Industrial machines with hydraulic arms such as backhoes typically rely on levers to control the motion of individual joints, which are coordinated to actuate an end effector. Professional operators require training and experience to become proficient. Teleoperation and autonomous control of hydraulic manipulators are difficult to achieve due to the inherent nonlinearities and complexities of hydraulic systems. In this work, we lay out an approach for simultaneous joint control of a hydraulic manipulator developed for humanitarian demining. The approach allows the end effector’s trajectory to be controlled directly by joystick or commands issued via a virtual reality interface. Direct control of the end effector is preferable to manual control of individual joints because of the precision required and delicate nature of demining. VR provides a means for the operator to view the shape of the terrain as well as the location of landmines and other objects buried beneath the surface in a single, composite image, and the manipulator’s position relative to these objects. We use a redundant manipulator, having three degrees of freedom in its planar portion. Velocity control is used for the planar joints and position control is used for the end effector pitch. A solution is presented for the inverse Jacobian problem for avoiding physical joint limits and maintaining manipulability. Neural network deadzone compensation ensures the planar joints begin to actuate simultaneously to prevent error in the initial motion of the end effector.
抄録全体を表示