image: Lehigh University’s Cristian-Ioan Vasile has earned an NSF CAREER Award to make robots more predictable and reliable. His research will help improve how autonomous systems, from drones to delivery robots, learn, adapt, and work together in complex environments.
Credit: Courtesy of Lehigh University
Self-driving vehicles, drones, and robotic assistants are transforming industries including transportation, logistics, and health care. With new developments in hardware, AI, and machine learning, these autonomous agents can sense their surroundings with greater accuracy, understand complex environments, and engage in sophisticated reasoning.
But despite such advancements, deploying robots in dynamic, real-world settings—and getting them to do what we want—remains difficult.
“The overarching problem deals with robot capabilities,” says Cristian-Ioan Vasile, an assistant professor of mechanical engineering and mechanics in Lehigh University’s P.C. Rossin College of Engineering and Applied Science and a faculty member in Lehigh’s Autonomous and Intelligent Robotics (AIR) Lab. “When you have complex interactions between new hardware and new software that comes from learning-based approaches, we now have new problems regarding how to use these components, like how to assign them tasks. And so how do we characterize these software and hardware systems in a way that we can integrate them into traditional planning workflows?”
Vasile recently received funding through the National Science Foundation’s Faculty Early Career Development (CAREER) Program to develop structured methods for assessing the capabilities of learning-enabled agents and use that information to improve planning and coordination of robots working in teams.
The prestigious NSF CAREER award is given annually to junior faculty members across the U.S. who exemplify the role of teacher-scholars through outstanding research, excellent education, and the integration of education and research. Each award provides stable support for a five-year period.
On the software side, “a lot of machine learning algorithms are very opaque,” says Vasile. “Even though they work well, we don’t know why or how exactly they do.”
That means it’s unclear how well robots using the software will perform outside of training environments in varying real-world conditions. Such uncertainty limits widespread robotic deployments across industries. For example, in some hospitals, robots are used for simple tasks like delivering medication, but with limited autonomy.
“They’re essentially just autonomous carts because we don’t know whether they’re going to do the correct thing,” says Vasile. “What if they go into a patient’s room and deliver the wrong medicine?”
Vasile’s research will focus on how to map and model an agent’s capabilities—particularly those involved in motion, manipulation, and perception such as navigating crowded hallways, checking on patients’ status, and cleaning up spills—and develop tools to reliably predict their behavior. The goal is to use that understanding to plan effectively for large teams of agents. This means understanding how performance varies depending on time, location, and situational context.
“It’s not just about whether a robot has a camera or an arm,” he says. “We want to know if it can still operate in the dark, in a crowded space, or navigate a narrow corridor.”
His approach involves three main tasks. First, the team is developing a formal framework to describe and learn an agent’s “capability profile,” which links hardware, software, and context to performance metrics like energy consumption or task completion time. For example, if a robot is operating in a grocery store, how well it can manipulate stocking shelves or navigating crowded aisles.
Engineers will be able to see how an agent’s performance is influenced by time, location, and context—and why it’s suited (or not) for a given task. “The key here is that these profiles will be interpretable even if they are complex,” he says.
The second task is to design planning methods that move beyond binary assumptions of capability. “Up until now, capabilities have been treated as either yes or no,” he says. “But we’re creating a spectrum of performance using this additional, rich context information, like whether a robot is operating in a kitchen at noon or a library at night.”
The third task involves detecting—and quickly recovering from—failures. If an agent is misassigned or underperforms, the system needs to recognize the mismatch and adapt. Such dynamic reassessment is essential for safe, large-scale deployment of autonomous systems.
The ultimate goal, says Vasile, is to enable widespread use of robots that are both efficient, and effective. Not, as some fear, to replace humans in the workforce.
“A lot of the jobs that robots could do are ones in which there are currently worker shortages, or are dangerous or hazardous to human health,” he says. “And in countries where the demographics are shifting and birth rates are declining, robots could potentially perform the physical, strenuous work that can free older adults to do more creative work.”
About Cristian-Ioan Vasile
Cristian-Ioan Vasile is an assistant professor in the Department of Mechanical Engineering and Mechanics at Lehigh University. His research interests include formal methods, motion and path planning, distributed and decentralized control, machine learning with applications to robotics, and networked systems and systems biology.
Prior to joining the Rossin College faculty in 2019, Vasile was a postdoctoral associate with Sertac Karaman and Daniela Rus in the Laboratory for Information and Decision Systems (LIDS) and the Distributed Robotics Laboratory, Computer Science and Artificial Intelligence Laboratory (CSAIL) at Massachusetts Institute of Technology (MIT). He obtained his PhD in 2016 from the Division of Systems Engineering at Boston University, where he worked with Calin Belta in the Hybrid and Networked Systems (HyNeSs) Group of the BU Robotics Laboratory. He earned his BS in computer science in 2009, an M.Eng. in intelligent control systems in 2011, and a second PhD in Systems Engineering in 2015, all from the Faculty of Automatic Control and Computers, Politehnica University of Bucharest.
Related Links
- Rossin College Faculty Profile: Cristian-Ioan Vasile
- Lehigh University: Autonomous and Intelligent Robotics (AIR) Lab
- NSF Award Abstract (2442644): "CAREER: Capability Inference and Planning for Teams of Learning-Enabled Agents"
- Lehigh University: Institute for Data, Intelligent Systems, and Computation (I-DISC)