Computational dynamics for robotic systems on land and under water

Abstract

Motivated by the need for more complex robotic systems to accomplish more difficult tasks such as hazardous waste cleanup, assembly operations, and deep-sea construction and exploration, this dissertation presents the development of efficient dynamic simulation algorithms for multiple chain robotic systems on land and under water. These algorithms can be used in the development of control algorithms, and with real-time rates, hardware- and human-in-the-loop applications also become possible. These uses of simulation can significantly reduce the cost of design, development, and operation of complex robotic systems. In Part I, land-based systems are examined which include multiple manipulator systems and multilegged vehicles. For multiple manipulator systems, an efficient closed chain dynamics algorithm based on the Composite Rigid Body (CRB) method is presented. To increase computational rates in an effort to achieve real-time, temporal and spatial forms of parallelism are implemented, and algorithms that are robust in the presence of manipulator singularities are developed. Then, the decoupled tree-structure (DTS) approach is used to develop a new efficient, CRB- based algorithm for legged vehicles on land.

In Part II, the development of a real-time simulation system for underwater robotic vehicle (URV) systems is presented. Hydrodynamic forces on submerged rigid bodies are investigated, and an efficient dynamic and hydrodynamic simulation algorithm based on the Articulated Body (AB) method is developed. Finally, an implementation of this algorithm capable of simulating a general class of tree structured mechanisms having star topologies is undertaken. Object oriented design techniques such as object hierarchies, encapsulation, inheritance, and polymorphism are applied to this task and a general but very efficient implementation results.