With the development of intelligent vehicle driving, the vehicles must be trained through deep learning method. Therefore, the vehicles will be driven on real roads and subjected to traffic and environmental conditions, such as traffic lights, traffic jams, rounds, stops, curves, road condition and gradient. The on-road driving tests make it difficult to predefine beforehand the route, since the conditions are changeable and unpredictable. On the other hand, it is quite time-consuming and expensive. However, the generation of methods and algorithms able to read the main important parameters from the route (speed and gradient profile) would give a twofold benefit: generation and optimization or appropriate routes considering the testing objective, and if a powertrain model available, simulation of vehicle performance and emissions. In such a context, this thesis aims to generate real driving speed profiles based on either Google Maps guidance or GPX experiment data, considering speed limit, curvature and traffic condition. Real driving cycle generation from the on-line platform Google Maps and GPX data shows advantage over traditional experimental data characterization in aspects of cost, universality and convenience. Based on this approach, it is possible to analyze the influence of driver characteristics, curvature and traffic condition on driving behavior, emissions and fuel consumption, supporting in the phase of intelligent vehicle strategy development. The generated profile is compared to real driving recorded GPX data and is verified to be generally realistic.