Pneumatic artificial muscles for human-friendly mobile service robots

Abstract

Developing robotic systems which can operate in complex, human centred environments is a major challenge, but one which holds the promise of dramatically improving human living standards in several areas including manufacturing/assembly, the service industry, elder care and domestic assistance. In such applications it will be necessary for machines not only to work in the same space as people, but also to interact with them, often involving physical contact. This raises many concerns regarding the safe operation of such robots and how it can be guaranteed. Of crucial importance is the actuation method used to control such a robot. The actuation paradigms which have been established for industrial robots (high stiffness, accuracy etc.) are altered for this application, with inherent compliance, light weight, small size and energy efficiency becoming more significant. Pneumatic artificial muscles (PAMs) are a powerful class of actuator with highly advantageous properties for physical human-robot interaction. They can exert large forces, while being lightweight. They are also inherently compliant and can be manufactured relatively cheaply in a range of sizes. However, due to their inherent inefficiencies, the requirement for a compressed air supply, limited contraction and difficulties in modelling and control, they have not frequently been used for mobile service robots. The sleeve PAM addresses some of these shortcomings by incorporating a solid internal structure which reduces the volume which must be filled with compressed air while increasing the force output and contraction ratio. This thesis proposes design improvements to the sleeve PAM which is applied to the case of the McKibben type muscle. The modular nature of the design simplifies manufacture and makes it more adaptable for various implementations. The design also allows force output along the central axis of the PAM which simplifies system integration. Efficiency was also improved over that of the previous implementation due to a greater proportion of the volume being occupied by the internal element. In order to alter the force/contraction characteristics of the PAM an optional internal pulley mechanism was installed in the PAM which may be useful in applications where space is at a premium. A dynamic testing apparatus was constructed in order to further characterise and model the sleeve PAM under a variety of operating conditions. The bandwidth of sleeve and traditional PAMs were compared. Under isometric conditions the pressure bandwidth with no contraction was increased by approximately 100%. When the muscle contracts this improvement gradually reduces as the proportion of the volume of the PAM which is occupied by the internal element decreases. When the force bandwidth of the actuator is tested under similar conditions there is a further 20% increase in bandwidth across the range of contraction due to the lower pressure required to achieve the same force output. Testing with a constant force profile as the PAM contracts showed a more modest increase in bandwidth. A phenomenological dynamic model of the traditional and sleeve PAM was constructed based upon a popular existing model in order to further understand the differences between both muscles and facilitate control. This model was fitted to the PAMs using both static and dynamic tests. An empirical model of the PAMs? volume was also constructed based on the measurement of pressure change in the PAMs as they contract or elongate under closed system conditions. Combining these models with an established model of gas flow through control valves it was possible to accurately predict the behaviour of both the PAMs. The use of sleeve PAMs in an antagonistic configuration to operate a revolute joint is also investigated. Comparing a traditionally actuated joint with two configurations of a sleeve PAM it is shown that the torque can be increased by between 25% and 100% (depending on joint angle/direction) or the joint range can be increased by 14%. While isobaric stiffness is comparable, closed system stiffness is increased by up to 300% using sleeve PAMs. A theoretical energy analysis of the operation of PAM actuated joints is presented which show substantial energy savings when sleeve PAMs are used.