The emerging area of "computer science in sport" and particularly the upcoming field of "ubiquitous computing in sport" are highly influenced by the progress of Information Technology (IT). In general, the current tendency of ubiquitous or also called pervasive computing describes the movement of information processing towards miniaturized, interconnected and intelligent computer devices as integral parts of everyday life. Given the continuous advances and recent boom in the mobile technology sector, in particular, nowadays also sports frameworks integrating pervasive equipment are in ongoing development. Especially mobile feedback systems are getting continuously widespread and important for the monitoring of sports performances. Sensor devices are becoming smaller, increasingly cable-free and, at the same time, smarter, enabling efficient methods for the acquisition of sports-related parameters. The diversity, capacity, networking ability and handy design of today's mobile devices, on the other hand, allow the implementation of effective supervision and instant intervention routines. The present thesis demonstrates the design and practical realization of an easy adaptable sports framework integrating innovative online analysis and real-time feedback techniques. A first goal of this work is to point out the implication of pervasive computing for sport and sport science and to demonstrate some of the most important developments. For such purposes, an overview of typical implementations and fields of application regarding the use of ubiquitous computing equipment is given and illustrated in detail by the realization of the developed mobile coaching framework, aiming at the instant support of athletes and coaches during the training process. The overall concept of the implemented system includes a server component for the bidirectional communication between sportsmen and coaches. Thereby, athletes are equipped with wireless sensors and an up-to-date handheld Personal Computer (PC) for data acquisition purposes, while coaches can access the measured information online and return appropriate feedback via the host computer to the athletes- devices. First implementations of the approach concentrate on the application in running and moreover for the use of pupils and teachers in school sport. The reception of the sensor data is based on ubiquitous technologies such as ANT TM (a wireless sensor protocol commonly applied in the field of sport)-enabled smartphones, facilitated either via an ANT Universal Serial Bus (USB) stick, a (mini/micro) Secure Digital (SD) card, an individually designed Bluetooth®-to-ANT adapter or a built-in ANT module. The collected data is immediately forwarded from the Internet-connected mobile device to the server, where it is permanently stored. In this way, coaches and other experts can access the acquired real-time characteristics via developed web applications, providing not only instant analysis but also prompt feedback routines. Thus, specialists can assist athletes and intervene in their training by looking at the performance outcomes, thereby optimizing the achievements and also avoiding fatigue or injuries. At the same time, with the increasing measurement possibilities and hence the growing data amount, also the computer-based analysis of the collected information and the immediate return of feedback become more significant. Consequently, intelligent methods are needed in order to extract significant patterns out of the measured items and send automated real-time notifications to the performing athletes. Therefore, another major aim of the thesis is to propose sophisticated routines for the computerized analysis of the gathered parameter values. Regarding the server-based mobile coaching framework, Artificial Intelligence (AI) techniques running on the host component appear to be an efficient approach for data evaluation purposes. Accordingly, a particular adaptation of the system aims at the use in fitness and the automatic assessment of the executed exercises on weight training machines on the basis of machine learning techniques including classification algorithms like Artificial Neural Networks (ANNs) or fuzzy logic concepts. In running, on the other hand, the focus is set on the integration of an antagonistic meta-model for the analysis of physiological adaptation processes called Performance Potential (PerPot) for the optimization and enhancement of long-distance runs like marathons. The main goal thereby is to detect a possibly occurring fatigue in real-time and, in this way, to avoid exhaustion at an early stage. But also related application scenarios such as monitoring routines for medical purposes gain in importance and are therefore addressed throughout the thesis.