Actuator design for safe human-robot interaction
Traditionally robots have employed electromechanical and hydraulic actuation techniques. These work particularly well in the majority of industrial settings, where they can deliver large forces with impressive positional accuracy. However, in the case of robots which operate in close proximity or in cooperation with humans they raise a number of difficulties. Their positional accuracy is largely due to high stiffness a property which greatly increases the likelihood of injury should the robot come in contact with a person. A more compliant system would be preferable, whereby the actuators themselves can react to externally applied forces, such as a collision and limit the force imparted. A compliant system is also advantageous in unstructured environments, such as a home, where an accurate model of the environment is not available. For tasks involving sweeping or wiping, the ability of the actuator to maintain a constant force output regardless of deviations in the surfaces encountered allow it to operate effectively when the surface properties are unknown. For these reasons, pneumatic artificial muscles have been explored for use in mobile service robots.
Pneumatic artificial muscles (PAMs) have operating characteristics similar to that of human muscles. They contract longitudinally and are frequently used in antagonistic pairs. They consist of a tubular elastic membrane surrounded by a braided mesh. When the tube is inflated it pushes out on the mesh, increasing its diameter. This in turn causes the muscles length to contract, exerting considerable force. Due to the compressibility of air, these muscles are inherently compliant and behave like non-linear springs, the stiffness of which may be adjusted by varying the air pressure. They are also lightweight, relatively inexpensive and can easily be adapted to a range of sizes as the application demands. While these muscles have been used on a number of robots and certain industrial applications, the requirement for a large compressed air supply has hindered their adaption on mobile platforms. Our work aims to increase the efficiency of PAMs thus making them a more compelling proposition.
We have developed a novel design for a sleeve pneumatic muscle. A sleeve muscle contains a solid central cylinder which reduces the internal volume of the muscle. This in turn reduces the mass of compressed air which must enter the muscle to achieve the required pressure and therefore increases the efficiency. The central cylinder also reduces the area over which the pressure in the muscle acts upon the muscle ends, increasing the force output of the muscle. Using the central cylinder as a structural member of the actuated system, the total volume occupied by the system can be reduced. The sleeve muscle design we have implemented builds on previous sleeve muscle embodiments to deliver increased reduction in muscle internal volume. We have focused on retaining the ease by which a muscle may be adapted for different lengths and use cases, by keeping our design as modular and reconfigurable as possible. The addition of an internal pulley mechanism has allowed the muscle to deliver increased contraction with reduced force which is useful in compact systems. We have also incorporated a pressure sensor in the muscle to enhance the accuracy of this important control variable.
Tests have been conducted to establish the static and dynamic characteristics of the sleeve muscle and compare them to a standard muscle. This has confirmed sizable increases in muscle force output, contraction and efficiency. Additionally, the speed at which the muscle pressure may be varied is increased as less mass of air must flow through the control valves. This in turn allows muscle force and contraction to be varied more quickly. Tests have been conducted on muscles operating independently and in an antagonistic configuration to investigate this. The sleeve muscles have been used in a novel arm design, which makes use of additive manufacturing techniques to decrease manufacturing time.