Detection Method of Lane Change Intentions in Other Drivers Using Hidden Markov Models

Woo, Hanwool; Ji, Yonghoon; Kono, Hitoshi; Tamura, Yusuke; Yamashita, Atsushi; Asama, Hajime · 2015 · OpenAlex-citations

DOI: 10.1299/jsmeicam.2015.6.253

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the critical safety challenge of detecting lane change intentions in other vehicles, a primary cause of traffic accidents. While previous research has utilized Hidden Markov Models (HMMs) for this purpose, existing methods often rely on steering angle and steering velocity data, which are difficult to measure for external vehicles. Alternatively, methods using only lateral position data have struggled to distinguish slow lane changes from lane-keeping maneuvers due to indistinguishable data distributions. To resolve these limitations, the authors propose a novel HMM-based detection method that utilizes only lateral position data, which is measurable via distance sensors or GPS, thereby enabling the detection of both standard and slow lane changes. The proposed method models a lane change as a sequence of three states: "keeping," "changing," and "adjustment." The system estimates the transition from "keeping" to "changing" using the Viterbi algorithm. A key innovation is the direct use of lateral position values rather than their distributions. This approach allows the system to distinguish lane changes from lane keeping even when the distribution of lateral positions is ambiguous, a common issue in slow lane changes. To handle rapid lane changes that produce large deviations and potentially small emission probabilities, the method sets specific limits on deviations. The features used are lateral positions at specific longitudinal distances ahead of the vehicle, derived from current lateral and longitudinal velocities. The method was evaluated using a real-world dataset published by the Federal Highway Administration. The assessment involved 100 lane change samples and 100 lane keeping samples, none of which were used in the training phase. The results demonstrated high accuracy: 99% for lane change detection (99 out of 100 correct) and 95% for lane keeping recognition (95 out of 100 correct). The authors compared these results with previous techniques. While Kuge et al. achieved 98% accuracy, their method required unmeasurable steering data. Mandalia et al. achieved 97.9% using Support Vector Machines and only 80.2% using HMMs with lateral data. The proposed method outperformed the previous HMM approach and matched the accuracy of methods requiring more complex sensor data, while successfully recognizing slow lane changes that previous lateral-position-only methods failed to detect. The significance of this work lies in providing a robust, high-accuracy solution for detecting lane change intentions in other drivers using only measurable external data. By overcoming the limitation of slow lane change detection, the proposed method enhances the reliability of Driving Safety Support Systems (DSSS) within Intelligent Transport Systems (ITS). This advancement allows for more effective warning systems that can accurately infer the behavior of surrounding vehicles, thereby improving overall driving safety without relying on internal vehicle data that is unavailable for external targets.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• lane changing