Data-Dependent Hidden Markov Model with Off-Road State Determination and Real-Time Viterbi Algorithm for Lane Determination in Autonomous Vehicles (2505.04763v2)

Abstract: Lane determination and lane sequence determination are important components for many Connected and Automated Vehicle (CAV) applications. Lane determination has been solved using Hidden Markov Model (HMM) among other methods. The existing HMM literature for lane sequence determination uses empirical definitions with user-modified parameters to calculate HMM probabilities. The probability definitions in the literature can cause breaks in the HMM due to the inability to directly calculate probabilities of off-road positions, requiring post-processing of data. This paper develops a time-varying HMM using the physical properties of the roadway and vehicle, and the stochastic properties of the sensors. This approach yields emission and transition probability models conditioned on the sensor data without parameter tuning. It also accounts for the probability that the vehicle is not in any roadway lane (e.g., on the shoulder or making a U-turn), which eliminates the need for post-processing to deal with breaks in the HMM processing. This approach requires adapting the Viterbi algorithm and the HMM to be conditioned on the sensor data, which are then used to generate the most-likely sequence of lanes the vehicle has traveled. The proposed approach achieves an average accuracy of 95.9%. Compared to the existing literature, this provides an average increase of 2.25% by implementing the proposed transition probability and an average increase of 5.1% by implementing both the proposed transition and emission probabilities.

Data-Dependent Hidden Markov Model for Lane Determination in Autonomous Vehicles

This paper presents a novel approach to lane determination in Connected and Automated Vehicle (CAV) systems, leveraging a data-dependent Hidden Markov Model (HMM) for estimating vehicle lane position with improved accuracy and reduced computational burden. The authors address key gaps in existing literature, which largely relies on empirical definitions and user-modified parameters in HMM-based lane determination methodologies that lack robustness, particularly in scenarios where vehicles deviate from predefined road lanes.

Contributions and Methodology

The principal contribution of this research lies in the development of a time-varying HMM that incorporates both the physical properties of roadways and stochastic characteristics of vehicle sensors. The approach evaluates both emission and transition probabilities directly from sensor data without requiring parameter tuning. It also accounts for scenarios where the vehicle is off the road, eliminating the need for post-processing or HMM breaks—a common issue in previous models. The authors enhance the Viterbi algorithm by adapting it for real-time application within this data-dependent HMM framework, thus ensuring that the most likely sequence of lane transitions is determined as the vehicle navigates through varying road conditions.

Key Findings and Results

The proposed approach achieves an average lane determination accuracy of 95.9%, representing a significant improvement of 5.1% over existing methods that rely on parameterized or empirical HMM definitions. By conditioning both emission and transition probability models on real-time sensor data, the proposed method demonstrates superior performance in both static batch and dynamic real-time settings. This result highlights the benefit of integrating physically meaningful and data-driven probability calculations into the HMM framework.

Practical and Theoretical Implications

The practical implications of this research are profound. The lane determination method developed offers enhanced reliability and accuracy crucial for the deployment of CAV systems, particularly in complex traffic environments where lane-level precision is indispensable. The capability to process standard GNSS data without extensive post-processing or parameter tuning offers significant adaptability and ease of integration into existing vehicle navigation systems.

From a theoretical standpoint, the introduction of data-dependent probability conditioning provides a novel lens for enhancing HMM applications beyond lane determination, potentially extending to other domains reliant on real-time data integration and sequential state estimation.

Future Directions

Future research can expand upon this work by incorporating additional sensor data modalities such as LIDAR or visual imaging, potentially further enhancing the robustness and accuracy of lane determination. Additionally, exploring the integration of this model into comprehensive CAV systems for tasks beyond navigation—such as traffic pattern analysis, platoon trajectory planning, and intersection management—could reveal further applications and benefits.

In conclusion, this paper delivers a substantive improvement in lane determination methods for CAV systems through the innovative use of data-dependent HMMs, providing a scalable and efficient solution that promises to enhance autonomous vehicle navigation and safety.