Ben Grocholsky Doctor of Philosophy The University of Sydney March 2002 Information-Theoretic Control of Multiple Sensor Platforms This thesis is concerned with the development of a consistent, information-theoretic basis for understanding of coordination and cooperation decentralised multi-sensor multi-platform systems. Autonomous systems composed of multiple sensors and multiple platforms potentially have significant importance in applications such as defence, search and rescue, mining or intelligent manufacturing. However, the effective use of multiple autonomous systems requires that an understanding be developed of the mechanisms of coordination and cooperation between component systems in pursuit of a common goal. A fundamental, quantitative, understanding of coordination and cooperation between decentralised autonomous systems is the main goal of this thesis. This thesis focuses on the problem of coordination and cooperation for teams of autonomous systems engaged in information gathering and data fusion tasks. While this is a subset of the general cooperative autonomous systems problem, it still encompasses a range of possible applications in picture compilation, navigation, searching and map building problems. The great advantage of restricting the domain of interest in this way is that an underlying mathematical model for coordination and cooperation can be based on the use of information-theoretic models of platform and sensor abilities. The information theoretic approach builds on the established principles and architecture previously developed for decentralised data fusion systems. In the decentralised control problem addressed in this thesis, each platform and sensor system is considered to be a distinct decision maker with an individual information-theoretic utility measure capturing both local objectives and the inter-dependencies among the decisions made by other members of the team. Together these information-theoretic utilities constitute the team objective. The key contributions of this thesis lie in the quantification and study of cooperative control between sensors and platforms using information as a common utility measure. In particular, • The problem of information gathering is formulated as an optimal control problem by identifying formal measures of information with utility or pay-off. • An information-theoretic utility model of coupling and coordination between decentralised decision makers is elucidated. This is used to describe how the information gathering strategies of a team of autonomous systems are coupled. • Static and dynamic information structures for team members are defined. It is shown that the use of static information structures can lead to efficient, although sub-optimal, decentralised control strategies for the team. • Significant examples in decentralised control of a team of sensors are developed. These include the multi-vehicle multi-target bearings-only tracking problem, and the area coverage or exploration problem for multiple vehicles. These examples demonstrate the range of non-trivial problems to which the theory in this thesis can be employed.

Information-Theoretic Control of Multiple Sensor Platforms

In this paper the problem of fully decentralized data fusion and control for a modular wheeled mobile robot (WMR) is addressed. This is a vehicle system with nonlinear kinematics, distributed multiple sensors, and nonlinear sensor models. The problem is solved by applying fully decentralized estimation and control algorithms based on the extended information filter. This is achieved by deriving a modular, decentralized kinematic model by using plane motion kinematics to obtain the forward and inverse kinematics for a generalized simple wheeled vehicle. This model is then used in the decentralized estimation and control algorithms. WMR estimation and control is thus obtained locally using reduced order models with reduced communication requirements, in a scalable network of control nodes, If communication of information between nodes is carried out after every measurement (full rate communication), the estimates and control signals obtained at each node are equivalent to those obtained by a corresponding centralized system. Transputer architecture is used as the basis for hardware and software design as it supports the extensive communication and concurrence requirements which characterize modular and decentralized systems. The advantages of a modular WMR vehicle include scalability, application flexibility, low prototyping costs and high reliability.

Estimation and control for a modular wheeled mobile robot

Eric Nettleton Doctor of Philosophy The University of Sydney 2003 Decentralised Architectures for Tracking and Navigation with Multiple Flight Vehicles This thesis is concerned with the development and demonstration of decentralised data fusion (DDF) algorithms in airborne applications. The decentralised architectures described in this thesis require no central fusion centre and no common communications medium. Sensor nodes are joined in a decentralised network without any global knowledge of the network topology or other node capabilities. Each sensor node is able to form a global estimate based on local sensor observations and information communicated to it by adjacent sensor nodes in the network. In a DDF architecture, no sensing, processing or communication component is critical to the operation of the overall system so a failure of any single element results in only an incremental decrease in performance rather than catastrophic system failure. As no node requires knowledge of the global network topology, the system can be scaled simply by connecting new sensing nodes to the system. The main contribution of this thesis is the development of decentralised algorithms and architectures able to deal with asynchronous and intermittent communications characteristic of real-world sensor nodes communicating over radio networks. The algorithms presented in this thesis are developed using the information form of the Kalman filter running on a variety of structured and unstructured decentralised sensor networks. Although this thesis is primarily concerned with structured communication topologies, that are able to make complete use of global information, sub-optimal algorithms for use in networks with dynamic connectivity changes are also proposed. Exact solutions to the delayed and asequent data problems are developed for the information form of the Kalman filter that enable operation of DDF algorithms in a broad range of real sensor networks. The DDF architecture is applied to the problem of tracking multiple ground targets using a network of airborne sensing platforms. The objective of this problem is for the sensing nodes on each platform to build a composite global picture of targets in an environment using both local sensor observations and information communicated from other platforms in a decentralised manner. The second important contribution made by this thesis is in the development of algorithms for multi-vehicle Decentralised Simultaneous Localisation and Mapping (D-SLAM). Closed form solutions to the SLAM covariance and information matrix for simplified single and multiple vehicle D-SLAM problems are presented. These show that; i) in general platform

Decentralised Architectures for tracking and navigation with multiple flight vehicles

Introduction to Decentralised Data Fusion

A Beginners Guide to Decentralised Data Fusion

In this paper the Nonlinear Information Filter is derived from the Extended Kalman Filter. A nonlinear system is considered. Linearizing the state and observation equations, a linear estimator which keeps track of total state estimates is conceived; the Extended Kalman Filter. The linearized parameters and filter equations are expressed in information space. This gives a filter that predicts and estimates information about nonlinear state parameters given nonlinear observations and nonlinear system dynamics. The Nonlinear Information Filter derivation is contrasted to that of the Linear Information filter. Pitfalls of a naive extension of the later to the former are thus identified. Furthermore, the Nonlinear Information filter is decentralized and distributed, to give the Distributed and Decentralized Nonlinear Information. Application is real decentralized data fusion and distributed control is proposed. Specifically, realtime distributed/decentralized control of a navigating, modular wheeled robot is considered.

Arthur G. O. Mutambara

Papers

Nonlinear information space: a practical basis for decentralization