Pervasive computing envisions the achievement of seamless and distraction-free support for tasks by means of context-aware applications. Context can be defined as the information which can be used to characterize the situation of an entity such as persons or objects which are relevant for the behaviour of an application. A context-aware application is one which can adapt its functionality based on changes in the context of the user or entity. Location is an important piece of context because a lot of information can be inferred about the situation of an entity just by knowing where it is. This makes location very useful for many context-aware applications. In outdoor scenarios, the Global Positioning System (GPS) is used for acquiring location information. However, GPS signals are relatively weak and do not penetrate buildings well, rendering them less than suitable for location estimation in indoor environments. However, people spend most of their time in indoor locations and therefore it is necessary to have location systems which would work in these scenarios. In the last two decades, there has been a lot of research into and development of indoor localization systems. A wide range of technologies have been applied in the development of these systems ranging from vision-based systems, sound-based systems as well as Radio Frequency (RF) signal based systems. In a typical indoor localization system deployment, an indoor environment is setup with different signal sources and then the distribution of the signals in the environment is recorded in a process known as calibration. The distribution of signals, also known as a radio map, is then later employed to estimate location of users by matching their signal observations to the radio map. However, not all the different signal technologies and approaches provide the right balance of accuracy, precision and cost to be suitable for most real world deployment scenarios. Of the different RF signal technologies, WLAN and Bluetooth based indoor localization systems are the most common due to the ubiquity of the signal deployments for communication purposes, and the accessibility of compatible mobile computing devices to the users of the system. Many of the indoor localization systems have been developed under laboratory conditions or only with small-scale controlled indoor areas taken into account. This poses a challenge when transposing these systems to real-world indoor environments which can be rather large and dynamic, thereby significantly raising the cost, effort and practicality of the deployment. Furthermore, due to the fact that indoor environments are rarely static, changes in the environment such as moving of furniture or changes in the building layout could adversely impact the performance of the localization system deployment. The system would then need to be recalibrated to the new environmental conditions in order to achieve and maintain optimal localization performance in the indoor environment. If this happens regularly, it can significantly increase the cost and effort for maintenance of the indoor localization system over time. In order to address these issues, this dissertation develops methods for more efficient deployment and maintenance of the indoor localization systems. A localization system deployment consists of three main phases; setup and calibration, localization and maintenance. The main contributions of this dissertation are proposed optimizations to the different stages of the localization system deployment lifecycle. First, the focus is on optimizing setup and calibration of fingerprinting-based indoor localization systems. A new method for dense and efficient calibration of the indoor environmental areas is proposed, with minimal effort and consequently reduced cost. During calibration, the signal distribution in the indoor environment is distorted by the presence of the person doing the calibration. This leads to a radio map which is not a very accurate representation of the environment. Therefore a model for WLAN signal attenuation by the human body is proposed in this dissertation. The model captures the pattern of change to the signal due the presence of the human body in the signal path. By applying the model, we can compensate for the attenuation caused by the person and thereby generate a more accurate map of the signal distribution in the environment. A more precise signal distribution leads to better precision during location estimation. Secondly, some optimizations to the localization phase are presented. The dense fingerprints of the environment created during the setup phase are used for generating location estimates by matching the captured signal distribution with the pre-recorded distribution in the environment. However, the location estimates can be further refined given additional context information. This approach makes use of sensor fusion and ambient intelligence in order to improve the accuracy of the location estimates. The ambient intelligence can be gotten from smart environments such as smart homes or offices, which trigger events that can be applied to location estimation. These optimizations are especially useful for indoor tracking applications where continuous location estimation and accurate high frequency location updates are critical. Lastly, two methods for autonomous recalibration of localization systems are presented as optimizations to the maintenance phase of the deployment. One approach is based on using the localization system infrastructure to monitor the signal characteristic distribution in the environment. The results from the monitoring are used by the system to recalibrate the signal distribution map as needed. The second approach evaluates the Received Signal Strength Indicator (RSSI) of the signals as measured by the devices using the localization system. An algorithm for detecting signal displacements and changes in the distribution is proposed, as well as an approach for subsequently applying the measurements to update the radio map. By constantly self-evaluating and recalibrating the system, it is possible to maintain the system over time by limiting the degradation of the localization performance. It is demonstrated that the proposed approach achieves results comparable to those obtained by manual calibration of the system. The above optimizations to the different stages of the localization deployment lifecycle serve to reduce the effort and cost of running the system while increasing the accuracy and reliability. These optimizations can be applied individually or together depending on the scenario and the localization system considered.