Design Optimization of Tree-type Robotic Systems Using Exponential Coordinates and Genetic Algorithms

Abstract

This paper proposes a new way to optimize the robotic systems using a combination of exponential coordinates and genetic algorithms. The exponential coordinates are an intuitive parametrization to describe the kinematics and dynamics of robotic systems, which only relies on a few parameters called the exponential parameters. The tree-type robotic systems composed of a set of serial link chains are considered in the design. In addition to the physical and the exponential parameters, we also optimize the joints connection (tree structure) by employing the chain matrix. These parameters can be efficiently coded into the strings of the genetic algorithm. Using the strings, we can conduct a simultaneous design optimization including the joint configuration (system geometry) and the joint distribution (system topology). A design example of a 2D arm-hand manipulator system illustrates the proposed method.