with exhibits. This is to boost engagement and provide an interactive experience based on the attendees’ feelings, perspectives, and interests. For instance, at the Hirshhorn Museum, Pepper will pull up a piece of art, based on how a person is feeling, and subsequently answer any questions related to the art or exhibit. Pepper utilizes emotion recognition AI from the company Affectiva, which allows Pepper to understand different feelings being portrayed, such as laughter, joy, and sadness.

Pepper: The Smithsonian’s humanoid docent. (Courtesy of the Smithsonian Institution)

CARMA

The Federal Highway Administration (FHWA) presented on their Cooperative Automation Research Mobility Applications (CARMA), which is aimed to encourage collaboration with goals of improving both safety and efficiency. CARMA facilitates Automated Driving Systems (ADS), commonly known as self-driving vehicles, to maneuver through roadway infrastructure and other vehicles through communication and data exchange. CARMA is developed as open-source software and is available on GitHub. CARMA is vehicle and technology agnostic. Enabled by CARMA, vehicles would be able to communicate with one another on roads to improve road safety and cut down on delays/bottlenecks. The platooning of vehicles on roads is being heavily researched, as a way to increase road capacity. For example, as self-driving cars communicate with one another, vehicles that are headed to a similar destination would have the ability to platoon and travel very closely. This is a way to fit more cars on the road and cut down delays.  In addition to platooning, the CARMA software platform has plug-ins that support the following human-like driving tactics: cruising, yielding, lane changing and merging, and speed harmonizing. Data is fed to CARMA through a vehicle’s devices and microcontrollers through the vehicle’s Controller Area Network. CARMA also interfaces with the onboard unit which can provide dedicated short-range communications for other vehicles and infrastructure.

Understanding Data and Humanoids

The US DOT is trying to have a better understanding of safety data and how to better process this data. This is evident by the number of data driven innovation opportunities being put out compared to years prior. The vast amount of data being generated has significantly increased due to the sharp growth in the number of data sources coming online via sensors, systems, etc. The current focus is more on prescriptive analytics which differs from predictive analytics. Prescriptive analytics has the ability to advise us on possible outcomes, while predictive analytics aims at telling us what might happen. Volanno is currently engaged with the DOT’s Small Business Innovation Research Program (SBIR) to develop a new way to present vehicle safety data. Under the program, Volanno is evaluating various input data streams, identifying relevant sources, and developing an approach to consolidate data from such sources. Our techniques allow us to deliver information in response to user preferences, and give these users an opportunity to get prescribed information to conduct necessary automobile research.

The keynote speaker of the event, Secretary of Transportation, Elaine Chao, noted that “AI can also be leveraged to improve the Department’s internal operations, enhance customer service and stakeholder outreach, and strengthen research.” The DOT’s main focus at the moment pertains to safety, innovation, and infrastructure. The DOT also believes that innovations in technology can be used to solve problems, enabling progress. The transportation industry continues to be data driven, so it is important for the DOT to be able to keep up with new sources of data and make updates as needed.

DOT emphasized that now, more than ever, it is depending on partnerships with organizations to help move them forward. Safety is a large concern for the DOT, especially on the roads, as accidents have not been decreasing despite numerous improvements.

AI in the transportation industry is valued at approximately $1.2 billion, and is expected to have steady growth in the next ten years. With urbanization growing at a fast pace, our roads are being heavily impacted. In 2017, there were 268 million cars registered to be on the road. This increased to 272 million at the beginning of 2018, and is expected to continually rise (statista.com). With the popularity of ride sharing and millennials relying less on public transportation, ride sharing companies have created unintended consequences and are congesting our roads even further. If the DOT is able to take their data and apply it to traffic management through AI, they could have the potential to streamline traffic patterns, create smarter traffic light algorithms, and continue to increase road safety.

DOT emphasized that now, more than ever, it is depending on partnerships with organizations to help move them forward. Safety is a large concern for the DOT, especially on the roads, as accidents have not been decreasing despite numerous improvements.

Nonetheless, AI creates new sources of business value for executives in companies and throughout the federal government. The AI examples above are some of the current adopters actively using these new intelligences to see impressive results in their businesses, operational strategies, and organizations as a whole. As these AI humanoids and tools are fed more and more data, their potential to compute and process data increases tenfold. For those seeing the incredible computational power of AI, the fear of the humanoid AI replacing the human workforce is decreasing. AI humanoids such as Smithsonian’s Pepper not only work with the staff, but also create new jobs to help manage the program. AI is creating hope and success for industries throughout the world.

Ultimately, AI allows machines to be fed new data and to learn from experiences, adjust to new data, and create human-like responses. For the DOT, AI promises to have a positive impact on the transportation industry in the very near future. It is of the upmost importance that data must be available and ready in order to perform any sort of AI Analysis. AI will have resonating effects across the board, from safety, decision making, traffic patterns, autonomous vehicles, and capacity.

About Volanno

Founded in 2003, Volanno Inc., is an award-winning, Women-Owned Small Business (WOSB) based in Washington, DC providing software development, data analytics, project management, and technology implementation services to the transportation industry.

Recently at a Data Conference at the University of Delaware, Volanno presented a paper called “Evolution of Operational Analysis Using Discrete Data Streams and Big Data Approach: case Study- Prediction of Train Arrival Times” where we are using statistical models and artificial intelligence techniques to improve ETA calculations. For more information, contact us at bizdev@volanno.com