Virginia Center for Autonomous Systems

Making UAVs more robust by mixing up the control development process

Related stories

A VaCAS team is designing and implementing control algorithms to improve the resilience of unmanned aerial vehicles (UAVs). Mazen Farhood of aerospace and ocean engineering and Sandeep Shukla of electrical and computer engineering have added a new twist: they are designing the control algorithms and implementation software at the same time.

Farhood and Shukla are designing and implementing the algorithms concurrently, in a way similar to hardware-software codesign in computer engineering. They have already built a UAV platform for advanced control development, and are in the process of testing their designs.

Three researchers with a UAV

From left: Graduate student Ony Arifianto, Mazen Farhood, and Sandeep Shukla are using a Telemaster UAV to test their control systems.

“Currently, in academia as well as industry, the collaboration between control and computer engineers occurs late in the process, after the control algorithms are designed and transformed into software,” Shukla notes. As a result, both sides lose the opportunity to experiment with each other’s most recent research and ultimately leverage the newest innovations to create optimized systems.

Their motivation is the interest by the U.S. Air Force to deploy expendable, low-cost UAVs instead of large and expensive aircraft. Even with less space, lower power, and a reduced budget, these small airplanes must be as reliable as their larger counterparts. Consequently, an optimal use of the resources, according to some predefined measure, is necessary to ensure a reliable UAV system that can achieve the desired objectives. “We plan to accomplish this optimal behavior by designing the software and hardware parts of the UAV jointly,” Farhood states.

The VaCAS team is particularly interested in making the control system more resilient to disturbances by wind and moving obstacles, and more secure against malicious cyber-attacks. According to Farhood, their initial interest was to design control systems that operate reliably in spite of various random and/or independent uncertainties and exogenous disturbances. Then they decided to leverage their understanding of the dynamics of UAVs to devise intrusion detection algorithms based on the physical system, which complement Shukla’s tools for securing software against cyber attacks.

Farhood and Shukla are using event-driven computational models, allowing the software platform to interact with software components asynchronously. “Designing software components assuming a fully synchronous environment is easier for the software designer,” explains Shukla, “but then the platform designer has to implement time synchronization across the platform— which often leads to a nonoptimal system.” Control algorithms are also usually timedriven, and this new model allows Farhood to explore new methods.

By working together throughout the entire project, Farhood and Shukla hope to create an experimental environment for integrated design of cyber physical systems in general and UAVs in particular. According to Farhood, knowledge of Shukla’s software allows him to exploit certain optimization opportunities that he wouldn’t have known about otherwise. “We end up with an integrated approach that results in optimized control software,” he says.

Seed funding for the project came from an ICTAS program to encourage collaboration between junior and senior faculty. The work is also funded by the Office of the Secretary of Defense’s software producibility initiative and the Air Force Research Labs in Rome, N.Y. The team is collaborating with researchers from INRIA in France and University of Kaiserslautern in Germany.