(A) new approach for solving the inverse kinematics of robot manipulators using incremental unit computation method

Abstract

In today``s intelligent robots, one of necessary features is an adaptability to cope with unknown working environments. This calls for on-line path modification based on various sensory informations from the latter. However, since robot manipulators are usually servoed in the joint level, there is a need to transform the sensory informations in terms of Cartesian coordnate into the joint space. The inverse (Cartesian-to-joint) transformation involves an extensive computation, thus the computational requirement for real-time processing of it places a heavy burden on computational resources. This dissertation proposes a new efficient method to solve the inverse kinematics for robot manipulators in real-time, termed incremental unit computation method. First, the major linkage of anthropoid manipulator is considered as the combination of two types of 2 DOF robot. For each 2 DOF robot, the method difines incremental units in joint and Cartesian spaces, which represent the position resolutions in each space. By introducing these units, the repetitive evaluation of inverse Jacobian matrix can be realized through a simple sign table, which can be readily implemented in a combination of TTL logic gates. Furthermore, the direct kinematics can be solved by using DDA integrators at high speed. Through an iterative comparison of the incremental units with convergence rules, the inverse kinematics algorithm is established. The proposed algorithm has many attractive features from the hardware implementation viewpoint. For a drastic speed-up of solution rate, a hardware architecture that implements the algorithm has been proposed and realized by adopting an EPLD. In order to demonstrate the feasibility of the implemented hardware, a series of experiments are conducted. The experimental results show that sophisticated tasks such as telerobot control, Cartesian path control, and compliance control can be easily accomplished through the hardware.