Current research projects: (This is a stub)
At the heart of my research is the question of how artificially intelligent algorithms can deal with uncertainty in the perception of their environment and how they can learn from experience.
I use Object Petri Nets as a model of communication and as a mathematical model for analysing system properties. Petri nets combine the mathematical rigour with the visualisation required to convey complex scenarios to the lay person. They have proven incredibly useful in many collaborations.
Verifying Petri Net Models of Agent-based Systems
We use high-level Petri nets following the nets-within-nets paradigm to model complex systems and analyse essential system properties. The class of Agent Petri nets used in this project is designed to capture the agents’ current knowledge of and visualise the perception of the environment, the reasoning and any communication leading to a change in the agents’ beliefs. Mobility of agents and limited resources are naturally modelled at net level. Due to the nested Petri net structures used, encapsulation, abstraction/refinement, and hierarchical development of multi-agent systems are supported.
Engineering Agent Software (for the IoT and smart environments)
The project introduces a component-based framework for creating software agents with a focus on the Internet of Things (IoT) and smart home applications. The Layered Agent Framework (LAF) provides a separation of responsibilities that supports real-world requirements, such as low-level support for encryption, unobtrusive application of updates to the system, and recovery. Agents are autonomous software entities that can apply reasoning to achieve goals they have been set, or that they have set themselves proactively or in negotiation with other agents. Our agents follow the Beliefs, Desires, and Intentions (BDI) paradigm that is widely accepted as standard in the Artificial Intelligence and Agents communities.
The BDI Blocks implementation of the framework supplies re-usable components for building BDI-compliant deliberation cycles as a basis for agent implementations on embeds able systems.
Modelling the Cell Cycle
We model the cell cycle and effects of medication with a view to analyse the potential effectiveness of treatment prior to clinical trials. Data is taken from in vivo and in vitro experiments carried out by our research partners. Apart from providing a mathematical model to predict the inheritance of properties by subsequent generations of cells in an evolving population based on stochastic data, our models are used to visualise the processes involved. This helps to convey causalities hidden in other mathematical models to an audience not familiar with the stochastic models.
Smart Technology for Dementia Support in the Community
To delay full-time institutional care for dementia patients by use of smart-technology (smart- phones, wearables, smart-homes). To investigate solutions that can decrease the risks dementia-affected persons are subject to, and at the same time increase their independence, autonomy, self-confidence, and mobility. Design and implement autonomous algorithms/agents on smart-devices to making decisions without continuously transmitting potentially sensitive data to a central unit.