Vehicle driving simulation, collision detection, and collision simulation of rigid bodies are not new in their corresponding literature, but the integration of all of these techniques is a challenging and interesting topic. Some special requirements arise when they are combined, especially when multiple vehicles, located at different places on a network, are involved in collision simulation. This thesis implements a network-based vehicle collision detection and response simulation system. This system has all the components that are required by vehicle driving simulation. It supports vehicle-to-scene and vehicle-to-vehicle collision detection and response simulation in real-time required by human-in-the-loop driving simulation. Additionally, it supports collaborative driving simulation for multiple vehicles in the same virtual environment operated from different physical locations. It provides consistent and realistic collision response for vehicles that collide. A network-based collision server is developed to accommodate vehicle-to-vehicle collision detection and response simulation. A general collision algorithm supplies consistent collision result for all collided vehicles. The method takes advantage of Open Scene Graph's (OSG) built-in collision detection functionality for vehicle-to-scene collision detection. For vehicle-vehicle collision detection, a two stage process is introduced which employs a bounding circle localization technique and Cohen-Sutherland clipping. A simple but realistic vehicle-to-vehicle and vehicle-to-scene collision simulation algorithm is developed with a friction model and a modified coefficient of restitution model, based on a vehicle collision simulation algorithm presented by Macmillan. Implementation testing shows that the network-based collision simulation system can provide a real-time, realistic, and robust system in a network with relatively small time lags delay, such as a LAN, a city network, and some inter-city networks. The implementation has demonstrated support for simultaneous collision simulation with up to 32 vehicles operating at reasonable speed in local area network.