The present work focuses on the development of a pneumatic gecko robot capable of moving on surfaces with different inclinations, using linear actuators. The main applications are inspections, cleaning and repairs to buildings, ship hulls and large solar panel structures. The work relies on the design of vacuum claws using mechanical and pneumatic concepts to escape the high expense of compressed air provided by the common vacuum components on the market. In addition, the present work applies the neural network algorithm called backpropagation with tests on the prototype to obtain a computational learning of the robot's operation, after observing the behavior of the actuators through simulations with a kinematic model.
