Font Size: a A A

Intuitive Control Based On Virtual Model For Quadruped Robot Trot Gait

Posted on:2016-10-19Degree:DoctorType:Dissertation
Country:ChinaCandidate:H X XieFull Text:PDF
GTID:1108330509960980Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
Legged robots have higher adaptive capacity and higher motility on non-structural terrains than wheeled and tracked robots. As an important category of legged robots, quadruped robots are advanced in stability and load-capacity. So they can quickly and stably navigat rough natural terrain while carrying significant payloads. These advantages make quadruped robots of great potential value in military field and civil use, and hence getting attention from more and more researchers and institutions. However, quadruped robot is a multi-input, multi-output nonlinear system which works toward unknown non-structural environment. It is very difficult to achieve totally auto-balanced control for a quadruped robot. One of the key problems in current quadruped researches is the basic algorithms for locomotion control, especially the control methods for dynamic balance in fast running. This difficulty is also the bottleneck that blocks the development of quadruped robots in our country. To find a simple and effective locomotion control approach for quadruped robots, this dissertation researches the intuitive control for qudruped trotting, with the main ideas and contents as following.(1) Based on the study of common used gaits for quadruped animals, a groundcontact based method for typical gaits’ planning is proposed. First, the common used quadruped gaits are introduced, with focus on the typical gaits of crawl and trot. Then Raibert’s method is adopted to calculate the target foothold of the swing foot, and the classic cycloid function is chosen to plan the swing trajectory. And then the sequence of leg motion is intuitively organized based on the ground contact signals of the feet. The logical sequence and judgement process of the crawl and trot gaits are planned, and the gait switch between the two gaits are also discussed. The planned trot gait here is the main gait for the research of this dissertation, supporting the related simulations and experiments.(2) Based on an intuitive method called virtual model control(VMC), a decomposed VMC method is proposed and extended to the implementation of qudruped trotting control in 2D space. In order to solve the problem of VMC implementation for qudruped control, here the trotting control is first studied in 2D space(sagittal plane). After the principle and features of VMC are introduced with details, the difficultes of VMC implementation for quadruped robot are analyzed. Then a decomposed VMC is proposed to individually control each leg and the VMC laws for stance phase and swing phase are respectively derived. Last, the simulation model and VMC controller are built, and, associate with the planned trot gait above, simulations of locomotion control are conducted to verify the proposed approach.(3) A full VMC approach in 3D space is proposed based on the decomposed VMC, focusing on the solutions of attitude control and lateral balance in quadruped trotting. In 3D space, the decomposed VMC law for the trunk’s translational motion control is first derived. Then a special attention is paid to the body attitude change in trotting, analyzing the feature of attitude change and its cause. And then an intuitive method, using the toruqe of the side-sway hip joint to balance the rolling motion of the body, is proposed for attitude control. By integrating the attitude control into the single-leg VMC, the complete stance VMC can be obtained. Combining with the swing VMC, the complete virtual model controller can be established. With the controller, trotting simulations in 3D space are run to test the effectiveness of the attitude control and lateral balance.(4) Aiming at the adaptive control of a quadruped robot on unperceived sloped terrain, four adaptive strategies are proposed to improve the robot’s capability of terrain adaptation. In order to quickly and efficiently go over the unperceived sloped terrain with trot gait, inspired by behavious of quadruped animals in nature, four intuitive strategies are proposed to solve the problems that the quadruped robot meets when on sloped terrain. These adaptive strategies are integrated with the above VMC intuitive method, and are compared and verified by simulations of locomotion control.(5) A real quadruped robot is designed and built, and the preliminary experiments with a single-leg system are conducted. To further demonstrate the proposed VMC intuitive control method with real robot, a small-size quadruped robot that actuated by electrical motors is built. The robot has light-weight and modular leg design, resulting in simple and compact structure and wide ranges of joint motion. Then a test platform based on the single leg module is built and tested by the proposed VMC method for the stance control and swing control.The decomposed VMC has very simple control laws and hence high computational efficiency, avoiding complex model derivation and heavy computation. The simulation results demonstrate that the proposed method can effectively control the dynamic balance of the quadruped while trotting, with certain capability in disturbance rejection and terrain adaptation. The experiments with the single leg also verify the effectiveness of the method for the stance and swing controls. In addition, the proposed approach is not limited by certain leg structure of a quadruped, making it easy to be extended for other quadruped robots. This dissertation provides a simple and effective solution for quadrupedal dynamic balance in trotting.
Keywords/Search Tags:Quadruped robot, Dynamic balance, Virtual model control, Intuitive control
PDF Full Text Request
Related items