My core area of research is in building computer architectures that enable unique ways for improving compute efficiency, network performance and provide reactive capabilities to adapt to changing environments, through seamless interaction of software and hardware. Such architectures can then be tailored to different domains such as automotive embedded systems, cognitive radio systems and internet of things. Reconfigurable hardware is a key enabler for my research.
Automotive Embedded Systems
Automobiles are mobile computing devices, with multiple funtions like engine control, steering, stability control and others, implemented on distributed array of embedded modules, commonly known as Electronic Control Units (ECUs). A modern high-end vehicle incorporates some 100 ECUs, which perform both safety-critical and comfort functions, and this number is rising rapidly. The rising number increases power consumption, space and weight, while also affecting reliability and determinism of the critical systems. Our research aims at building architectures for next generation electric vehicles, which aims to address these challenges through the benefits of reconfigurable hardware. We look at alternate techniques for function consolidation and fault-tolerance while also looking at system level validation and optimisations at the architectural level.
Vehicles incorporate multiple networks like the Control Area Network (CAN), Local Interconnect Network (LIN) and Media Oriented Systems Transport (MOST) to support the requirements for the variety of embedded systems on-board. With increasing demand for bandwidth and reliability, these networks are being gradually replaced by emerging time-triggered schemes like FlexRay and Synchronous Ethernet. Through our research, we look at network layer enhancements and extensions that aims to improve the overall communication network, determinism and security. We look at custom network controllers that support extended features, cross layer enhancements and reconfigurability of the platform to achieve these.
Security Infrastructure for Automotive Systems
Increasing connectivity (Wireless) in vehicles add new functionality to vehicles for comfort and safety. With techiques like Vehicle to Vehicle Communication (V2VC) and Connected Services (like GM's OnStar, BMW's Connected Drive) becoming more popular, security of critical embedded systems in vehicles are a rising concern. Researchers have already identified security flaws with existing vehicles and devised methods to remotely hack connected vehicles by exploiting their wireless interfaces. We look at techniques that address these challenges and build an infrastructure that enables multiple levels of security for critical embedded systems and the messages exchanged on critical networks, without increasing the computational load on the ECUs.
Dynamic Cognitive Radios
Available radio spectrum for communication is limited and has to be utilised intelligently. Many spectrum bands that are licensed to specific users (example TV bands, GSM bands) have utilisation that vary with time and could be seriously underutilised. Cognitive Radios make use of such unused spectrum in a dynamic manner: they sense the available spectrum and uses bands that are sparingly used for communicating with other users. This requires adaptive RF and baseband capabilities, which are usually implemented in software running on PC (see GNURadio). This limits the portability of such systems, making them less attractive for mass adoption. This research focusses on compact Cognitive Radio architectures with hardware-layer adaptability that is tightly coupled to a software-based cognitive engine. We explore the possibilities of utilising the hybrid FPGA platforms with closely coupled software control with abstract mechanisms to adapt the baseband modules through high-speed reconfiguration. We also explore the application of such platforms in flexible IoT architectures and upcoming avionics standards for efficient air-to-ground communication.