PID controllers are widely used in industry. While this may be appropriate for many systems, eventually a more complex or reliable algorithm has to be designed to improve performance. Common praxis is to take advantage either of physical properties (e.g. passivity) or of a mathematical model. For this last case it may prove to be hard to get an accurate description of the system dynamics. In this article we experimentally analyse the behaviour of pneumatic actuators by employing control algorithms available in the literature with little or no model information at all, including an adaptation of a robot control law which is shown to work very well for the test bed.