This site will work and look better in a browser that supports web standards, but it is accessible to any browser or Internet device.

Home | Bio | RESEARCH | Publications | AVIA | Dragonfly | Green Flight Challenge
The research that my students and I conduct has focused on improving the capabilities of autonomous vehicles by developing algorithms and system designs that address the problem of perception and persistence. Our research focuses on real-world operation: we fly what we design. We have developed and flown:
  • algorithms for energy-optimal flight path planning
  • methods to enable mission-integrated atmospheric modeling and flight planning using only in-situ gathered data
  • algorithms that enable coordinated transport of payloads by a team of autonomous rotorcraft
  • vision-based systems for autonomous landing
To enable this research we have also developed design tools for conceptual and preliminary design of both fixed-wing and rotary-wing flight vehicles. Short summaries are given below. For more details please contact me or see the AVIA Lab's Publications page.

Vehicle design and prototyping

Some of our lab's unmanned aircraft are based on commercially available airframes that we modify, adding an autopilot module and flight computer. However, in many cases there are no suitable commercially available airframes, and we design, build, and flight test new aircraft from blank page to successful flight.

We use a combination of free analysis and design tools (XFoil, AVL, and OpenVSP) and in-lab developed tools for vehicle design and performance prediction.

In all cases we develop aircraft and algorithms that are self contained: all autonomy functions are hosted on single-board computers carried on-board the aircraft, with telemetry allowing us to monitor the system's decision making and performance. Flight systems are designed so that the operator can step in at any level of the autonomy system: we can take over low-level control to fly the aircraft home in case of a severe failure, or we can intervene at a high level to give the aircraft a specific instruction.

Enabling new science

We are part of a team working to send a drone to Saturn's moon Titan. This aircraft (called Dragonfly) will conduct surface science operations at various geological settings on Titan by flying from one site to the next: the combination of low gravity and high atmospheric density makes Titan a great place to fly! We are working on vehicle design, performance, and atmospheric flight planning. We have built a half-scale demonstrator to validate performance, navigation, and flight control algorithms. More information about Dragonfly is available at the Johns Hopkins Applied Physics Lab page:

We are also working with researchers in Penn State's Geosciences department to study the Helheim Glacier in Greenland. We have built a drone that has flown over the glacier, gathering data from sensors placed on the ice.

Persistence: autonomous soaring flight

Small autonomous vehicles are typically limited in the amount of on-board energy storage and in the sensing payload that can be carried. As a result a fundamental trade-off is often made between sensing (to maximize the information gathered on a mission) and fuel (to maximize mission duration). Additionally, small uninhabited aerial vehicles (UAVs) suffer from reduced performance due to generally low operating Reynolds number, which makes design of efficient (i.e. high L/D) aircraft difficult.

In the case of UAVs significant energy is available from the atmosphere. Large birds such as eagles, hawks and vultures routinely employ soaring flight to remain aloft for many hours with flapping wings (and of course hang glider and sailplane pilots also soar). We have developed algorithms that improve flight duration by an order of magnitude by exploiting vertical air motion. All flight algorithms run on a single-board computer carried on board our vehicle, called AutoSOAR.

Uncertainty: atmospherically-aware flight

We have developed atmospheric modeling and flight planning algorithms that balance gathering information and conserving energy while flying a mission. These algorithms can be run using only measurements and computational resources available onboard the aircraft, enabling a small UAS to efficiently conduct long range missions in remote areas with no a priori knowledge of the environment.

Coordination: cooperative transport of a slung load

We have developed methods to enable load-leading control for a team of small rotorcraft cooperatively carrying a large payload. Rather than treat the load as a disturbance acting upon the formation, the load controls the formation by determining the cable forces (tension and cable direction) that will enable trajectory following. Each vehicle in the formation is then responsible for controlling its own cable tension. A human operator can thus effectively operate a large formation by controlling the payload.

Perception: autonomous landing and collision avoidance using vision

We have developed methods that combine computer vision, state estimation, and flight planning to enable autonomous landing on pitching, heaving, rolling ship decks. Laboratory flight demonstrations have resulted in 80 safe landings out of 80 attempts on both stationary and moving decks. As with AutoSOAR, all flight, estimation, and planning algorithms are hosted on the vehicle.

This autonomous landing research grew out of my Ph.D. work, which focused on collision avoidance and GPS-denied navigation of a robotic aircraft flying among obstacles.

Home | Bio | RESEARCH | Publications | AVIA | Dragonfly | Green Flight Challenge