Computer engineering research
Research areas in computer engineering are:
- Embedded platforms and Systems design
- Positioning and navigation
- Application-specific processor (ASP) design
- Software-defined radio
- Video encoder architectures
- Wireless sensor networks
Embedded platforms and Systems design
Our research combines embedded processor architectures, Network-on-Chip, middleware for monitoring and controlling the platform, and is also moving towards parallelizing compilation onto such platforms and extension of existing operating systems for manycore architecture. The platforms are prototyped on FPGA, and may be complemented by application-specific accelerators in addition to programmable processors.
Moreover new design methods and prototyping tools are developed to enhance and accelerate embedded system development. We emphasize analyzing and comparing design choices, model-based design, fast iteration, and requirements management. The goal is achive seamless, automated flow from high-level model to FPGA prototype. Currently, research group concentrates on IP library management, handling legacy codes, as well as utilizing IP-XACT and MCAPI standards in design process. Kactus2 is an open source IP-XACT tool which also augments the standard concepts towards SW and HW/SW design and management. The tools and methods are utilized also in our teaching. Another open source tool is called Woke which is meant for documenting and analyzing design processes within a company. Previously our group has, among other things, developed an UML2.0-based design flow for MP-SoC devices (multiprocessor System-on-Chip). Research group home page: http://funbase.cs.tut.fi/
The old way to design (left) and the right way (right).
Positioning and navigation
Positioning and navigation group concentrates on GPS and GALILEO positioning receiver hardware and software architecture and implementations, use of Signals of Opportunity (such a WLAN) for positioning, and sensor-based localization methods. The group has gained a deep understanding of performance of micro-electro-mechanical-system (MEMS) sensor in various navigation applications. In addition, the group has introduced novel ways to use the GPS and sensor data in context-aware applications. See for example a demonstration video on motorcycle posture sensing. Research group home page: http://www.tkt.cs.tut.fi/research/nappo/index.htm
Indoor positioning becomes musch more accurate when pedestrian dead reckoning (PDR, green line) is accompanied with appropriate particle filtering (PF, blue line)
Application-specific processor (ASP) design
Research on application-specific processor (ASP) design methodology is based on Transport Triggered Architecture (TTA) processors and design tools for compiling and mapping applications. The research on ASP design methodology concentrates mostly on compiler code generation issues using LLVM framework, methods on fast generation of customized soft-core multiprocessors and processor architecture features leading to energy efficient designs. The tools are available as open source: http://tce.cs.tut.fi/
TTA toolset offers versatile design and analysis tools, such as simulators at HW level (left) and instruction level (right).
Software-defined radio implementations uses manycore, TTA and reconfigurable platforms. It deals with efficient partitioning and mapping of receiver baseband algorithms on programmable and reconfigurable processing elements, context switching, and cognitive radio protocols.
Video encoder architectures
Research on video encoder architectures has presented several parallel hardware architectures and software implementations for H.263/4, MPEG-4/AVC on over ten commercial platforms. Architectures make use of novel parallel memory architecture that solves the memory bandwidth bottleneck. A software framework for managing image processing pipeline reduces total memory requirements and reduces pipeline development time by months. Research group home page: http://ultravideo.cs.tut.fi/
Test video sequences are available in many format (1080p/4K, 8b/10b, AVC/HEVC, YUV etc.)
Wireless sensor networks
Wireless sensor networks (WSN) research area focuses on energy efficient multi-hop MAC and routing protocols, energy-optimized platforms, embedded software architectures, and prototype implementations. Performance and reliability of WSN technologies are developed through large-scale application pilot studies. For novel applications, automated processing of data reserves and real-time measurements are developed for the various areas of applications containing measuring and controlling, positioning and distributed data processing on distributed server infrastructure.
The concept of wireless sensor network: nodes, network, application server and user interface.