Reliable Heading Tracking for Pedestrian Road Crossing Prediction Using Commodity Devices

TL;DR

A new heading tracking algorithm, the Orientation-Heading Alignment (OHA), which leverages a key insight: people tend to carry smartphones in certain ways due to habits, such as swinging them while walking, and achieves 3.4 times smaller heading errors across nine scenarios than existing methods.

Abstract

Pedestrian heading tracking enables applications in pedestrian navigation, traffic safety, and accessibility. Previous works, using inertial sensor fusion or machine learning, are limited in that they assume the phone is fixed in specific orientations, hindering their generalizability. We propose a new heading tracking algorithm, the Orientation-Heading Alignment (OHA), which leverages a key insight: people tend to carry smartphones in certain ways due to habits, such as swinging them while walking. For each smartphone attitude during this motion, OHA maps the smartphone orientation to the pedestrian heading and learns such mappings efficiently from coarse headings and smartphone orientations. To anchor our algorithm in a practical scenario, we apply OHA to a challenging task: predicting when pedestrians are about to cross the road to improve road user safety. In particular, using 755 hours of walking data collected since 2020 from 60 individuals, we develop a lightweight model that operates in real-time on commodity devices to predict road crossings. Our evaluation shows that OHA achieves 3.4 times smaller heading errors across nine scenarios than existing methods. Furthermore, OHA enables the early and accurate detection of pedestrian crossing behavior, issuing crossing alerts 0.35 seconds, on average, before pedestrians enter the road range.

Figures (13)

• Figure 1: Smartphone's LCS initially aligned with GCS, (a) roll first by $\psi$, (b) pitch second by $\theta$, and (c) yaw third by $\phi$, to achieve (d) last orientation.
• Figure 2: Pedestrian heading $\theta_h$ represents angle between PRF and GCS. A smartphone in front of the pedestrian in LCS. This image was created with the assistance of DALL$\cdot$E 2.
• Figure 3: OHA processing pipeline: smartphone orientation data, along with unreliable and low-frequency headings (e.g. 1Hz), are processed in two phases to compute reliable and high-frequency headings (e.g., 50Hz). A weighted average is employed to merge unreliable headings with drift errors, and high-frequency headings with possible cumulative errors. Learning Phase: The weighted average heading and smartphone orientation data are combined to compute and calibrate the smartphone's relative orientation at current timestamp. Prediction Phase: High-frequency smartphone orientation data and the smartphone's current relative orientation are combined to compute a reliable and high-frequency heading.
• Figure 4: Overall heading estimation error CDF for three common smartphone placements: (1) handheld in front, (2) in a trouser pocket, and (3) swinging by the body side with approximately 60-degree swing angles. We compared OHA with GPS bearing and integrated gyroscope (IG) method, based on IMU and GPS data collected from 5 participants over a total of 135 minutes.
• Figure 5: Heading estimation mean and interquartile range errors across 9 movement scenarios from 5 participants. Scenarios combine three smartphone placements: (1) hand, (2) pocket, and (3) swing, and three walking patterns: (1) Straight walking with Occasional Turns (SOT), (2) Stationary While Rotating (SWR), and (3) walking in Multiple "S"-shaped Paths (MSP).