As computational and engineering applications become more sophisticated, the need for autonomous systems that can act intelligently without direct human intervention increases. Yet the autonomous control at the heart of many such systems is often ad-hoc and opaque. Since the cost of failure in critical systems is high, a more reliable, understandable and consistent approach is needed. Thus, in this project we aim to provide a rational agent architecture that controls autonomous decision-making, is re-usable and generic, and can be configured for many different autonomous platforms. In partnership with the industrial collaborators we aim to show how such reconfigurable autonomy can be achieved in relevant applications. The main objectives of the project are as follows.
1. to define a core hybrid agent architecture, based on that introduced within a previous project on Engineering Autonomous Space Systens, but incorporating key elements from intelligent systems work at Surrey;
2. to investigate how a library of “environmental interfaces” can be incorporated into this architecture, where typical elements within the library corresponding to special control systems, networks interfaces, or infrastructure;
3. to investigate and incorporate multi-agent aspects when multiple autonomous components are involved;
4. to consider the (formal) analysis of reliability of the core/generic architecture;
5. in collaboration with the industrial project partners, to show how this generic architecture can be instantiated to some of the areas they are interested in; and
6. to demonstrate a selection of real autonomous systems in relevant (possibly hardware) testbeds and to evaluate the practical efficacy of the instantiated architecture.

Sheffield Robotics People:

Sandor Veres (University of Sheffield)
Jonathan Aitken (University of Sheffield)
Michael Fisher (University of Liverpool)
Elisa Cucco (University of Liverpool)
Yang Gao (University of Surrey)
Affan Shukat (University of Surrey)

Groups:

Autonomous Control Lab
ASRG