Navigation

Research

For a list of my research group members, please click here. For the membership of RTML, which I direct, please click here.

We thank our research sponsors:

Current Research Thrusts

My research interests lie in all aspects of embedded real-time systems and wireless/sensor networks. In the domain of embedded real-time systems, my interests include but are not limited to operating systems, scheduling theory, resource management, wired/wireless networking protocols, quality of service management, hardware/software architecture, model-based design tools and power management. In the context of wireless/sensor networks, my research interests span hardware, devices, power-efficient networking protocols, run-time environments, large-scale system architectures, visualization and administrative tools.

RMA, Resource Kernels and QoS

A primary focus of my research is to build practical and functioning systems which can be analyzed/proved to be correct (in terms of timeliness, jitter, power efficiency, quality of service metrics, etc.). (If you are a practitioner, and you consider analyzable/provable and functioning systems as an oxymoron, I fully understand where you come from - let's talk!)

I was one of the principal contributors to Rate-Monotonic Analysis (RMA) (also see What Happened on Mars). RMA is supported by an impressive list of standards including POSIX Real-Time Extensions (IEEE 1003.1), the Real-Time Specification for Java, Real-Time UML (UML 2.0), Real-Time CORBA (CORBA 2.0), Ada 95, Ada 83, and automotive standards like OSEK and CANbus. I was the principal architect of tools like TimeWiz (from TimeSys), which supported schedulability analysis. I was the primary founder of TimeSys along with two other co-founders. (The company founded in 1996 now focuses on embedded Linux with an added emphasis on real-time features - thanks to my group's work on real-time extensions such as Linux/RK).

Current Projects

Our work on resource kernels abstracts the notions of RMA, EDF and other real-time scheduling policies while enforcing the assumptions these theories make. It also extended resource management from processors to network bandwidth, disk bandwidth and physical memory resources. The resource set abstraction of resource kernels turns out to be an ideal construct for managing and minimizing power consumption without violating the responsiveness characteristics of applications. We have lately started to extend this notion to distributed systems (using Distributed RK), and to multicore processors. Early publications are available.

I am a strong proponent of model-based design and development. Our tool, SysWeaver, embodies our approach and has been applied in multiple application domains including avionics, software radios, distributed automotive systems and sensor networks.

We then went a layer above and studied in depth tradeoffs across multiple applications each with multiple Quality of Service (QoS) dimensions using the same set of (finite) resources. This resulted in the QoS-based Model Allocation Framework (Q-RAM).

In recent years, my group has also spent a considerable amount of effort to build predictable wireless sensor networks. The FireFly sensor networks and our large-scale deployment of FireFly across the Carnegie Mellon campus (called Sensor Andrew) are concrete outcomes of this work. Also, please see FireFly Platform page.

Finally, thanks to our focus on real-time embedded and networked systems, we work with General Motors on Vehicular (V2V) Networks and on robust real-time platforms for autonomous driving. Yes, I serve as the Co-Director from Carnegie Mellon of the General Motors-Carnegie Mellon Information Technology Collaborative Research Laboratory and the General Motors-Carnegie Mellon Autonomous Driving Autonomous Driving Collaborative Research Laboratory.

My current research can be broadly classified into the following two categories. (Updated links will be available soon - meanwhile, please use the links provided above and below).