Large-scale Artificial Socio-Cognitive Systems
Autonomy (in computer systems) is an inevitable feature of the emerging physical + cyberspace being created by our eager but poorly understood uptake of technological artefacts to operate and manipulate both physical and cyber worlds. Modelling and building such systems requires a novel approach to the design and understanding of self-managing partially-observable complex systems, for which existing network and (numerical) mathematical models are inadequate and unsafe because of their over-simplifications and incapacity to explain intrinsic but unexpected emergent outcomes.
The primary tool of our approach is (computational) logic, embedded in distributed intelligent systems, to realize massive multiagent
simulations for the class of global systems science problems. Explicit representation of knowledge and reasoning processes allows a
very high degree of entity individuation and the use of abduction to uncover how and why particular events occurred.
The research will include development of new theory, building on the above, but also, depending on the evolving focus of the research, the testing of that theory through software for cloud platforms in areas such as:
1. Large scale intelligent agent-based simulation. Current platforms cannot utilise multi-core or cluster hardware and struggle supporting even 1000 agents on a single node.
2. In-silico experiment observation and visualization. Target domains are continuous systems, where detection of trends or
disturbances, particularly those that might upset the stability of the model, or cause it to shift to a new equilibrium, is essential.
3. Platform self-management. Unlike conventional procedural codes, such a platform relies upon large numbers of asynchronous components with unpredictable reasoning times that, left unchecked, can cause platform malfunctions. This apparent fragility is not a fault, but highlights the gap between conventional computer job management and that needed for next-generation simulation/decision-support systems.
The successful candidate will be a Home/EU student who holds a 1st class or high 2:1 undergraduate degree (or any EU equivalent) in a relevant discipline. This PhD studentship will cover the full tuition fees at the Home/EU rate, as well as a training support fee and a standard tax-free stipend (at least £14,057/annum) for 3.5 years.
Applicants must satisfy RCUK residency rules for the full studentship. The successful candidate is expected to start on 28th September 2015.
We also invite applications from worldwide students who can self-fund.