An Extension of the MB-RRT Method to Solve the Inverse Kinematics Problem for Real Robot Manipulators

In previous work, a solution to the inverse kinematics (IK) problem of manipulator robots, called the Manipulator-Based Rapidly-exploring Random Tree (MB-RRT), was proposed. Unlike the classical RRT, it grows the tree as a spatial representation of the manipulator in the workspace rather than in the configuration space, reducing the search space to three dimensions. The technique also introduces a new probability model for sampling new nodes and a novel metric for finding the nearest nodes. The two main advantages of the method over other IK solvers are its ability to handle obstructed environments easily and the invariance of its search space with respect to the number of joints in the robot. However, the original method is not suitable for robots with one-dimensional joints operating in three-dimensional space, which is common in real scenarios. This paper proposes an extension of the method to address these scenarios. The extension retains its effectiveness in handling obstructed environments and the invariance of its search space to the number of joints. Furthermore, the proposed solution obtained high success rates in the IK problem of two robots, 6-Dof and 7-Dof, in both obstructed maps and a obstacle-free map.