Stacey, Geoff2017-05-242017-05-24b43751830http://hdl.handle.net/1885/117037The research in this thesis addresses the subject of sensor-based formation control for a network of autonomous agents. The task of formation control involves the stabilisation of the agents to a desired set of relative states, with the possible additional objective of manoeuvring the agents while maintaining this formation. Although the formation control challenge has been widely studied in the literature, many existing control strategies are based on full state information, and give little consideration to the sensor modalities available for the task. The focus of this thesis lies in the use of a generic arrangement of partial state measurements as can commonly be acquired by onboard sensors; for example, time-of-flight sensors can be used to measure the distances between vehicles, and onboard cameras can provide the bearing from one vehicle to each of the others. Particular aspects of the problem that are addressed in this thesis include (i) ways of modelling the formation control task, (ii) methods of analysing the system's behaviour, and (iii) the design of a formation control scheme based on generic arrangements of sensors that provide only partial position information. A key contribution in this thesis is a generalisation of the classical notion of rigidity, which considers the use of distance constraints between agents in R^2 or R^3 to specify a rigid body (or formation). This enables the concept of rigidity to be applied to agent networks involving a variety of (possibly non-Euclidean) state-spaces, with a generic set of state constraints that may, for example, include bearings between agents as well as distances. I demonstrate that this framework is very well-suited for modelling a wide variety of formation control problems (addressing goal (i) above), and I extend several fundamental results from classical rigidity theory in order to provide significant insight for system analysis (addressing goal (ii) above). To design a formation control scheme that uses generic partial position measurements (addressing goal (iii) above), I employ a modular passivity-based approach that is developed using the bondgraph modelling formalism. I illustrate how adaptive compensation can be incorporated into this design approach in order to account for the unknown position information that is not available from the onboard sensors. Although formation control is the subject of this thesis, it should be noted that the rigidity-based and passivity-based frameworks developed here are quite general and may be applied to a wide range of other problems.enformation controlvehicle formationsagent formationssensor modelingsensor modellingpartial state measurementsrigidity theorygeneralised rigiditygeneralized rigiditypassivity based controlbond graph modellingbond graph modeling201610.25911/5d723d5106684