FindAPhD LIVE! Study Fair

Oxford | Leeds

University of Bristol Featured PhD Programmes
Imperial College London Featured PhD Programmes
Norwich Research Park Featured PhD Programmes
University of Kent Featured PhD Programmes
FindA University Ltd Featured PhD Programmes

Optimal cue integration in biomimetic navigation algorithms

This project is no longer listed in the FindAPhD
database and may not be available.

Click here to search the FindAPhD database
for PhD studentship opportunities
  • Full or part time
    Dr P Graham
    Dr A Philippides
  • Application Deadline
    No more applications being accepted
  • Competition Funded PhD Project (European/UK Students Only)
    Competition Funded PhD Project (European/UK Students Only)

Project Description

Navigation is a vital task for animals and robots and efficient navigation requires robustness to environmental and sensory uncertainty, with flexibility to integrate multiple sources of information. The behaviour of navigation specialists such as ants and bees, shows that they can solve these problems despite limited neural resources. A deeper understanding of the neural and theoretical basis for cue integration and robustness to sensory uncertainty would be a major benefit to the field of bioinspired navigation, where insects are templates for low-computation, low-power robots, capable of prolonged autonomy.

This exciting interdisciplinary project will form part of the EPSRC funded Brains on Board (BoB) program grant, a multi-university project in which we aim to create robots with the learning abilities of bees. You will join the Sussex BoB team and contribute to the vision of the project, while benefitting from years of experience in bio-inspired algorithms as well as cutting-edge facilities and equipment.

Detail:

The basic approach will be to benchmark biomimetic models of insect navigation against navigation models from robotics. In engineering approaches to robot navigation, a family of algorithms called SLAM (Simultaneous Localisation and Mapping) provide optimal cue integration in the building of spatial representations. This optimality depends on reliable extraction of visual information and accurate information about sensor uncertainty. For this project we propose an exploration of the conditions under which biomimetic controllers approximate the performance of SLAM models. More importantly, we will investigate which properties of biomimetic controllers lead to increased performance. The outcomes will be improved navigation controllers for low-computation robots, and deeper understanding of how robust vision and cue integration can be implemented in neural circuits.

We envisage that the project will involve two major research strands, as a successful applicant you will be able to influence the details and relative weighting of these strands, depending on your prior experience and research interests. Furthermore, there will be flexibility in the project so that we can respond to exciting results and new ideas.

In the first strand, you will explore robustness in the visual system of a navigating robot. Visual approaches to navigation sit along a continuum from feature based (extracting discrete identifiable features from a scene) to appearance based (scenes are described by parameters independently of specific objects). Natural visual systems depend on visual filters that can be tightly tuned to pick out specific features, or broadly tuned to represent appearance information. Thus, one can build a visual system that is capable of being feature or appearance based and investigate its performance as part of a biomimetic or SLAM-like navigation system. By measuring performance of navigation systems as a function of the visual input, one can understand where robustness comes from.

In the second research strand, you will investigate cue integration. Robots and animals integrate self-generated and environmental information as part of navigation tasks. SLAM approaches use confidence estimates for information sources as part of optimal Bayesian integration. In low-computation biomimetic systems, one can approximate this process with simple heuristics. We will explore the computation-to-accuracy trade-off for optimal versus heuristic cue integration. This research strand would also offer the opportunity to explore how cue integration is implemented in neural circuitry.

To excel in this project you will have either a strong quantitative background with interest in bio-inspired solutions or a biological background with strong maths and computing skills and a keen desire to improve them.

How to apply:

Please submit a formal application using our online application system at http://www.sussex.ac.uk/study/phd/apply, including a CV, degree transcripts and certificates, statement of interest (clearly stating supervisor’s name and the project title) and names of two academic referees. On the application system use Programme of Study – PhD Biology.

Requirements:

Applicants will have an excellent academic record and should have or be expected to receive a relevant first or upper-second class honours degree. The EPSRC award is available to UK and to EU students who have been ordinarily resident in the UK for the previous 3 years. EU candidates who do not meet this criteria will be eligible for a fee waiver only. Overseas (non EU) students are not eligible to apply for EPSRC funding, but they are welcome to apply if they have access to other sources of funding.

Funding Notes

This project is one of a number ear-marked for funding by the University of Sussex Engineering and Physical Sciences Research Council (EPSRC) Doctoral Training Partnership to commence in September 2018. This project is in direct competition with others for funding; the projects which receive the best applicants will be awarded the funding.

For enquiries about the application process contact Anna Izykowska ([Email Address Removed])
For enquiries about the project contact Dr Paul Graham ([Email Address Removed])

References




FindAPhD. Copyright 2005-2018
All rights reserved.

Let us know you agree to cookies

We use cookies to give you the best online experience. By continuing, we'll assume that you're happy to receive all cookies on this website. To read our privacy policy click here

Ok