Unstructured terrain adaptive navigation of self-reconfigurable quadruped robot

Abstract

This work formulates adaptive navigation for a quadruped robot by reconfiguring its 5-DOF limbs according to analysis of the terrain's topographic features. Four issues are contributed: a measure of the terrain's complexity useful for the robot; an automatic polynomial path generation approach based on the safest local route; automatic selection of the legs' configuration; and nearly holonomic trajectory tracking control. The process of limbs' self-configuration consists of changing the robot's linkage for suitably navigation according to the unstructured terrain's complexity. A formulation to track the safest local route combined with a set of navigational functions: trigonometric, exponential or polynomials is proposed. For determining terrain and route complexities, 5 factors were analysed: curvature of the generated route, quantification of obstacle's height, surface emptiness, terrain slopes, and ground hardness. The novel limb linkage allows to adopt different kinematic configuration such as wheels, legs, and mixed types. This work describes mathematical formulation validating feasibility and reliability of terrain's evaluation for adaptive navigation and control by simulation results.