Adaptive Scheduling, Planning & Control of Autonomous Vehicles
The goal of this project is to develop techniques to schedule and plan unmanned vehicles operations to minimize energy and computation while guaranteeing safety (i.e. something bad will never happen) and liveness (i.e., something good will eventually happen) properties.
Modern unmanned vehicles such as aerial and ground vehicles rely on constant periodic sensor measurements to detect and avoid obstacles. However, constant checking and replanning are time and energy consuming and are often not necessary especially in situations in which the vehicle can safely fly in uncluttered environments without entering unsafe states. This project researches self-triggered techniques that leverage reachability analysis, control and motion planning techniques, and risk-based analysis to schedule the next time to check sensor measurements and perform replanning while guaranteeing safety under noise and disturbance effects.
Attack Resilient Control of Autonomous Cyber-Physical Systems
Modern cyber-physical systems are not build with security in mind. The tight interaction between information technology and the physical world, coupled with system design complexity and the need for a more open control architecture, have introduced security vulnerabilities in the form of physical and cyber-attacks (e.g., sensor, actuator, controller, communication and environment attacks). Examples of such vulnerable CPS include modern automobiles, unmanned aerial and ground vehicles, vessels, medical devices, power plants, and smart buildings.
This project is centered on developing resilient techniques to detect cyber-atatcks, estimate correctly the state, and recover/reconfigure the system to guarantee safety and performance.
Modern robotic systems leverage the growing advances in control, computer vision, and machine learning coupled with technological advances in sensing, computation, and communication to achieve complex autonomous operations often in unknown environments. The lack of understanding of when and how machine learning works makes it hard to provide guarantees for such systems during safety- critical operations.
This project focuses on the development of techniques to provide safety assurance at run-time as robotic systems are deployed to perform complex tasks in which learning is required due to uncertainities
Mapping, Localization, & Environmental Monitoring
One of the distinguish features of autonomous vehicles like UGVs and UAVs is that they can perform operations with minimal human supervisions in environments that can be hazardous for human beings. Search and rescue, mapping, surveillance, and monitoring are some of the best examples of operations in which robotics have great advantage over other technologies.
This project focuses on the development of techniques to enable more efficient mapping, sensing, and planning of autonomous ground and aerial vehicles to maximize data diversity. Some of our recent experiments on this topic include the deployment of autonomous vehicles to explore an hazardous historic tunnel in Virginia, the mapping of buildings and bridges, and monitoring of environmental states like temperature, humidity, light, and noise inside working environments.
Coordination of Heterogeneous Robotic Systems
Heterogeneous multi-robot systems, characterized by members with diverse dynamics, sensing, computation, and energy characteristics, offer a variety of benefits over teams composed of homogeneous members. Appropriate coordination of these systems can combine a team’s differing capabilities and enhance a mission by extending individual agent limitations and optimizing performance. However, cooperative efforts may also lead to multi-agent failures. An aerial vehicle dependent on a ground vehicle for power recharging, for example, will fail if the ground vehicle fails.
This project focuses on the development of online scalable scheduling and planning techniques to deploy heterogeneous systems to prioritize safety while accomplishing different tasks.
Modern cyber-physical systems like autonomous cars include a strong human in/on the loop component. The human needs and preferences need to be taken into account when designing such modern systems. Safety, trust, decision making, autonomy, user interface, user emotions, and security are some of the problems to consider when dealing with human-machine interactions.
This project focuses on the development of techniques to close the loop between human and robot interactions to guarantee safety.
Co-design and Rapid Prototyping of autonomous mobile robots
In the near future3-D robotic systems can be produced and designed using 2-D desktop technology fabrication methods. If this feat is achieved, it would be possible for the average person to design, customize and print a specialized robot in a matter of hours. Currently, it takes years and many resources to produce, program and design a functioning robot.
This project focuses on automate the process, from sketches on-demand, anywhere, and with the skill of a team of professional engineers, leading to potential transformations in advanced manufacturing. The project addresses broad classes of constructible cyber-physical systems: the development of tools for functional specification and automated co-design of the mechanical, electrical, computing, and software aspects of the device; the design of planning and control algorithms for the assembly of the device and for delivering the desired function of behavior, and tools for the analysis of these algorithms that take into account all the necessary resources, including actuators, sensors and data streams from the world; the methodology to generate device-specific and task-specific programming environments that provide safeguards for programs written by non-expert users to enable them to operate the machines safely; and the development of novel approaches to the automated production of new devices which may be based on the synthesis of programmable materials with customizable electrical or mechanical properties. This research is highly multidisciplinary, primarily leveraging the disciplines of computer science, electrical and mechanical engineering, materials and manufacturing science.