Pedestrian Dead Reckoning (PDR) is a frequently used and established method for indoor positioning of humans with smartphones. The device's motion can be derived from the accelerometer, gyroscope, and magnetometer sensor data in this method. In the acceleration data of the inertial measurement unit (IMU), strides or steps are usually determined using various methods to divide the translational movements of a user into straight-line movement sections. Because of the challenge of separating the acceleration signals into the vertical and the horizontal directions during the movement and excluding the gravitational acceleration, numerous methods approximate the stride length based on key features of the movement or based on user-specific parameters. The novel translation determination method presented in this paper does not require a step detection. Furthermore, it provides a higher accuracy than established stride length determination methods from the literature, and it estimates the translation without the need for user characteristics. The presented method, called "gia", achieves the highest accuracies for the relative stride length error, the root mean square error, and the 95% error for 20 consecutive strides compared to 15 other methods for four benchmark datasets.
Regulating free-floating micromobility services, such as e-scooters, presents a unique challenge for cities. This work aims to support cities in developing regulations by assessing similarities between and within urban regions based on micromobility trip patterns, built environment characteristics, and socioeconomic and demographic profiles. We develop three random forest classification models across these dimensions for five U.S. cities - Baltimore, Denver, Detroit, Portland, and Washington DC. We demonstrate that cities with similar built environment characteristics or socioeconomic and demographic profiles do not always exhibit comparable micromobility characteristics. Similarly, we find that cities exhibit more generalizable micromobility patterns than they do through built environment or socioeconomic and demographics. Our regional similarity models offer a means to compare micromobility across regions, while contrasting how regions differ in their built environment or socioeconomic and demographic characteristics. By enabling micromobility focused comparisons, cities with similar built environment characteristics or socioeconomic and demographic profiles may draw conclusions as to which policies may be most effective.
Efficient budget allocation is essential for sustaining mobile crowdsensing systems (MCS), where resource distribution must balance cost-effectiveness and participant motivation to ensure optimal task execution. This paper introduces two novel winner selection mechanisms designed to maximize system rewards while adhering to financial constraints: a) Targeted Regional Incentive (TRI), and b) TRI with Critical Value (TRI-CV). The proposed incentive mechanisms are structured through regional divisions within the area of interest for effective task allocation. TRI employs a computationally efficient, low-overlap prioritization strategy that supports real-time task allocation across spatial regions. TRI-CV builds on this foundation by introducing region-specific critical value thresholds and a lightweight audit reassignment step, which reallocates budget from overrepresented to underserved regions to enhance spatial coverage and fairness. Simulation results show that TRI improves platform utility by 45-50% and reduces auction costs by 35-36%, while TRI-CV achieves 50-58% higher utility and 40-42% lower costs compared to established mechanisms such as FIRE, GRUS, and BFM. Additionally, TRI-CV demonstrates superior participant retention and scalability, positioning it as a robust and fairness-aware solution for dynamic Mobile Crowd Sensing environments.
Despite advances in route optimization and navigational technologies, pedestrians still face challenges navigating complex urban spaces. This study aims to identify and quantify the multidimensional factors that influence pedestrian navigation, with the focus on why people get lost. The collected data from an online survey in which 64 participants reported the locations and contextual information of their getting lost events. Building upon literature, expert interviews, and collected data, we identify and quantify 14 environmental, situational, and personal factors influencing pedestrian navigation. We utilize a dual-analytical approach that combines expert-led analytic hierarchy process (AHP) analysis with data-driven regression models to derive distinct weighting schemes for the factors. While the approach based on experts' opinion (i.e. AHP) demonstrates that familiarity, self-orientation skills, and access to reliable navigation tools are the most important contributing factors, data-driven models additionally highlight the significance of environmental complexity, such as angular distance between exits and number of landmarks near decision points. Importantly, expert-led and data-driven methods reveal different but complementary influences on pedestrian navigation, underscoring the value of combining specialist knowledge with insights derived from observed data to improve our understanding and guide the creation of more navigable urban environments.
