Transport assignment models play an important role in the research and practice of transport planning. Traditionally, transport assignment models are link-based, predicting traffic volumes on network links. However, route-based transport assignment models, which work with route sets and determine traffic volumes of routes, became more important in recent years. They support advanced analysis procedures that require route volumes as input, e.g. estimation of OD-matrices and evaluation of intelligent transport systems (ITS). The aim of this research is to develop a general framework for a route-based transport assignment, which emphasizes the quality of the resulting route sets. To achieve this goal, this framework differentiates the assignment procedure into two spate stages. In the first stage, it generates the route sets for all OD-pairs explicitly, and then in the second stage it distributes the travel demand between these routes. With this two-stage structure, a comprehensive route set generation procedure can be incorporated considering specific constraints. In order to generate realistic route sets, it is essential to have a clear definition of the quality of route sets. After examining the assumptions on the travellers’ behaviour in identifying and choosing routes, a set of indicators is introduced to quantify the quality of route sets. These indicators concern two aspects of the quality of route sets: the quality of individual routes and the composition of route sets. Constraints such as global detour factor, local detour factor, local optimal threshold for route sections and road hierarchical index are introduced to ensure the quality of individual routes while constraints such as choice set size and minimal dissimilarity address the composition of route sets. With these indicators in place, a two-stage route set generation procedure is developed to generate route sets whose quality satisfies certain constraints. The first stage employs the branch-and-bound technique to enforce the constraints on individual routes. For a given OD-pair, it enumerates all possible routes systematically and discards routes whose indicators do not satisfy the constraints. The second stage enforces the constraints on the route set composition. It produces route sets that contain a given number of routes where the routes are dissimilar with each other. Using the route set generation method a framework for distributing travel demand on given route sets is presented. It applies a nested logit model that takes advantages of the hierarchical structure of the route choice sets. For the feedback between route choice and route travel time it uses a simple convergence procedure based on successive average. A pilot implementation of this assignment framework is developed and some implementation issues such as network pre-processing are discussed. The pilot implementation is validated with a small test network and a larger test case in the Munich region. Here the generated route sets are compared with observed route sets. The result shows that the framework is capable of reproducing observed route sets with sufficient accuracy and that it is flexible enough to be calibrated for different situations.