A.I. driven adaptive cameras that can be a game-changer for applications ranging from autonomous delivery drones to privacy preserving sensors.
Using neural networks to directly control LIDAR sampling that can revolutionize perception tasks for robots and autonomous systems.
In our current Science of Autonomy grant, we linked a hierarchical vision system with a navigation controller to improve navigation of autonomous underwater vehicles (AUVs) in docking maneuvers. Our inspiration to design the object recognition system was the human visual system , because organisms have evolved to survive in an unpredictable world, and so unlocking the principles they use to roam the world has proven to be quite productive. Our multidisciplinary team successfully proposed new paradigms to design the vision system architecture to detect objects in video  and link it directly to the navigation controller as a nested feedback system .
Aerial and marine robots operate in a fluid medium which carries significant information not often utilized in path planning and cooperation of such robots. Motivated by this observation, we believe that modeling the macroscopic dynamics of a multi-agent system using the vast and historically grounded knowledge of fluid dynamics will open up many opportunities for a better understanding and subsequent enhancement of the behavior of such multi-agent systems. In addition, by including the dynamics of background flows, significant improvement in mobility, reachability and success rate can be expected in applications such as dynamical target tracking, foreign threat interception, and large-scale environmental sensing. Starting with the consideration of the underlying background flows and the sensor network as two tightly coupled dynamical systems, we leverage the fluid dynamical description of a system to understand the macroscopic motions of the sensor network as a continuum. We present designs for cooperative control strategies governing both local interactions within the multiagent systems and the emergent swarm behaviors. The proposed fluid-based swarm modeling and control scheme is well-suited for aerial or marine missions including multi-vehicle transportation, cooperative surveillance, collaborative target tracking and capturing, etc. More specifically, we will be focusing on the following research tasks:
-stability analysis and adaptive control of multi-agent systems modeled as continuous fluids;
-multi-agent system dynamics coupled with background ow dynamics;
-experimental evaluation with mobile aerial vehicle swarms
The design and development of highly-maneuverable aircraft has been a long-standing engineering challenge. This challenge presents itself in almost all flight regimes, from supersonic fighter jets down to low-speed, smaller-scale unmanned aircraft. Despite the very different operating conditions, there is at least one common feature among aircraft designs aiming to provide very agile, yet stable, flight. Namely, the planforms (the shape and layout of an airplane’s wing) of such aircraft are of low aspect ratio. Recent studies suggest that the aerodynamic and gust-response of such low aspect ratio fliers are significantly different than larger flyers and not well understood. This issue constitutes a critical gap in aerial vehicle development, and this research project addresses a critical gap in the development of reliable and fully controllable aerial drones. The researchers will also enhance course curricula with results from this research, and a course on unsteady low Reynolds number aerodynamics will be developed. A summer program is proposed that will enable local high school students to learn aerodynamics and flight concepts and to participate in a design/build/operate competition integrating fluid dynamics, aerodynamics, and aircraft design.
This project will create new control methods to maximize the effectiveness of a commonly prescribed rehabilitation therapy for individuals with neurological conditions (NCs), including stroke, spinal cord injury, and traumatic brain injury. Functional Electrical Stimulation (FES) cycling uses an externally applied sequence of voltages to cause the individual’s leg muscles to contract to propel a recumbent cycle. The repetitive, coordinated motions of cycling can help restore limb function. An electric motor is available to augment the person’s own muscles, if needed. This project will determine how to switch between different muscle groups and the motor to ensure desired behaviors, despite differences in muscle strength and endurance between individuals. For example, the project will examine methods to enable an FES cycle to adapt to the individual attributes of a new participant within a known time interval. The results will be validated in populations of individuals with NCs to demonstrate clinical efficacy. This project will advance the national health by improving the quality of life for individuals with NCs. In addition to these direct benefits, the project will introduce top undergraduate students to advanced research methods in this critical area of biomechanics.
Cameras are pervasively used for surveillance and monitoring applications and can capture a substantial amount of image data. The processing of this data, however, is either performed a posteriori or at powerful backend servers. While a posteriori and non-real-time video analysis may be sufficient for certain groups of applications, it does not suffice for applications such as autonomous navigation in complex environments, or hyper spectral image analysis using cameras on drones, that require near real-time video and image analysis, sometimes under SWAP (Size Weight and Power) constraints.