The mechanical design of legged robots

This thesis has focused on the mechanical design of statically stable multi-legged CLimbing And WAlking Robots(CLAWAR) machines and, since crab-like machines represent an important sub-class, in this thesis they were chosen as the subject of the study. Through the review of leg mechanisms, it is clear that for many years, navigation, gait generation and control, rather than mechanical design, have been the main concerns of many CLAWAR researchers. During the development of prototypes, it has often been assumed that the mechanical design principles are known and the researchers' jobs are simply to apply them. In practice, this is far from the truth, as the performance of existing prototypes testifies. The most common design approach is to copy the geometry of insects and mammals with little or no scientific justification. Although there has been some very good work on leg mechanism design, the relationships between leg design and overall machine layout have been neglected. In this thesis, the design is considered as a whole with no artificial decoupling of leg geometry and overall machine geometry. Furthermore, the design process is treated as a series of coupled optimisation problems. The main achievements and conclusions that have resulted from the research described in this thesis can be summarised as follows: • Based on the review on leg mechanisms, a classification on legged machine layout was proposed; • A clear understanding of the effect of leg configurations and geometric design parameters on the performance of crab-like CLAWAR machines has been achieved. • A novel design methodology that breaks the problem into: a) kinematic design; and b) performance optimisation, was presented. The design methodology is based on satisfying kinematic requirements (constraints) and optimisation of kinetic performance measure, such as minimising the joint torques. • Although the design methodology has only been applied in the 2D case, it has been shown that it could be applied in the 3D case and the necessary analysis methods have been established. • Methods for using foot force distribution as well as design to optimise performance were developed. • Novel reformulations of the Moor-Penrose pseudo-inverse for optimising the foot force distribution in the 3D case were developed, which could be applied in real time control as well as in design.