NRI: Biologically-inspired, hybrid quadruped robot control

点击翻译按钮获取中文摘要

This research project is about designing better legs for legged robots. When a robot needs to move across rough terrain, as would be the case when searching for victims in a disaster area, legged robots have advantages over wheeled or tracked robots. They can traverse such terrain more easily, for example, stepping over fallen beams and bricks. However, legged robots are inherently more complex. They have many more joints and are more difficult to control. The goal of the project is to use ideas from nature to overcome some of these difficulties. The shape and compliance of four-legged animals adds to their stability passively when they trot or gallop. These ideas will be incorporated into a four-legged robot, making its locomotion abilities more robust. A major issue to duplicate from nature is the animal's ability to adaptively change the compliance of their joints and legs. This is a primary objective of this project. In addition to the research objectives, this project also includes educational components involving graduate and undergraduate students as well as outreach to secondary schools.The approach taken in this project will simplify the control of quadruped robots by the use of directionally compliant legs to passively stabilize the robot under dynamic gaits such as trotting. Active control of the stiffness and joint equilibrium positions will allow for adaptive parameter changes to changing terrain, gaits, and robot variation (carry a load for example). The system will employ a hybrid controller using passive, under-actuated legs during gaits, but actively changing the passive characteristics. Stability of the robot is measured by the pitch and roll of the robot body. The adaptive controller will minimize errors in the desired pitch and roll by changing the joint stiffness, equilibrium angles and positions of each joint.