CAREER: Toward Unified Nonlinear Dynamics and Control of Robot Systems

点击翻译按钮获取中文摘要

9625143 Whitcomb This program of research pursues two complementary ideas toward the synthesis of robots possessing exceptional yet flexible dynamical dexterity: (I) precise physical models and (ii) adaptive learning. The first idea is to incorporate precise physical models of the robot and its environment into a robot's computer control system. By knowing the full physical dynamical model of its limbs, a model-based robot controller can automatically compensate for the nonlinear dynamical effects (e.g., inertial, Coriolis, friction, hydrodynamic, and gravitational forces) acting on the robot during motion. The second idea is for robots to quantitatively learn the parameters of their own physical structure and their environment, and to adaptively compensate for changes in these parameters. The result is superior speed, precision, and dexterity. Theoretical problems in robot arm control (Section 1.3) are remarkably similar to the problems of underwater vehicle position (Section 1.4) -- both industrial robots and underwater vehicles are second-order Lagrangian systems exhibiting highly structured nonlinear dynamics. These ideas are to be employed in these two real-world problem domains: Nonlinear Adaptive "Learning" Control for Industrial Robot Arms Nonlinear Model-Based Control of Underwater Robot Vehicles Education Innovation in Robotics Teaching ***