Urban Search and Rescue is a dangerous task for rescue workers and for this reason the use of mobile robots to carry out the search of the environment is becoming common place. These robots are remotely operated and the search is carried out by the robot operator. This work proposes that common search algorithms can be used to guide a single autonomous mobile robot in a search of an environment and locate survivors within the environment. This work then goes on to propose that multiple robots, guided by the same search algorithms, will carry out this task in a quicker time. The work presented is split into three distinct parts. The first is the development of a nonlinear mathematical model for a mobile robot. The model developed is validated against a physical system. A suitable navigation and control system is required to direct the robot to a target point within an environment. This is the second part of this work. The final part of this work presents the search algorithms used. The search algorithms generate the target points which allow the robot to search the environment. These algorithms are based on traditional and modern search algorithms that will enable a single mobile robot to search an area autonomously. The best performing algorithms from the single robot case are then adapted to a multi robot case. The mathematical model presented in the thesis describes the dynamics and kinematics of a four wheeled mobile ground based robot. The model is developed to allow the design and testing of control algorithms offline. With the model and accompanying simulation the search algorithms can be quickly and repeatedly tested without practical installation. The mathematical model is used as the basis of design for the manoeuvring control algorithm and the search algorithms. This design process is based on simulation studies. In the first instance the control methods investigated are Proportional-Integral-Derivative, Pole Placement and Sliding Mode. Each method is compared using the tracking error, the steady state error, the rise time, the charge drawn from the battery and the ability to control the robot through a simple motion. Obstacle avoidance is also covered as part of the manoeuvring control algorithm. The final aspect investigated is the search algorithms. The following search algorithms are investigated, Lawnmower, Random, HillClimbing, Simulated Annealing and Genetic Algorithms. Variations on these algorithms are also investigated. The variations are based on Tabu Search. Each of the algorithms is investigated in a single robot case with the best performing investigated within a multi robot case. A comparison between the different methods is made based on the percentage of the area covered within the time available, the number of targets located and the time taken to locate targets. It is shown that in the single robot case the best performing algorithms have high random elements and some structure to selecting points. Within the multi robot case it is shown that some algorithms work well and others do not. It is also shown that the useable number of robots is dependent on the size of the environment. This thesis concludes with a discussion on the best control and search algorithms, as indicated by the results, for guiding single and multiple autonomous mobile robots. The advantages of the methods are presented, as are the issues with using the methods stated. Suggestions for further work are also presented.
展开▼
