Over the past few years there has been an dramatic increase in mobile data tra c demand, a trend that is expected to continue in coming years. Traditional macrocellonly networks are incapable of providing the quality of service that modern subscribers expect from a mobile broadband service. Increasing network densi cation with the deployment of low power base stations has proven to be an e ective solution in this regard. The resulting multi-tier topology is known as heterogeneous networks or HetNets. This new topology brings a series of new and important challenges, since traditional practices applied for macrocell-only networks no longer provide optimal results. There is a need to increase the understanding about the operation of these systems and develop new techniques to properly plan, design and optimize HetNets. These new techniques should focus on the e cient use of resources during network planning, reducing costs of deployments, and facilitating the con guration and maintenance of HetNets. This thesis has focused on exploring novel solutions to challenges in two main areas regarding the operation of HetNets: planning and self-optimization. Regarding the planning of HetNets, the thesis starts by treating the issue of improving the accuracy of site-speci c path loss prediction models for outdoor microcell deployments. The prediction of coverage areas based on path loss estimations are essential for network operators during the planning and design of new deployments. The thesis proposes two novel tuning algorithms intended to optimize the propagation model parameters based on information from a limited set of physical measurements. Also in the area of network planning, it is fundamental for network operators to understand typical user mobility patterns and accurately estimate the quality of the service as users move. For this purpose system level simulations are typically carried out. This thesis proposes a downlink system level simulator that incorporates a mobility model as well as a tra c model where users are categorized according to their type of demand. We were able to demonstrate that an appropriate tra c model can signi cantly increase the accuracy in the estimation of the the user experience. Regarding the self-optimization of HetNets, the thesis treats two key challenges: load balancing and the optimization of handover parameters. Proper load balancing among base stations is fundamental in order to leverage the bene ts in network capacity that HetNets can provide. In this thesis a novel and practical load balancing algorithm is proposed. With this algorithm, each base station can solve locally a load-aware utility maximization problem. As opposed to current approaches, the algorithm minimizes the required level of coordination among base stations, hence reducing the impact on the signaling load of the network and potentially reducing the e ect on power consumption. Finally, the thesis proposes a novel methodology to optimize handover parameters for in-building systems. The goal of the methodology is to minimize handover failures and the triggering of unnecessary handovers, while maximizing the quality of service provided to users approaching the cell-edge. With this methodology, a base station can customize the handover parameters according to the current load level and the speci c radio frequency conditions of the cell-edge that a user will experience as it moves out of the service area. With the research work described in this thesis, we have expanded the understanding about the operation of HetNets. The algorithms and methodologies proposed in this thesis have the overall objective of maximizing the bene ts that HetNets can provide through the e cient use and coordination of the resources in every tier.