Research in the Microelectronics Design Lab focuses on VLSI circuits and systems, especially vision sensors, novel computer architectures, cellular processor arrays, analogue and mixed-mode signal processing, brain-inspired systems and chip design for biomedical electronics.
Below is a brief overview of major research areas, further information, and the list of current projects can be found on the Microelectronics Design Lab website.
For further details and research opportunities please contact: Dr Piotr Dudek
Vision chips are microelectronic devices, which combine image sensing and processing on a single silicon die. In a way somewhat resembling the vertebrate retina, these semiconductor chips perform preliminary image processing directly on the sensory plane. They can be used for computer vision applications in areas such as autonomous vehicle guidance, robotics, industrial inspection or surveillance. We investigate the design of vision chips, using CMOS technologies. Integrating a processing element (PE) within each pixel of the image sensor array results in thousands of processors working concurrently, which enables the processing speeds of billions of operations per second to be easily achieved, at very low power consumption.
Cellular processors arrays offer high computing performance and are ideally suited to perform pixel-parallel image processing tasks. Our research focuses on computer architectures based on fine-grain parallel processor arrays and asynchronous cellular automata. We have developed several approaches, e.g. SCAMP is family of mixed-mode pixel-per-processor SIMD array devices, ASPA devices are massively parallel processor arrays which work in asynchronous/synchronous mode running wave-propagating algorithms, yet another approach investigates fine-grained MIMD processor arrays. We implement integrated circuits in state-of the-art and emerging silicon technologies, such as sub-100nm CMOS and stacked 3D wafers.
Brain-inspired VLSI circuits may one day replace conventional microprocessors as more robust and intelligent systems. They will be also used as controllers for autonomous robots, and in prosthetic devices. We research digital FPGA based, and analogue full-custom VLSI based acceleration engines for neural computation. We work with neuroscientists, psychologists and computer scientists, investigating computation in biological neural networks and designing neuromorphic chips and systems implementing natural computation in hardware.
Analogue signal processing circuits can offer higher efficiency (in terms of performance, silicon area and power dissipation) than their digital counterparts. We investigate algorithmically programmable general-purpose analogue microprocessors build using switched-current circuit techniques, and analogue implementations of probabilistic inference in Bayesian networks. We are interested in ultra-low power computing hardware. We are also developing sensors, sensor interface circuits and signal processing circuits for biomedical applications, such as high-frequency medical ultrasound systems, and wearable ECG sensors.
Novel algorithms are developed to take full advantage of the massively parallel hardware. We investigate ways of efficiently mapping low- and mid-level image processing algorithms onto processor arrays. We are interested in computer vision and machine intelligence, especially in the context of autonomous robots. We work on brain-inspired algorithms for object recognition and efficient simulation of large biologically plausible neural network models using a range of techniques including processing using GPUs and other parallel hardware as well as custom processors. We are also developing systems and software tools for controlling and programming fine-grain processor arrays and vision chips