For most college seniors, the final months before graduation involve job interviews or deciding whether to go to graduate school. It typically doesn’t involve meeting with the U.S. Air Force to discuss design changes for an unmanned outdoor security robot.
But that’s the case for three seniors from Worcester Polytechnic Institute (WPI), who plan on meeting with officials from the Air Force Research Lab (AFRL) later this spring to discuss their security robot prototype, nicknamed “Cerbi.”
The Computer Enabled Robotic Base Enhancing Remote Unmanned Security (CERBERUS) project is a mobile, all-terrain Trackchair modified with additional cameras and sensors. The unmanned system can monitor outdoor locations and provide security alerts. The project was the result of an AFRL challenge for teams to create an autonomous robot that could protect places, like a missile silo area, where no humans were working.
The team of three undergraduates from WPI created “Cerbi” as part of their Major Qualifying Project (MQP), a long-term project that is required for them to graduate. Unlike regular coursework for robotics that completes projects within a seven-week term, the MQP can be a multi-year project.
Students work with WPI advisors and outside researchers — in this case, the AFRL and members of the Ohio Aerospace Institute, which runs the challenge for the Air Force — in the project planning, budgeting, and production stages.
“In classes, you’re doing project work with professors,” said Jeff Tolbert, one of the team members. “This [MQP] is an example where you have to work with advisors, but also an outside sponsor.”
The longer term and collaboration required for the security robot project differentiate it from the student’s usual work, Tolbert said. The team’s other two members are Marissa Bennett, a major in robotics engineering and a minor in computer science, and Ken Quartuccio, a robotics engineer major.
Security robot development process
After hearing about the security robot challenge, the three began writing a proposal and working with their advisors, Kenneth Stafford, a teaching professor at WPI and director of the university’s Robotics Resource Center; and Alexander Wyglinksi, an associate professor of electrical and computer engineering. After being accepted by the AFRL, the team received some parts from Trackchair, and then began making modifications to their design.
Another aspect of the challenge was to work within a set budget, a $10,000 limit. “The Air Force instructed us to be competitive with our pricing,” said Tolbert. “We treated it just like in industry where you have to bid for a job, and went through our plan several times looking for places to cut costs. We will likely come in under our initial estimate, and we are pretty proud of that.” The final cost of Cerbi is about $8,000, the team said.
The robot is built with sensors, electronics, and video cameras that help drive its autonomous navigation. The goal of the challenge is when an alert is received, the security robot undocks itself from a charging base station and autonomously navigates to the location of the alert, avoiding obstacles. Cerbi is also designed to work in varying weather conditions and both in daytime and at night.
Once at the alert location, a remote operator can access the security robot’s video feed and controls to determine the nature of the alert and decide whether to send human personnel or disable the alert in the case of a false alarm.
The team built Cerbi with the Robotic Operating System, devoting a lot of research to picking sensors that already had ROS support.
“We spent a lot of time designing a charger that would interact seamlessly with our project goals,” Tolbert said. “It had to be durable enough to survive in the same climate as the robot and be able to attach no matter how dirty the robot was.”
The final design features two plates that are a fixed distance from each other and able to move linearly as well as pivot from side to side. Fixed magnets attach to plates on the robot, with an electrical system that can detect when the robot is disconnected to the charger in order to remove the voltage differential across the plates.
The big challenge: Autonomy
Making the security robot autonomous is one of the biggest challenges for the team.
“You don’t want somebody on this robot full time, controlling it,” said Quartuccio. “You want the robot to be able to handle most of the tasks, and only have human interaction when it’s absolutely necessary. That’s probably the biggest core challenge that we face.”
Communication between the robot and the ground station has also been challenging, the team said. Different programming languages and communicating with network latency and other bugs have been big hurdles, they added.
The three also needed to think about a range of weather conditions when designing the robot.
“Our advisors made us consider installations in North Dakota, where it can go from -40 degrees to 100 degrees over the course of a day,” Bennett said. “If we just had to accommodate for one climate we would be able to simplify our design dramatically, but it wouldn’t be a design that would be desirable in the real world.”
The long-term project also gave the team opportunity to experience design changes based on feedback from the outside groups or end users.
“In our initial concept, we had planned on the teleoperation to be done with a joystick,” Bennett said. “However, former members of Air Mobility Command recommended using commercial gamepad controllers, like the Xbox 360 controller. When we brought these changes back to our contacts at the AFRL, they were ecstatic.”
Post-graduation plans vary
After the team meets with the Air Force, the future of Cerbi is unclear. All three team members are graduating, with only Bennett staying to pursue a graduate degree. The project could continue with other students, which could complete the work and take the security robot from a prototype to something that looks more like a finished project, team members said.
As per university policy, all of the project work is openly published for other researchers to review, Stafford said. Commercialization of the robot could eventually occur, with the Air Force and WPI sharing intellectual property rights.
For now, working on the MQP has been challenging and satisfying for the team members, giving them real-world experience outside of the classroom.
“[With the MQP], we can go more in depth and try to solidify the design and really bring in all three aspects of robotics – computer science, electrical and computer engineering, and mechanical engineering, to create this platform, which I find really cool,” Bennett said.