It has become commonplace for mobile computing systems to be constructed using low-cost, commodity components for localization. While the expected error in consumer-grade sensors can still be acceptable for localization at human scale, all have fundamental limitations which manifest in different ways in different environments. Cooperative localization techniques can compensate for these hardware limitations, facilitating robust positioning in location-sensitive mobile applications.Four major challenges exist for developers of localized mobile systems. First, the nature of connectivity in mobile ad-hoc networks can often be highly sporadic, with frequent disconnects and changing topology. Second, sensor error frequently occurs in unexpected ways, is produced by multiple sensors working in tandem, or exhibits much different behavior depending on the type of sensor. Third, cooperative techniques for localization are difficult to implement effectively due to the lack of effective tools to measure spatial separation among mobile nodes. Finally, mobile applications that rely on localization of remote network nodes can fail without accurate and precise positioning.The foundation of this work is a rigorous evaluation and discussion of sensor error as encountered in practice. This work examines sensor error, which can manifest in unexpected ways outside of controlled environments and applications, focusing on its effect on human-scale localization. Data collected from both empirical measurement and outdoor exercises are used to construct error models, which are used to evaluate new ideas in mobile cooperative computing. A simulation environment for mobile ad hoc networks incorporating various models of localization error is presented.Next, the utility of sharing such location information among cooperating users is explored. Two methods are presented which can account for and reduce localization error using shared data, exploiting the independence of error among nodes in close proximity. A scenario-based evaluation approach is used to demonstrate possible techniques for using shared location information. System parameters required for effective utilization such data is also discussed. Simulation trials show that up to 50 percent reduction in overall localization error can be realized in many cases using only commercial-grade sensors.Finally, the effect of robust localization and error reduction at the application layer is studied. This is examined in the context of a new method for selecting available peers in a mobile network for the purpose of short-term data storage and retrieval. By weighting the utility of each remote node based on error metrics and the confidence level of those metrics, an increase in the effective availability of data in the system can be demonstrated.