Development of the control system for an inflatable robot = Development and Implementation of a Control System for an Inflatable Robot

Since the advent of robotics, manipulators have been made of rigid materials, and most developments in the field have pursued increasingly precise and dynamic robots. However, in recent years there has been a growing interest in soft robotics, which has led to the development of new mechanical structures based on flexible and deformable elements. Thanks to these properties, soft robots are able to perform new tasks in new fields of application, ranging from non-invasive surgical techniques to the manipulation of delicate and irregular objects.One of the fields in which soft robots can excel and bring significant benefits is space application: lightweight robotic systems and the use of flexible and deformable elements allow the manipulator to be easily contained in a small package and deployed when required. Such approaches also offer an increase of the payload-to-weight ratio, making it possible to reduce the cost of getting a manipulator into orbit or onto the surface of another celestial body such as the moon or Mars. Despite the promising properties, there are particular criticalities that make the development of soft manipulators challeging: great attention is paid to the study of innovative materials to create flexible structures and actuators or to new control approach to such structures. Indeed, both the modelling and control of rigid robotics cannot be applied successfully to soft robots because of their highly non-linear dynamics. The goal of this thesis is to develop a control system for the 3-degree-of-freedom manipulator, the POPUP Robot, whose mechanical structure has been designed and developed by the research group of professor Stefano Mauro. The manipulator is an anthropomorphic arm with electrical actuation, where the typical rigid link have been replaced by inflatable cylinders, made out of PVC. It therefore presents complex kinematic and dynamics equations that are difficult to obtain analytically. The proposed controller is able to calculate the amount of deformation of the inflatable links by processing the measurements of a set of sensors opportunely placed along the robot structure. The control algorithm is able to counteract the deformation both in equilibrium conditions (point-to-point applications) and during the execution of planned trajectories. Two methods are then explored: the first method aims at correcting the positioning error due to link deformation through a linear regression on the sensor data. The second method makes use of an Artificial Neural Network to counteract the non-linearities of the system and to make the control independent of the kinematic and dynamic model of the manipulator. Finally, the two methods are compared, both in terms of performance (positioning error, trajectory error) and in terms of computational load: the system is in fact entirely executed by a microcontroller and it is therefore fair to compare the computational effort required to execute the control algorithms.