Dr Hayes Martin

Project Supervisor: Dr Martin Hayes Project No: MH1
Project Title: Feature extraction from vehicle based cameras
Course Suitability: BE

Project Description:

Human error remains the main cause of road accidents. Intelligent driver systems that can monitor a driver’s state and behaviour in order to aid collective safety are now being proposed.

https://www.google.com/selfdrivingcar/how/

This project will analyse data (primarily camera but also other sensor data if appropriate) to see whether particular features can be extracted reliably. In particular the identification of humans from camera data is critical for any Driver Assist or other decision making software.

A particular outcome from the project will be the correct identification of whether a pedestrian is in a safe (i.e., on the footpath) or at risk position (on the road) wherein corrective action may be required.

The project will involve :-

  1. Selection of appropriate test camera data.
  2. Familiarisation with the basics of Feature Extraction.                  
  3. Development of Feature Identification algorithms from camera data.
  4. Are pedestrians in a safe or at risk location?
  5. How does the availability of extra sensor information affect the identification process of Step (iv)? How can the Sensor Fusion problem that exists when multiple sensor sources are available be managed intelligently?
Project Supervisor: Dr Martin Hayes Project No: MH2
Project Title: Wavefront Algorithm implementation for dynamic maze solution using two Redbots
Course Suitability: BE / BSc

Project Description:

This project looks at the solution of simple maze type problems using shared or co-operative learning by each robot. The project will implement the wavefront algorithm and the project should analyse how the amount of time taken to solve a maze problem is reduced when two robots working as a team is compared with one device working on its own. If time permits we can also look at various leader follower and intruder detection problems using the Redbot as a basic low cost test robot.

The project will involve :-

  1. Analysis of the functionality of off the shelf redbot robot (kits?)
  2. Execution of simple standalone maze solution algorithms (like eg:- https://en.wikipedia.org/wiki/Maze_solving_algorithm )
  3. Implementation of wavefront algorithm using the Redbot
  4. Formal investigation of “information sharing” between robots
  5. Assessment of improvement possible using a team based approach.
  6. Development of intruder alert / monitoring app using the Redbot.
Project Supervisor: Martin Hayes Project No: MH3
Project Title: Modification of the EPV possession metric for use in field sports
Course Suitability: BE / BSc

Project Description:

This project looks at data analytics that are now a fact of life in the analysis of Sports performance. EPV denotes the (Expected Points Value) for a particular possession by a particular player in a particular location of the court in Basketball. As a starting point for the work we will consider Don Cervone’s MIT Sloan paper “Pointwise: Predicting Points and Valuing Decisions in Real Time using Optical Data” and consider how this work might apply to football. In football the quality of possession at any time is likely to be lower/more difficult to predict than when, say, a basketball player is dribbling because in general (unless you happen to be Messi) the ball is under less control when in transit.

The project will involve :-

  1. Analysis of the EPV metric and the reproduction of data from the Cervone trials.
  2. What are the added/different variables that are in play during a game of Football
  3. Research whether any EPV analogues exist in Football analysis
  4. Formal investigation of Match data to see whether the EPV metric is applicable in field sports.
  5. Development of a simple pilot app that considers an EPV metric for simple training data. A small amount of open source game data will be considered.
Project Supervisor: Dr Martin Hayes Project No: MH4
Project Title: Development of IoT App for MV ESB Distribution Poles
Course Suitability: BE/BSc

Project Description:

Creosote poles are used nationwide by the ESB for the distribution of Electrical Power at Medium Voltages (600 to 69kV). The problem with Creosote is that it is a (very) low grade conductor leading to a build up of voltage over time within the pole that can pose a risk to maintenance staff.

This project will use an Internet of Things (IoT) enabled approach to specify and implement on a pilot basis a solution to this problem. Test probes will be deployed at various points about a sample pole, diagnostic tools will be developed and a low bandwidth networking solution (VT network or similar) will be proposed so that a clear picture is available at a distribution centre in relation to potential shock/discharge problems that might obtain on the MV distribution network.

The Project will involve the :-

  1. Development of simple embedded system that can act as instrumentation for this problem.
  2. Design of a pole based IoT module that can provide diagnostics in relation to charge build up and distribution across a pole.
  3. Communication of Object information in an efficient (low bandwidth) networked fashion.
  4. Efficient cloud based storage of data.
  5. Investigate the aggregation of data for use by Eirgrid personnel.