| In a complex interactive training or simulation system, a large number of simulation entities and a large set of geographical data are involved. Networked virtual environments are being built to provide the capability, within realtime graphical simulation, of reconstructing landscape architecture or rearranging the terrain surface. Under those actions, objects must be faithful to the laws of physics and behave appropriately in response. The geographical database and simulation entities are desired to be partitioned and distributed across loosely coupled networks of simulators to meet requirements of low cost and high efficiency.; This research work is dedicated to resolutions of the above issues. Physically-based soil models are established to provide physical realism of a realtime simulation. Distribution strategies are also presented and estimated for performances of systems. The results of the research are summarized in two parts. For physically-based models, soil properties and behaviors are investigated. Realtime graphical models for soil slippage and manipulations are introduced to simulate excavating activities on the dynamic terrain. For distributed computations, approaches for distributing terrain databases and simulation entities across loosely coupled networks of simulators are proposed. An analytical model is then defined, which can be used to estimate quantitatively expenses of a distributed system. Experimental simulation study is also conducted to provide a verification of the correctness of the analytical method. |
