news
Two PhD positions in Robotics and AI
Two PhD positions are available in the Intelligent Robotics Laboratory (IRLab) at the University of Birmingham
Position 1: Learning models of human behaviour
This project is part of an FP7 project called Strands (http://www.strands-project.eu/), involving 30+ people, of whom a team of 6 work at the Birmingham IRLab (http://www.cs.bham.ac.uk/research/groupings/robotics/). In the overall project we investigate the task a robot that must perform 4D mapping. By this we mean producing not just a map of space (3D mapping), but also of the activities that occur within that space. Such activity maps are necessary for robots to act within a space shared with humans.
The activities performed in an environment by many humans over a period of time can typically be viewed as being composed of many smaller complex events and interactions. Many of these events occur periodically. There are differences across repetitions. The challenge of this PhD is how to recognise and exploit this periodicity in order to build compact models of complex behaviour involving multiple humans over long timescales (hours and days). The work will consider how the ordering and frequency of sub-events can be learned, taking account of the fact that sometimes particular sub-events may not occur, or that their ordering may be changed.
Position 2: Gaze control during robot manipulation of objects
The PacMan project (www.pacman-project.eu) is concerned with modelling objects as hierarchies of parts in 2D and 3D, and using these models to guide manipulation. One important question is given such models how to gather information during grasping.
In this PhD we will look at how to gather information about objects using vision. We will build on a model of visual gaze control previously developed by ourselves. The approach poses the problem of where the robot should look next as one of choosing gaze to maximise the increase in reliability of robot grasping. Speci cally the robot maintains a scene model with a belief filter for each item of interest (e.g. an object) and models the expected e ffect on belief of diff erent candidate gazes. The winning gaze is the one that will lead to the largest expected gain in manipulation performance on a task which is itself modelled as a rewarding process. In this PhD we will do two main pieces of work. First we will extend our gaze control model to consider gazing at parts of
objects suggested by the hierarchical model of object shape that are relevant for grasping. Second we will implement and evaluate the resulting gaze control model on a real robot. Our current model only deals with uncertainty in pose. In this task we will extend that to cover object shape, including missing surfaces.
Environment:
The IRLab at Birmingham is a leading European lab working in many aspects of intelligent robotics. The group has six faculty, seven research fellows and fifteen research students. The lab has a state of the art equipment with access to a wide range of advanced platforms for mobility and manipulation.
You should have a Batchelors or Masters in Computer Science, Electronic Engineering, Mathematics or Physics. You should have excellent mathematical and coding skills. You will have graduated in the top 5% of your class.
In the first instance contact Jeremy Wyatt (jlw@cs.bham.ac.uk), with your CV and transcript. Put “Strands PhD application” or “PacMan PhD application” as appropriate in the subject line. Deadline for applications is the 30th October, late applications will be considered if the appointment process hasn’t gone too far.