Gait generation for a six-legged walking machine

This thesis focuses on the study of a six-legged walking machine. It consists of a chassis and six legs with three revolute joints each. Instead of a feet, each leg features a wheel, so that it is not only capable of walking but also driving. Furthermore, the robot is equipped with a 5-DOF arm and a gripper. The robot uses an Arduino platform for low-level control of the drives and an x86-64 computer for high-level tasks such as flow control, image processing, navigation or obstacle recognition. The aim of this work is to implement a gait generation algorithm in a six-legged walking machine and simulate the behaviour of the system. First of all, a thorough survey of literature considering static gait of multi-legged walking machines, planning-based and reactive controllers needs to be conducted. After deciding on a gait generation approach, the corresponding algorithms will be adapted to the specifications of the robot and implemented for simulation. A kinematic model of the robot is derived. A dynamic model is not needed since we focus on gait planning for statically stable walking. There are two possible modes of operation: remote controlled and autonomous. In the former mode, the robot is controlled from a base station with a gamepad and movement commands are transmitted to the mobile robot. We think that the easiest way to control the robot is to assign the curvature of the desired path (turning radius) and the orientation of the robot with respect to the motion direction (crab angle). In fact, only two inputs are needed in order to perform any general trajectory. In order to achieve the best result for the gait planning, some different procedures are developed. Their performances are compared considering both the stability of the motion and the computation time. Finally, once the gait is generated, inverse kinematics is used to determine motors motion law.