Distributed dynamic system simulation using a computer network

Over the years, much work has been done in the area of multibody simulations. Some examples of such work are found in the areas of robotics and biped locomotion. Complex multi-body systems can be simulated with systems of connected microprocessors at a fraction of the cost of using mainframe computers.;Simulation of linked rigid body dynamics is well suited for multiprocessor systems. The dynamic equations for a single body can be replicated to simulate a multiple body system of any number of components. When such a system is implemented on a group of connected processors operating concurrently, a large speed-up in processing time occurs.;Experiments were performed in implementing single-link and two-link planar multi-body systems using the Newton-Euler formulation of dynamics on a two-computer network. These experiments showed that dynamic simulations are feasible over such a network. The most important factor for this feasibility is the simulation implementation through state space equations which provides minimum communication when only state vector information is transferred. Two other important factors are the choice of a communications protocol which provides for reliable data transmission and numerical integration techniques which perform well when a single processor does not have immediate access to all state vector components.;To show the effects of load distribution involving larger systems, mathematical simulations of networks of up to five processors implementing a dynamic simulation of a planar nine-link body were performed. These simulations show that load distribution produces faster simulations given a sufficiently fast communication network. A technique for analyzing the load distribution problem over different types of architectures is presented. This technique is illustrated as it is applied to the nine-link simulation on fully-connected and star computer networks.