Quadruped robot autonomous navigation and stair climbing

• Main Author: Wang, Jiong
• Other Authors: Chen I-Ming
• Format: Thesis-Master by Research
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Mechanical engineering::Robots
• Online Access: https://hdl.handle.net/10356/157938
• Institution: Nanyang Technological University

Quadruped robots are widely used in many scenarios because of their superior obstacle crossing and complex environment adaptability. The application of quadruped robots in the construction site gradually increases. However, it’s challenging to achieve climbing or walking on other unstructured terrains, which are common in construction scene. This report studies the autonomous navigation of quadruped robots and the planning and implementation of gait when climbing stairs, in order to enable quadruped robot move without unmanned operation and detect the environment of stairwells such as cracks on the stairway walls when climbing stairs. Firstly, based on the structural characteristics of the legged robot, a set of autonomous navigation programs that can run on the legged robot have been developed, and real-time tests have been completed to verify the autonomous navigation function. Secondly, it studies the gait of the quadruped robot. In order to meet the stability requirements, the step sequence of the quadruped robot is studied. Moreover, the kinematics model is established and the kinematics analysis is performed. By solving the forward and inverse kinematics equations, functional expression of every joint variable and the gait algorithm of the quadruped robot walking on flat ground were completed and the test was successful. The climbing simulation is achieved on Gazebo and field testing is also conducted after that. Future work will further improve navigation and the performance of climbing. For the navigation, multi-sensor fusion technology could be used to improve the accuracy of localization. For the robustness of climbing, the force control and trot gait might be taken into consideration. The control parameters may be further optimized by reinforcement learning, and various situations that may occur in the construction site scene may be considered to propose a response plan.