Gradient Profiling for Pedestrian Services Shubham Jain WINLAB, Rutgers University, North Brunswick, NJ ABSTRACT Mobile systems have long been the cause of distraction to pedes- trians. Motivated by the safety challenges this distraction poses, we aim to develop a sensing technology based on smartphones and wearable devices, for fine-grained location classification in urban environments. Particularly, we are using shoe mounted inertial sen- sors to sense the ground a pedestrian is walking on. To begin with, it seeks to detect transitions from sidewalk locations to in-street lo- cations. This can be used for warning a distracted pedestrian or alerting an oncoming vehicle to the presence of a pedestrian in street. In addition, it can also be used for precise sidewalk-level localization in dense urban environments. Categories and Subject Descriptors H.4 [Information Systems Applications]: Miscellaneous Keywords Pedestrian Safety; Smartphone; Localization; Inertial Sensing; GPS; Accelerometer; Gyroscope 1. INTRODUCTION Smartphones have caught our attention for a long time. Unfor- tunately, recently they have turned into a major distraction [1] and are having a negative impact on pedestrian safety. We envision to build smartphone and wearable systems based techniques to en- able various services for pedestrians, primarily safety related. The challenge for these services, however, is to determine the pedes- trian’s precise location. Most importantly, this would require the ability to distinguish when a pedestrian is potentially at risk, par- ticularly walking in street, to walking in relatively safe areas on a sidewalk. To this end, we developed a shoe sensing approach that can determine the gradient of the ground that a user is walking on. This helps us detect when a pedestrian transitions from sidewalk to street, by walking off ramps or stepping off curbs. In the future, we intend to communicate this information to oncoming vehicles. Furthermore, we are also looking at how gradient profiles can be used to precisely localize a pedestrian in dense urban environments where GPS positioning has large errors. In this abstract we present Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage, and that copies bear this notice and the full ci- tation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s). Copyright is held by the author/owner(s). MobiSys’15 Ph.D. Forum, May 18, 2015, Florence, Italy. ACM 978-1-4503-3497-6/15/05. http://dx.doi.org/10.1145/2752746.2752784. our progress so far, and discuss some potential techniques that we intend to build in the future. Related Work. In prior work, the Walksafe project has pre- sented preliminary results on using cameras on pedestrians’ smart- phones for detecting oncoming vehicles [2]. The energy consump- tion of continuous camera operation is likely to remain a challenge. None of the existing works have explored the potential of shoe- mounted sensors in outdoor environments. Robertson et al. [3] ex- plore indoor localization for pedestrians using foot-mounted iner- tial sensors. Jimenez et al. [4] use ramp detection in indoor en- vironments to provide drift correction in indoor locations. Many car producers [5] are now integrating night vision, active break- ing and automatic steering solutions in their new models to reduce pedestrian accidents. Honda is developing a Vehicle-to-Pedestrian technology that is able to detect a pedestrian with a DSRC enabled smartphone [6]. 2. SENSING SOLUTION In our prior work [7, 8], we demonstrated the limitations of smartphone based GPS positioning for pedestrian risk detection, especially in urban environments. In this work, we address the pedestrian safety challenge through shoe-mounted inertial sensors. Our approach exploits the existing trend of shoe mounted exercise tracking devices. Since a pedestrian is usually safe when walking on the sidewalk, our approach is to identify pedestrians that are in the roadway. These pedestrians may be in the way of approaching vehicles and hence potentially at risk. Crossing Detection Algorithm. In our recent work [9], we have proposed a crossing detection algorithm that can generate targeted electronic alerts. Our primary idea is to distinguish street and side- walk locations of the pedestrian through inertial sensing of ground features, particularly by sensing the sidewalk design features that demarcate roadways and sidewalks. In urban environments, that follow consistent design guidelines, these features are primarily ramps and curbs. We leverage these design features and develop a sensing system that can automatically detect transitions from a sidewalk into the road. This includes stepping over a curb, which often occurs when crossing a street at midblock locations. More importantly, our solution can track the inclination of the ground and detect the sloped transitions (ramps) that are installed at many dedicated crossing to improve accessibility. We focus on ramps be- cause they are common in urban environments; we believe that the smoother transition makes it more likely that a distracted pedestrian fails to recognize the transition into the street. We have developed a prototype sensing system based on an iner- tial device affixed to a shoe to evaluate the effectiveness of this ap- proach. This device comprises an accelerometer, gyroscope, mag- netometer, battery and bluetooth. The algorithm extracts changes 11