The advent of mobile and ubiquitous systems has enabled the development of autonomous
systems such as wireless-sensors for environmental data collection and teams of collaborating Unmanned Autonomous Vehicles (UAVs) used in missions unsuitable for humans. However, with these range of new application domains comes a new challenge – enabling self-management in mobile autonomous systems. The primary challenge in using autonomous systems for real-life missions is shifting the burden of management from humans to these systems themselves without loss of the ability to adapt to failures, changes in context, and changing user requirements.
Autonomous systems have to be able to manage themselves individually as well as to form self-managing teams that are able to recover or adapt to failures, protect themselves from attacks and optimise performance.
This thesis proposes a novel distributed policy-based framework that enables autonomous systems to perform self management individually and as a team. The
framework allows missions to be specified in terms of roles in an adaptable and reusable way, enables dynamic and secure team formation with a utility-based approach
for optimal role assignment, caters for communication link maintenance among team members and recovery from failure. Adaptive management is achieved by employing an architecture that uses policy-based techniques to allow dynamic modification of the management strategy relating to resources, role behaviour, team and communications management, without reloading the basic software within the system.
Evaluation of the framework shows that it is scalable with respect to the number of roles, and consequently the number of autonomous systems participating in the
mission. It is also shown to be optimal with respect to role assignments, and robust
to intermittent communication link disconnections and permanent team-member
failures. The prototype implementation was tested on mobile robots as a proof-ofconcept
demonstration.