Biomedical control

We develop control systems for medical applications, to improve quality and safety of care through automation. Our work is strongly interdisciplinary and harnesses the strength of model-based automatic control and machine learning, two fields that play a vitally important role in engineering but that tend to have different goals and tools if taken separately. Working at the intersection of these two fields makes it possible to personalize treatments and to adapt modeling, data analysis and automation of complex biomedical systems to real-life scenarios in order to improve health outcomes and minimize risks and harms to patients.

Type 1 diabetes (T1D) is a prime candidate for the development of a treatment strategy using biomedical control. In this disease, the auto-immune destruction of the beta cells in the pancreas prevent the body from producing insulin, a hormone that is required for the glucose homeostasis feedback loop. The current standard of treatment is for people with T1D to measure their blood glucose concentration several times per day and manually deliver corresponding doses of insulin.

In our lab, we aim at artificially recreating the glucose control feedback loop using a combination of medical devices, wearable sensors and a smartphone, to realize a fully automated insulin delivery system.

 

Ongoing Projects:
  • Data-Enabled Neural Multi-Step Predictive Control (DeMuSPc): a Learning-Based Predictive and Adaptive Control Approach for Complex Nonlinear Systems (Founding source: NSFpress release)
  • First-Principles Informed Data-Enabled Predictive Digital Twin of Human Physiology (Founding source: NSFpress release) [overview video]
  • Learning blood glucose predictors with artificial neural networks for decision support systems [overview video] and closed-loop glucose control [overview video]
  • Decision support  for insulin dosage optimization in multiple-daily-injections therapy for people with type 1 diabetes
  • Development of personalized insulin infusion failure detection algorithms combining tissue counter pressure and blood glucose data for closed-loop diabetes management, in collaboration with Diatech Diabetes (Founding source: NIH R43DK130036press releaseJDRF IDDP)

Safe autonomy for cyber physical systems (CPS) in open environments

Cyber-physical systems (CPS) integrate physical processes with communication and control, and interact with the physical world via actuators and sensors. Over the last few years, there have been intense efforts to increase the capability, adaptability, scalability, resiliency, security, and usability of CPS. As a result, today, CPS control vehicle, aeronautical and aerospace systems, inspect critical infrastructures, monitor energy resources, and guide medical interventions, among many other applications. As CPS are increasingly being expected to operate autonomously in critical tasks and missions, in poorly understood environments and often in the vicinity of humans and other intelligent systems, assuring safety of their operations becomes a top priority. Examples of safety-critical CPS include military and aviation systems and drones, autonomous vehicles, smart networked systems, robotic systems interacting with humans, smart medical devices, and many others. Failure to ensure safe operations may result in an unacceptable loss, such as loss of control, physical damage to the system under control, loss of human life, human injury, property damage, environmental pollution, and failure of the mission. In our lab, we aim at improving the ability of unmanned vehicles operating in complex, possibly contested open environments to effectively react to new contexts, unexpected disturbances, and system malfunctions, while ensuring safety at all time.We validate our control architectures using various types of CPS in the flight arena of our lab.

Ongoing Projects:
  • Run time assurance (RTA) approaches for safe control of multi rotors UAVs in adverse and hazardous conditions (Founding source: Air Force Research Lab – AFRL)[AFRL ML-RCP Fall 23 Newsletter , student engagement]
  • Safe kynodynamic motion planning for CPS
  • Development of novel algorithms for collaborative satellite autonomy
  • Adaptive Space Systems (Founding source: NASA MIRO Inflatable Deployable Environments and Adaptive Space Systems (IDEAS2) Center, press release)

Superconductor manufacturing

We perform research on process control and machine-learning methods for real-time data analysis to achieve high manufacturing yield for superconductor or metal additive manufacturing projects at the Advanced Manufacturing Institute at the University of Houston.

Metal Organic Chemical Vapor Deposition (MOCVD) offers great potential for high yield rapid fabrication of High Temperature Superconductor (HTS) tapes. The MOCVD process is able to produce high quality HTS tapes allowing large deposition area. In this process, semiconductor precursors in gas phase thermally decompose upon contact with a heated substrate forming a superconductor film. MOCVD is a versatile process allowing large scale production of HTC coated conductors (CCs). However, superconducting tapes manufactured with conventional MOCVD suffer from large variations in the current critical density Jc due to potentially significant variability in the processing conditions of the MOCVD reactor that occur over time. The MOCVD process is sensitive to these variations resulting in structural and chemical compositional disparities along the length of the tape. A novel advanced reel-to-reel MOCVD (AMOCVD) system that has been developed at UH is able to scale up production of moving high quality long thick REBCO tapes meeting higher critical current density Jc requirements.  Manufacturing of high-quality REBCO tapes of large thickness at scale with excellent homogeneous magnetic field properties based on desirable crystalline orientation and film morphology remains a challenge.

Ongoing Projects:
  • Real-time learning and process control for high performance advanced MOVCD system