A study on the design of quadruped walking robot and its push recovery strategy

Abstract

A reliable and robust quadruped robot platform that enables the implementation of stable and fast static/dynamic walking on even or uneven terrain and walking while carrying a payload is developed. Developed robot platform has 12 degrees of freedom and its height, length and width are 0.86m, 0.8m, 0.34m respectively. It weighs 42kg with a battery. The robot consists of one trunk and four equal legs and the legs are mounted symmetrically on the corners of the trunk. The legs have the same kinematical structure and each leg has 3 active d.o.f. : hip roll (abduction/adduction), hip pitch and knee pitch (flexion/extension). A lightweight but highly rigid platform is achieved via a double supported beam type joint design and trunk frame design using aluminum alloy tube. A small, light but high output power joint actuator is designed with a harmonic reduction gear and electric motor which has high speed and output power. Because hip pitch and knee joint need more power while robot walks, two motors are allocated on the hip pitch and the knee joint. The details of mechanical design and concepts for achieving design objectives effectively are described in this paper. A distributed control system is used since it can reduce the computational burden of the main computer and easily extend the electrical components. Microprocessor-based sub-controllers are developed for motor control and sensor signal feedback. The main computer which is mounted on the trunk, communicates with the sub-controller via the CAN (Controller Area Network) protocol. We used Windows XP as an OS and have established a real-time control system in Windows XP by using RTX (Real Time eXtension) software. In order to demonstrate the performance of the robot platform, simple experiments were implemented. Forward/backward walking pattern is designed considering a quadruped mammal`s walking motion. Also, we develop the push recovery strategy for dynamic walking quadruped robot. Because trot gait always dr...