Stevens' Summer Research Institute Marks Tenth Year of Student Innovation in Maritime Safety and Security
SRI connects multidisciplinary student teams with government and industry professionals to develop real-world solutions to safety and security challenges on U.S. waterways
From June 3 to July 26, Stevens Institute of Technology's Maritime Security Center (MSC) conducted its tenth annual Summer Research Institute (SRI), marking a decade of successful maritime and port safety-and security-focused experiential learning and applied technological research.
Representing 11 academic disciplines, this year's cohort of 25 students hailed from seven schools, including two minority-serving institutions: the University of Puerto Rico at Mayagüez and the University of North Carolina–Pembroke, the latter of which participated in the SRI for the first time.
Each year, project topics for the eight-week program are developed in conjunction with the Department of Homeland Security (DHS) stakeholders, representing the current needs and interests of the homeland security enterprise.
"Our goal is to create something that the U.S. Coast Guard or other DHS component agencies can use immediately in accomplishing their challenging missions," says Hady Salloum, MSC director and electrical and computer engineering professor. "We let the students work on their own as much as possible, and we encourage them to take their ideas and efforts to the next level, whether it's a startup or to pursue a patent or commercialization channel."
Of the 2019 cohort's six projects, two continued from last year, including one that saw the SRI's first invention disclosure. Three of those projects are highlighted here.
Augmenting manpower through unmanned aerial systems
A reality faced by the U.S. Coast Guard every day is the simple fact that water refuses to stand still.
By the time Coast Guard personnel are able to mobilize and reach the last known coordinates of a person reported to have gone overboard, for example, that person has already moved, drifting with the current to an unknown location potentially miles away.
Similarly, officers giving chase may be forced to choose between pursuing criminal suspects and retrieving contraband thrown overboard before the goods' confirmed location is lost.
That time delay between an incident's occurrence and a target's discovery could mean the difference between life and death or criminal conviction and exoneration.
One SRI project team's unmanned aerial systems (UAS) buoy system, however, could help homeland security practitioners "be" in two places at once.
Equipped with GPS and environmental sensors, the lightweight self-locating datum marker buoy is autonomously deployed from a drone, dropping into the water near a person or object's last known location. As the buoy drifts with the current, it follows the target's approximate path.
Potential applications for this UAS buoy system include search and rescue, ocean modeling, accident and pollution monitoring, and drug interdiction.
"The only difference between each application is the amount of time that it's in the water," says Stevens mechanical engineering undergraduate Maria Manoussakis. "We're not trying to home in on one specific application right now. We're just trying to get a prototype."
The 2019 UAS buoy project picks up where last year's left off. Using the 2018 team's ideas as a springboard, the 2019 team — returning Stevens software engineering undergraduates Eric Fernandes and Herb Zieger and Stevens mechanical engineering undergraduates Manoussakis and Jason Chang — spent the first couple of weeks redesigning the buoy and drone release mechanism.
Originally conceived as a separate piece in which the buoy sat, the release mechanism, the team realized, could be streamlined.
"One professor that works with aerodynamics and aerial systems said, '[The independent release mechanism] is unnecessary. Why make it harder for yourself?'" Manoussakis says. "So we abandoned that idea and made the buoy the release mechanism."
Working with research professor Barry Bunin, the team also found themselves addressing design issues on the fly while already immersed in the prototype's assembly stage. Having built fully functional drogues with which the buoy would float with the current, they discovered they'd failed to consider the need for springs to keep the drogue's arms deployed.
With neither springs on hand nor sufficient time to secure any, "we had to kind of improvise," Manoussakis says.
But materials and assembly of the buoy itself posed perhaps the team's biggest challenges.
They built the buoy's fiberglass body with guidance from unmanned autonomous systems manufacturer Duro UAS. But the meticulous process of making fiberglass proved more difficult and time-consuming than anticipated.
"You have to use a very precise amount of resin with hardener, and you have to use it within a minute or else it gets hard, and you can't use it," Manoussakis says. "It took a very long time for us to get used to it."
Using a combination of machined and 3-D-printed parts, the team was able to complete assembly of a functional, if not fully operational, buoy prototype system. Although not yet waterproof (an undertaking so immense it must wait for future iterations), the prototype brings together the UAS system's functions and components into a cohesive whole.
The 2018 UAS buoy project received the SRI's first invention disclosure. With the 2019 team's continued progress and refinements, government and commercial interest in the system remains high.
Visualizing data for more effective maritime risk management
The Coast Guard records safety, security, and law enforcement incident data in a nationwide database called MISLE (Marine Information for Safety and Law Enforcement). This data represents all major and minor incidents occurring on domestic and territorial waters, including environmental spills, criminal activity, allisions and accidents, and boating under the influence.
But the process of accessing and analyzing data from such an extensive database can be tedious and cumbersome, requiring multiple steps, spreadsheets, and even software systems. Changing just one search parameter could mean having to start the whole process over from scratch.
Fast, efficient data analysis is crucial for making proactive resource allocation and budgetary decisions that improve safety and security outcomes. With resources already at capacity, Coast Guard operational field command Sector New York — whose area of responsibility spans more than 6,000 square miles, including the Port of New York and New Jersey — looked to the SRI for assistance.
Enter the risk management dashboard.
"[Sector New York] is already analyzing their data," says Virginia Tech civil engineering graduate student Emily Jannace, "but not in an effective way. We wanted to make a visualization that quickly and efficiently gives a visual representation of the trends that occur in this data. That way they don't have to sift through, for example, all 2000 lines of search and rescue incidents in the 2018 data."
Working with Bunin and associate teaching professor and information systems master's program director Paul Rohmeyer, the risk management dashboard team — Jannace, Stevens mathematics and computer science undergraduate Mathew Seedhom, SUNY Maritime College maritime studies undergraduate Danielle Dobbs, and University of Alaska Anchorage data science and geographical information systems graduate student Tristan Goers — developed a prototype visual interface that provides a comprehensive overview of incident data and trends summarized through an easy-to-scan display of charts, graphs, and a regional map.
Lacking access to proprietary Coast Guard systems, the team built the dashboard to display through a website developed on a Python-based framework (Django) by pulling pertinent data from a PostgreSQL database. Interactive graphical elements are generated via JavaScript libraries.
The dashboard seeks to facilitate agile, quantitative decision-making, such as when and where to beef up patrols.
"When they go into the morning briefing, they can use these visuals and these trends and risk analysis to know how to allocate their resources that day, and then the next week, and the next month," Jannace says.
The team consulted regularly with Sector New York officials, whose insight into the practical needs of Coast Guard personnel directly contributed to the team's thinking.
That insight proved invaluable, even when it sent the team back to the drawing board. Envisioned as a tool that would update in real time, the original dashboard concept had to be completely reimagined when the team learned incident data in MISLE could take up to 72 hours to appear.
"From our initial design," Jannace says, "it changed everything."
But a bigger challenge arose when the team discovered that the huge volume of incident data, plus the time required entering it, sometimes resulted in incorrect or incomplete entries.
Some incidents, for example, were entered without dates and times, while others were left without categories. Most confounding were incidents with longitude and latitude coordinates for locations in Africa and Antarctica, far afield of Sector New York's jurisdiction.
Of more than 8000 lines of data, ultimately seven percent was unusable.
"That doesn't sound like a lot, but seven percent was 800 lines that we lost," Jannace says. "It was definitely a hard part of it, trying to figure out what data we could use."
Data integrity improvements figure into the team's final recommendations to Sector New York, including a flag or alert system to signal to a user when GPS coordinates have been entered that land outside the appropriate territory. Additional enhancement recommendations include strategic color-coding, custom graphs, and the separation of vessel categories (cruise skip, jet ski, ferry, etc.) to allow for more granular incident analyses.
The team also made system integration and business process recommendations to further reduce complication and to bring the prototype closer to that original vision of a tool that provides a near-live snapshot of incidents as they occur.
Although eight weeks was insufficient to apply predictive modeling to allow the dashboard to perform risk prioritization, the project is likely to see continued development next year, and positive feedback has been received even beyond its initial audience.
"The dashboard we created can be used anywhere," Jannace says. "We're specifically doing this for Sector New York, but there has been interest at the national level of the Coast Guard as well."
Gaming the way to improved border security
Imagine you want to smuggle contraband goods through an international border onboard a group of cargo ships. To avoid detection, would you concentrate the contraband together on one ship or distribute them across multiple ships?
Now imagine you're law enforcement searching for said contraband. Which ships would you search and how thoroughly? What strategies would you employ to maximize results while optimizing available resources?
The Red Team/Blue Team game project set out to answer just such questions.
Working with professor and associate dean of research Jeffrey Nickerson, the project team — Stevens software engineering undergraduate Liam Brew, Stevens biomedical engineering undergraduate Caroline Corr, Stevens engineering physics undergraduate Tanner Chiamprasert, Stevens computer science undergraduate Ronald Estevez, University of Puerto Rico at Mayaguez electrical engineering undergraduate Yousef Salaman Maclara, and New Jersey Institute of Technology computer science undergraduate Shreena Mehta — designed and developed a two-player adversarial computer game to mirror how the synthetic opioid fentanyl is illegally imported into the U.S.
"A lot of fentanyl is smuggled in through ships or cargo containers," says Mehta. "The Department of Homeland Security does their best to screen these, but the people that are doing the smuggling are getting crafty and creative with the ways that they hide and distribute the drugs."
To explore such hiding and seeking strategies and ascertain how they might evolve, the project team developed a game in which opponents play in groups of two, with one player assigned to the Red Team (the smugglers) and the other to the Blue Team (law enforcement). To begin each round, the Red Team secretly hides ten contraband units amongst ten numbered ships. The Blue Team must then decide which ships to search and how. Each player takes one turn in each round, for a total of five rounds.
Blue Team resources are not unlimited, however. Rather, the player must plan their strategy balancing the likelihood of success with the costs (game tokens) required to achieve it: one token to search 20 percent of one ship, three tokens to search 60 percent, and five tokens to search 100 percent.
By quantifying limitations as part of its parameters, the game simulates real-world scenarios and gets to the heart of the project's central question: how to allocate law enforcement resources in seemingly unpredictable situations.
"In real life, Customs and Border Patrol don't have unlimited resources, and neither do the drug smugglers. So we've put limitations on both sides," says Mehta. "You need to be able to gauge what you want to search and how much you want to search because, realistically speaking, you can't check every single spot of every single ship."
In-game teams weren't the only ones facing the limitations of reality, however.
Having nearly completed programming the basic game in the Python language, the project team came to realize that using JavaScript instead would allow them to progress faster and to make the game more easily accessible remotely through a web browser. So two weeks into the project, they rewrote all the code.
With the game now hosted online, game data would automatically be stored into a Firebase database and plotted onto graphs via Python scripts connected to the web application.
The team engaged approximately 60 friends and family members working in pairs to play the game, generating around 90 games' worth of usable data for the team to analyze.
Although the players' techniques lacked the benefit of advanced security training, Mehta feels even this limited data is valuable for drawing preliminary conclusions "at least human psychology-wise." With multiple playthroughs, even civilian players began to recognize patterns and respond accordingly.
"Generally, the Red Team (a.k.a the smugglers) had a larger amount of victories when using tactics of hiding all their contraband within one or two ships and leaving the others empty," Mehta says. "Blue Team took a while to catch onto this tactic, but when they did, distributing contraband as sparsely as possible proved to be Red Team's next best tactic. We also found that Blue Team had greater wins when they played multiple rounds with the same player on Red Team and were able to somewhat read that opposing player's strategies."
To improve the game's fidelity and better align it to DHS's needs, the project team has made numerous recommendations for going forward, including integrating in-field data, procedures, and intelligence not available to the students into the game's parameters. They’ve also recommended incorporating machine learning to improve both play speed and the quantity and quality of strategy analysis.
Ideally, it would evolve to become more like "an advanced form of a cat-and-mouse game," says Mehta.
Ultimately the Red Team/Blue Team game provides a framework through which larger adversarial strategy questions can be considered and tested beyond its original story line.
"Even though right now it is very maritime-oriented," Mehta says, "I think it could also be seen as an overall resource allocation game that could help in any field where resource allocation is needed."
Ten years and counting
Including the 2019 cohort, 187 students have now completed the SRI program from 30 universities.
"We're very proud of what we've accomplished," says MSC director of education Beth Austin-DeFares. "Not only is it the number of students we've brought through the program, but the outcomes from the research they've done has really had an impact in the field and on our stakeholders." Additional student projects offered this summer included the use of machine learning and neural network models to identify vessel traffic anomalies and the creation of a custom-designed, remotely operated underwater vehicle.
Manoussakis, Jannace, and Mehta all say they would consider returning to the SRI without hesitation.
"When you go into an internship, you're working with everyone that's in your industry, so [for example] you're only working with transportation engineers. But here I'm working with maritime, and computer science, and data scientists," says Jannace. "I loved my internships, but this is a totally different mindset. I'd definitely do one again."
The Maritime Security Center is a U.S. Department of Homeland Security Science and Technology Directorate Center of Excellence in port and maritime security.