Current WiFi positioning systems (WPSs) require databases - such as locations of WiFi access points and propagation parameters, or a radio map - to assist with positioning. Typically, procedures for building such databases are time-consuming and labour-intensive. In this paper, two autonomous crowdsourcing systems are proposed to build the databases on handheld devices by using our designed algorithms and an inertial navigation solution from a Trusted Portable Navigator (T-PN). The proposed systems, running on smartphones, build and update the database autonomously and adaptively to account for the dynamic environment. To evaluate the performance of automatically generated databases, two improved WiFi positioning schemes (fingerprinting and trilateration) corresponding to these two database building systems, are also discussed. The main contribution of the paper is the proposal of two crowdsourcing-based WPSs that eliminate the various limitations of current crowdsourcing-based systems which (a) require a floor plan or GPS, (b) are suitable only for specific indoor environments, and (c) implement a simple MEMS-based sensors' solution. In addition, these two WPSs are evaluated and compared through field tests. Results in different test scenarios show that average positioning errors of both proposed systems are all less than 5.75 m.

For vehicle navigation, Global Positioning System (GPS) provides long term accurate positions, but only when direct lines of sight to four or more satellites exist. Inertial Navigation Systems (INS), on the other hand, are self contained sensors that can provide short term accurate navigation information. The integration of the two systems can effectively provide continuous navigation data even during GPS signal outages. Traditional inertial systems were heavy, bulky and costly. In the past two decades, the use and development of light weight, compact and cost effective Micro-Electro-Mechanical Systems (MEMS) based inertial sensors has made the civilian integrated vehicle navigation systems more affordable. However, these sensors still have to make their way in the field due to their significant error sources such as turn-on biases or scale factors variations. Moreover, the performance characteristics of these sensors are highly dependent on the environmental conditions such as temperature variations. Hence there is a need for the development of accurate, reliable and efficient thermal models to reduce the effect of these errors that can degrade the system performance.

Vehicle navigation poses difficulties as it requires the uninterrupted availability of accurate positioning information, even in circumstances without an ideal condition. Global Positioning System (GPS) provides consistently accurate positioning solutions if four or more GPS satellites can be observed. Unfortunately, this condition is usually not satisfied if a vehicle is going through urban canyon, tunnel or forest canopy. Even with High Sensitivity and Assisted GPS receivers, reliable positioning using GPS alone in difficult urban situations is still a challenge. Inertial Navigation System (INS), consists of self- contained sensors that can continuously provide accurate short term positioning solutions. The integration of GPS and INS can overcome the GPS drawback and provide continuous navigation solutions even during GPS signal outages. Though newly developed MEMS-based INS sensors have relatively low accuracy, they are compact and inexpensive, which is very suitable for vehicle navigation. Hence, there is a growing interest in exploring the capabilities of these sensors in the field of vehicle navigation. This paper presents the integration of GPS with MEMS-based INS in a tightly coupled scheme. Tightly coupled integration can make use of GPS signals even if less than four GPS satellites are observed. Thus it offers better integration options. To further improve the GPS/INS integration results, non- holonomic constraints and heading observations were used in this study to improve the online positioning accuracies. The results showed the drift errors could be significantly reduced when non- holonomic constraints and/or heading information were used, during periods with GPS signal outage. In addition, a backward smoother called Rauch-Tung-Striebel (RTS) was also implemented for offline processing needs purpose. The integration results showed that the RTS smoother can significantly reduce the drift errors even if neither non-holonomic constraints nor heading information were used.

The Kalman filter is often used to integrate satellite navigation systems with inertial navigation systems. Such integrated systems are especially useful for navigation of vehicles in urban environments where satellite signals are frequently blocked by tall buildings. The filter weights the measurements of both navigation systems to provide an overall optimal solution. Unfortunately, an optimal solution is only achieved when the filter has been supplied with ideal a priori information such as proper measurement noise characteristics and system dynamics. If such parameters are not perfect they can be detected and compensated for using an intelligent navigation scheme which is adaptable to different sensors. As dynamics are encountered, satellite signal blockages are simulated to test the optimality of the filter. A neural network is then trained to learn any residual deterministic errors which are then removed from future system drifts during signal blockages.

For vehicle navigation, Global Positioning System (GPS) provides long term accurate measurements, but only when a direct line of sight to four or more satellites exists. Inertial navigation systems (INS), on the other hand, are self contained sensors that can provide short term measurements. The integration of the two systems can effectively provide continuous navigation data even during GPS signal outages. Traditional INSs are bulky and expensive, and therefore, can not be used for daily civilian applications. With the evolution of MEMS technology, MEMS-based INS sensors are evolving into more accurate, compact and inexpensive units. Hence, there is a growing interest in exploring the capabilities of these sensors in the field of vehicle navigation. Most of the research is targeted towards finding the best error models and integration techniques that can reduce the high drift and errors associated with these sensors. One of the important aspects of this integration is the optimal configuration for sampling frequency, number of bits and time delay during recording of the various sensor outputs. The very low cost of the MEMS sensors makes the cost of the signal sampling, i.e. analog to digital conversion (ADC), an issue. These parameters will reduce the on-board memory requirement, speed up the computation and hence, significantly reduce the final cost to the consumers.