Control Systems for the Precision Guidance of Large Scale Off Road Robotic Vehicles

Control Systems for the Precision Guidance of Large Scale Off Road Robotic Vehicles

The research in this area is directed at automating large scale ground vehicles that operate on the agricultural fields and their autonomous navigation to conform to traversing specifications. They are, in general, articulated vehicle systems, in which a large tractor is providing propulsion and steering, and large implements are passively dragged behind. The control inputs to such a system are inadequate to guide the implements to follow a specified path. The research is aimed at providing the necessary control inputs at the implements to enable the precision autonomous navigation of the entire system. Research includes: i) the kinematic modelling of articulated systems taking into account the lateral and longitudinal wheels slips, ii) determination of force inputs using approximate dynamic models and the kinematic system responses, and iii) the design of controllers.

Autonomous Ground Vehicle - Husky

Autonomous Ground Vehicle - Husky

Based on the successful MAGIC2010 platform, this project explores a hardware and software combination of an off-the-shelf Husky A200 mobile platform, the Robotics Operating System, and a full range of standard sensors and actuators for field robotics competition purposes. A series of GPS waypoints are autonomously navigated while staying within painted boundaries and avoiding obstacles.

Data fusion and optimization in high dimensional processes

Data fusion and optimization in high dimensional processes

The application of techniques of Stochastic Data Fusion in estimation problems that involve high dimensional state vectors and the need to be performed under real-time constraints is a difficult task. In order to solve this type of problem, efficient techniques such versions of the EKF for high dimensional spaces, such as CEKF (Compressed Kalman Filter) are adequate tools. The adaptation of those techniques for Multi-core systems (such as GPU units) can be of relevant benefit. We investigate the implementation of expensive numerical processes such Stochastic Estimation (e.g. SLAM) and optimization approaches such as Dynamic Programming, by applying a .divide and conquer. policy, making the algorithms feasible to be performed in low cost parallel processing units (GPU).

Cooperative control of non-homogeneous vehicles using swarm intelligence

Cooperative control of non-homogeneous vehicles using swarm intelligence

The aim of this project is to develop efficient algorithms, inspired by social interactions in living species in a swarm, for the control of a group of non-homogeneous vehicles to perform desired tasks. In particular, vehicles are not equipped with the same type of sensors or motion abilities. Challenges arising from these constraints will be tackled in this project. Methodologies based on the multi-agent concept have arisen as a promising paradigm for coordinating, controlling and modelling complex systems such as transportation, resource allocation and production planning. The success of cooperation interrelates and depends on the sharing and exchange of information among every agent within the group. This necessitates a multi-objective optimal solution where the use of intelligent computation techniques is attractive in terms of their effectiveness. To this end, the emerging agent-based algorithms can bring new dimensions to the cooperative control problem, as the biologically inspired capability of social interaction and self-experience prompts directly to dynamic strategies in dealing with complexity and optimality.

Robust Control of Vectored Thrust Aerial Vehicles via Sliding Mode Control

Robust Control of Vectored Thrust Aerial Vehicles via Sliding Mode Control

The aim of the research is to explore the applicability of Sliding Mode based methodologies to unmanned aerial vehicles. A vectored thrust aerial vehicle developed at UNSW in 2010 is used as the test platform for the research. The final aim is to develop a nonlinear robust control methodology based on Sliding Modes which has improved convergence properties and eliminate current disadvantages of SMC.

Autonomous Coordinated Control of Public Transport Platoons

Autonomous Coordinated Control of Public Transport Platoons

In many cities around the world, the vehicles of public transportation agencies significantly contribute to vehicle congestion. In some cities a new approach is being trialed where dedicated public transport road infrastructure, contrary to rail road or MTR infrastructure, is used. Due to substantially improved vehicle control, such scenarios are sufficiently attractive to implement autonomous public transport convoys. This work takes into account the dynamic modelling of individual elements of the platoons and develops coordinated control of intersecting platoons, over taking platoons and contingency management of autonomous platoon control. Further the study aims to develop optimized resource allocations for a public transport system consisting of multi-node transport sub-hubs. In the current simulations, a reduced set of Sydney's transport sub-hubs are used.

Force Controlled Ground Vehicles

Force Controlled Ground Vehicles

This project develops control methodologies that will determine the drive forces of a vehicle so that it can be directed to follow a path accurately despite the changes in terrain conditions. The prototype being developed is a four wheel steer four driven system with full 6D force sensors at each wheel. These force sensors will enable the initial development of lateral and longitudinal force modelling as well as rolling resistance modelling and they validation through direct measurement of forces. At a later stage the force vectoring at the driven wheels through independent steering will be used to navigate the vehicle to follow a predefined trajectory under varying terrain conditions.

 

Improved Yield Estimation for the Australian Wine Industry

The project aims to improve upon current yield estimation methods, with a focus on techniques drawn from the field of computer vision. Initially the project will assess, design and benchmark computer vision techniques in yield estimation from various stages of the wine growth cycle. The project will then continue into the development of a range of mobile low cost devices that can be used to help farmers and wineries gain a more accurate estimate of yield in the field.

Page generated: Friday 24 November, 2017