This thesis contributes to multiple stages along the typical lifecycle of broadband wireless networks by introducing novel concepts, models, and algorithms for the dimensioning, the planning, and the self-organized operation of 4G radio networks.First, the key components for wireless network modeling are introduced. Using path loss information as input, we show how the required bandwidth for data transmission can be computed via the discrete rate-power function that is given by the system link budget. Since the spectrum in OFDMA multi-cell networks is shared and limited, all developed optimization models are subject to an underlying multiple knapsack problem if the users expect a minimum quality-of-service and the corresponding rate demand exceeds the load limit of single cells.We consider multihop WiMAX networks and determine the infrastructure dimensioning with respect to the expected user distribution and rate demand. The related optimization problem is formalized as a mixed-integer linear program that covers all relevant technical system aspects. We show how an economical perspective can help to find a closed-form representation for conflicting objectives like the trade-off alignment between network coverage, network capacity, and deployment cost.Compared to the dimensioning approach, accurate cell site planning and network configuration require a higher precision, particularly in terms of the applied inter-cell interference model. For this purpose, a low-complexity interference approximation is developed that estimates the overall required bandwidth at eNBs and HeNBs subject to inter-cell and cross-tier interference. This approximate model serves as immanent component of the optimization models that we introduce for the planning and the self-organized operation of LTE HetNets. The interference approximation can be calibrated with respect to the dynamic changes in the network.For the optimal cell site planning of LTE HetNets, we consider macrocells and user-operated femtocells that are not necessarily active all the time. The objective of the corresponding optimization problem is to provide a minimum number of macrocells such that mobile services are area-wide guaranteed. On the other hand, it avoids dispensable cell sites for the sake of cost efficiency and low interference.Finally, we present an integrated approach for the self-optimization of coverage and capacity in time-variant LTE HetNets. The corresponding algorithms are designed according to a traffic light principle and they autonomously control site activity, transmission power, and antenna downtilt parameters. The presented approach finds Pareto optimal solutions for the joint optimization of coverage and capacity. It confines to a hierarchical constrained single target maximization in the case that a single performance metric predominates. Simulative evaluations demonstrate that our approach improves the overall performance when the HetNet has to cope with traffic hotspots, coverage shortage, and objective trade-off situations.