Predicting the motion trajectories of a modular snake robot performing various gaits.
Thesis DisciplineMechanical Engineering
Degree GrantorUniversity of Canterbury
Degree NameMaster of Engineering
Some studies have demonstrated that snake robots have more advantages than other mobile robots in real-world applications such as search-and-rescue operations and building inspections. A pitch-and-yaw modular snake robot is capable of performing various gaits such as linear progression, rolling, sidewinding and rotating. These gaits allow the robot locomotion to adapt to different types of terrain. Predicting the motion trajectories of the gaits is necessary in order to control the robots in path planning.
This thesis presents the pitch-and-yaw modular snake robot developed as a robotic platform to study snake robot locomotion. The joint control of the robot was optimised and optimal gait parameters of the sinusoid function generators, which are used to generate the gaits, were obtained. The reprojection errors of the motion capture, which was used to track the robot's joints, were determined.
The accuracy of the 2D and the 3D simple kinematic motion models, which are based on the virtual chassis, were analysed using the motion trajectories recorded from the laboratory and the ODE simulation. The models were optimised to find the average optimal ground parameters of the models by training the models with the experimentally recorded motion trajectories. The consistency in the motion trajectories for each gait was evaluated to determine the predictability of each gait. The results are presented in terms of errors in the travelled distance and direction, and the orientation of the robot. The 2D model predicted the motion trajectories of the linear progression gait and the turning gait with highest average errors of 5.4% and 10.1%. The 3D model achieved similar outcomes for the rolling gait and the rotating gait but with highest average errors of 16.5% and 10.2%. However, the failure of both the 3D model and the simulation to predict the sidewinding gait might be due to the large inconsistency in the motion trajectories of the gait and the lacking number of modules. The use of average optimal ground parameters, reprojection errors and inconsistency in the motion trajectories might have contributed to the errors of the models.