PhD and Visiting Researcher Projects

Applications for PhD study under my supervision within the Computer Vision & Autonomous Systems group  (CVAS) are now open for AY2017-2018. I am offering a number of projects in computer vision for robotics as outlined below, and these are in line with both the core research conducted by CVAS and my research interests. I am prepared to discuss alternative projects, however these must also be aligned with the core research undertaken by CVAS in vision and robotics.

Here are some examples of the sort of PhD projects I would like to supervise:

• An investigation of how to couple our software retina to a Deep CNNs classifier and  compare/structure this DCNN with the Human Visual System. In order to achieve acceptable learning performance the retina front end will likely require to be GPU accelerated.
• Foveated, flowfield-linked object recognition to support core robot vision. Investigation of real-time dynamic exploration and appearance learning in robot-arm mounted binocular sensors, where the machine learns by interacting directly with objects in the scene.
• Investigate how to integrate the our software retina on mobile phone platform. Our goal is to build a low-cost intelligent robot sensor that pre-computes the initial processing stages on-board in order to reduce I/O bandwidth with the host controller and fully utilise the phones GPU and sensory package. We propose to adopt a “Deep hardware” concept whereby the initial processing on the platform will feed Deep CNNs directly at low data rates. (Current collaboration with ARM Research, Cambridge, hosting a PhD intern)
• An investigation on how to construct a robot vision control architecture structured as a collection of deep nets in a ROS framework - this would be based on a combination of the functional processing units for 3D sensing and camera vergence & gaze control ROS nodes we developed under CLOPEMA and object recognition & manipulator control ROS nodes based on learned functions using appropriately structured DCNNs.
• Stereo-matcher integrated with classical shape-from-shading constraints or Deep CNNs (potential industrial sponsorship) to improve the robustness of our current GPU accelerated passive stereo-based high-resolution RGBD sensor to  implement a very low cost, but high quality real-time robot 3D vision system.
• Investigate the formulation of a real-time processing pipeline from visual capture to control of a dexterous multi-fingered robot hand based on foveated 3D sensing integrated with smooth particle modelling  and a repertoire of manipulation behaviours capable of handling both rigid and non-rigid objects for robot workcell applications. (potential industrial sponsorship)
• Explore combined vision and tactile array sensing in  robot hands controlled by DCNNs for and dexterous manipulation and interactive proprioceptive perception & exploration of the robot’s workspace.

PhD Applications

If you are interested in PhD study with me, please send me an email outlining your expertise and CV and indicate the research topics on which you would like to work. If appropriate we can then discuss a research topic by email or Skype, prior to making a formal application for PhD Study. Instructions on how to make a formal application for PhD study are located on the College of Science & Engineering Graduate School web page. The Graduate School web pages also contain information regarding scholarships  for PhD study.

Visiting Researcher Applications

I am happy to host visiting researchers in the CVAS group and any researcher wishing to explore becoming a visiting researcher should contact me by email me with their CV and  an outline of their expertise and research interests.  

CVAS Facilities

CVAS hosts two full-size humanoid robots: Dexterous Blue is a large industrial robot comprising two arms, supplied by Kawasaki Motoman, which are mounted on a rotating plinth. This robot has been equipped with a steerable binocular vision head and processing system developed in our group. This powerful and precise 750Kg machine is located in a custom laboratory with control/viewing gallery for safe operation of the machine on the 7th floor of the Boyd Orr Building. Currently Dexterous Blue is equipped with specialied end effectors developed for clothing manipulation within the CLOPEMA project and examples of previous PhD project work within the CLOPEMA project on robotic clothes folding are available at the CLOPEMA website (including videos, datasets and codes). We plan to replace one of these end effectors with a more general purpose 4-fingered gripper, e.g. as supplied by SAKE robotics to progress our research in interactive perception. (The SAKE grippers can pick up pretty much anything, the CLOPEMA effector is highly optimised for manipulating clothing, but not much else.)

We also have a Baxter Research Robot used also for undergraduate and MSc projects in a showcase laboratory in the foyer of the Sir Alwyn Williams Building. Baxter is an inherently safe robot (unlike Dexterous Blue which is operated in isolation from humans under a strict safety plan) that can can be operated in close proximity to operators and bystanders. We have now mounted a two-fingered SAKE gripper to one of Baxter’s arms and this is functioning and available for projects.

In addition to the above humanoid robots, the CVAS group also hosts a number of smaller robots, including a hexapod, and a number of binocular stereo vision systems. CVAS also holds an extensive library of computer vision software developed in-house, including  3D sensing and interpretation, robot vision, gaze control, and biologically motivated vision.