The purpose of my thesis is to give an insight into the world of the free-climbing robots. In many areas the usage of these robots is an efficient way to complete tasks. These robots can climb on uneven terrains as well as on steep areas. According to a new approach the robot grabs the holds of the terrain to climb like free-climbers. Purpose of my thesis is also to implement a path planning algorithm, which helps the robot to climb on uneven terrains.
My thesis is divided into three parts: double-link robot, three-limbed robot, and neural networks.
The topics which are related to double-link robots are: the parameters of a configuration, the configuration space (stance manifolds) determining algorithm, the multi-step planning, and the RDT algorithm. These algorithms are implemented in MATLAB then the results are evaluated.
The parts of the three-limbed robot chapter are: the model of the three-limbed robot, one-step planning, and the equilibrium test. Moreover, the reader can gain an insight into the three-limbed robot's simulation in MATLAB.
The final big topic will be about an area of artificial intelligence called deep learning and neural networks. In this chapter there will be a review of the basics of deep learning and neural networks, and a realization of a deep learning based algorithm in MATLAB, which supports the double-link robot's motion planning.