A comparison study of search heuristics for an autonomous multi-vehicle air-sea rescue system

Rafferty, Kevin John (2014) A comparison study of search heuristics for an autonomous multi-vehicle air-sea rescue system. PhD thesis, University of Glasgow.

Full text available as:
Download (4MB) | Preview
Printed Thesis Information: https://eleanor.lib.gla.ac.uk/record=b3059845


The immense power of the sea presents many life-threatening dangers to humans, and many fall foul of its unforgiving nature. Since manned rescue operations at sea (and indeed other search and rescue operations) are also inherently dangerous for rescue workers, it is common to introduce a level of autonomy to such systems. This thesis investigates via simulations the application of various search algorithms to an autonomous air-sea rescue system, which consists of an unmanned surface vessel as the main hub, and four unmanned helicopter drones. The helicopters are deployed from the deck of the surface vessel and are instructed to search certain areas for survivors of a stricken ship. The main aim of this thesis is to investigate whether common search algorithms can be applied to the autonomous air-sea rescue system to carry out an efficient search for survivors, thus improving the present-day air-sea rescue operations.

Firstly, the mathematical model of the helicopter is presented. The helicopter model consists of a set of differential equations representing the translational and rotational dynamics of the whole body, the flapping dynamics of the main rotor blades, the rotor speed dynamics, and rotational transformations from the Earth-fixed frame to the body frame.

Next, the navigation and control systems are presented. The navigation system consists of a line-of-sight autopilot which points each vehicle in the direction of its desired waypoint. Collision avoidance is also discussed using the concept of a collision cone. Using the mathematical models, controllers are developed for the helicopters: Proportional-Integral-Derivative (PID) and Sliding Mode controllers are designed and compared.

The coordination of the helicopters is carried out using common search algorithms, and the theory, application, and analysis of these algorithms is presented. The search algorithms used are the Random Search, Hill Climbing, Simulated Annealing, Ant Colony Optimisation, Genetic Algorithms, and Particle Swarm Optimisation. Some variations of these methods are also tested, as are some hybrid algorithms. As well as this, three standard search patterns commonly used in maritime search and rescue are tested: Parallel Sweep, Sector Search, and Expanding Square. The effect of adding to the objective function a probability distribution of target locations is also tested. This probability distribution is designed to indicate the likely locations of targets and thus guide the search more effectively. It is found that the probability distribution is generally very beneficial to the search, and gives the search the direction it needs to detect more targets. Another interesting result is that the local algorithms perform significantly better when given good starting points. Overall, the best approach is to search randomly at the start and then hone in on target areas using local algorithms. The best results are obtained when combining a Random Search with a Guided Simulated Annealing algorithm.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Keywords: Autonomous System, Search Heuristics, Optimisation, Air-sea Rescue, Helicopter Control, Sliding Mode Control
Subjects: T Technology > T Technology (General)
Colleges/Schools: College of Science and Engineering > School of Engineering > Autonomous Systems and Connectivity
Supervisor's Name: McGookin, Dr. Euan
Date of Award: 2014
Depositing User: Mr Kevin Rafferty
Unique ID: glathesis:2014-5292
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 23 Sep 2014 09:22
Last Modified: 23 Sep 2014 10:02
URI: http://theses.gla.ac.uk/id/eprint/5292

Actions (login required)

View Item View Item


Downloads per month over past year