Examples of student projects
Project Title: Making sense of routinely collected behavioural data.
Project Summary:The project involves designing an AI system that will collect, analyse and feedback relevant information regarding the participant’s daily behaviour using in-house developed wearable devices. This involves 1) designing the infrastructure for data collection and analysis, 2) co-design the interface together with end-users to maximise the relevance of the feedback and allow for effective interaction. The outcome will be a prototype system deployed in the IAI facilities.
Project Title: Sliding Doors - hardware and software for restricting the movement individual ants between nest chambers
Project Description: The Computational Ant Lab at the University of Bristol uses artificial intelligence and robotics to study distributed information processing and collective decision making in ant colonies; and it has supervised numerous student projects in relevant tasks ranging from computer vision, machine learning, equation discovery, spatial simulation, behavioural modelling and robotic nests .
In order to build on this expertise, we have recently teamed up with Nathalie Stroeymeyt, from the School of Biological Sciences, who has commissioned the construction of a several brand-new climatically-controlled ant arenas in which individual Formicidae ants can now be barcoded and tracked  (https://stroeymeyt-lab.ch/research/). The purpose of this project is now to develop hardware and software for automatically restricting the movement individual ants between various nest chambers using some form of intelligent gating system.
Although previous work successfully developed a means of gating the movement of RFID-tagged Temnothorax ants between a nest and a foraging area using an electro-magnetically suspended hinged metal strip , that system is not suitable for our work because the mechanism would obstruct the tracking camera, because it is designed for use in situations where ants are not densely packed together, and because the control algorithm (presence/absence of a single ant at a unique pre-defined location) is not suitable for the rich two-dimensional multi-entity data produced by the ant tracking system. Therefore the project will involve developing a gating mechanism that can be integrated with the output of the existing tracker to reliably prevent the movement of selected individuals within the colony in real-time without injuring or impeding the movement of any other ants. Because of the sensitivity of the climate control system, consideration would have to be given to the heat output of any proposed solution. The goal would be to deploy this gating system in an automated scientific study involving colonies of real ants.
The proposed solution is clearly interactive because it involves real-time tracking and interaction with ants in a living colony. The project will involve an equal mix of implementation (developing the physical gates and mechanisms to control them) and investigation (testing out which solutions work most reliably). This project requires a mix of hardware and software development along with an ability to interact with our project partners in Biological Science and Engineering Mathematics.
-  https://sites.google.com/site/computationalantlab/
-  https://github.com/formicidae-tracker/documentation/wiki
-  https://jeb.biologists.org/content/215/15/2653
Project Title: Explanation via textualisation in machine learning.
Project Summary: Understanding data, models and predictions is key for any machine learning application. Due to the limitations of our spatial perception and intuition, analysing high-dimensional data is inherently difficult. Furthermore, black-box models achieving high predictive accuracy are widely used, yet the logic behind their predictions is often opaque. The use of textualisation – a natural language narrative of the selected phenomenon – can tackle these shortcomings. When extended with argumentation and reasoning theory we could envisage machine learning models and predictions arguing persuasively for their choices.
Data-driven AI: deep learning. Knowledge-driven AI: logical reasoning. Human-AI Interaction: narrative and NLP. Responsible AI: interpretability and trust.