Local Minima Prediction using Dynamic Bayesian Filtering for UGV Navigation in Unstructured Environments footnoteinfo
Seung Hun Lee 2, Wonse Jo 1, Lionel P. Robert Jr. 1,3, and Dawn M. Tilbury 1,2,4 · 2025
Read Local Minima Prediction using Dynamic Bayesian Filtering for UGV Navigation in Unstructured Environments footnoteinfo by Seung Hun Lee 2, Wonse Jo 1, Lionel P. Robert Jr. 1,3, and Dawn M. Tilbury 1,2,4 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-14337.txt.
Introduction
Seung Hun Lee, Wonse Jo, Lionel Robert, and Dawn Tilbury wrote this paper in two thousand twenty-five. It teaches ground robots to foresee the dead ends, called local minima, that trap them in rough outdoor ground.
A robot with only a local view can roll into a pocket with no way onward. The method uses dynamic Bayesian filtering on the blocks it sees and the far goal, to gauge the odds of a trap ahead. With that warning, the robot can call for human help or plot a new path before it wedges. The work aims to make field robots less prone to stalls that waste a run.
Catalog record
| CREATOR | Seung Hun Lee 2, Wonse Jo 1, Lionel P. Robert Jr. 1,3, and Dawn M. Tilbury 1,2,4 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-14337
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Local Minima Prediction using Dynamic Bayesian Filtering for UGV Navigation in Unstructured Environments footnoteinfo Seung Hun Lee 2, Wonse Jo 1, Lionel P. Robert Jr. 1,3, and Dawn M. Tilbury 1,2,4 Abstract Path planning is crucial for the navigation of autonomous vehicles, yet these vehicles face challenges in complex and real-world environments. Although a global view may be provided, it is often outdated, necessitating the reliance of Unmanned Ground Vehicles (UGVs) on real-time local information. This reliance on partial information, without considering the global context, can lead to UGVs getting stuck in local minima. This paper develops a method to proactively predict local minima using Dynamic Bayesian filtering, based on the detected obstacles in the local view and the global goal. This approach aims to enhance the autonomous navigation of self-driving vehicles by allowing them to predict potential pitfalls before they get stuck, and either ask for help from a human, or re-plan an alternate trajectory. This research is supported in part by the Automotive Research Center (ARC) in accordance with Cooperative Agreement W56HZV-19-2-0001 U.S. Army DEVCOM Ground Vehicle Systems Center (GVSC) Warren, MI. (OPSEC #: OPSEC8547), and in part by the headquarters of the Republic of Korea Army (ROKA); (Corresponding author: Dawn M. Tilbury). keywords Autonomous Vehicle, Path Planning, Artificial Potential Fields, Local Minima Prediction, Dynamic Bayesian Filtering 1 Introduction As autonomous vehicles increasingly operate in off-road environments, navigating complex terrains presents significant challenges. A key aspect of this navigation is understanding and addressing the issue of local minima. Local minima can mislead optimization processes by appearing optimal while not leading to globally efficient paths. The local minima problem is particularly prevalent in unstructured environments, where traditional algorithms often struggle to adapt to varying obstacles, elevation changes, and dynamic conditions ( Meng and Fang, 2024 ; Gong et al., 2009 ) . Local path-planning techniques must respond quickly to unforeseen obstacles within their sensory range ( Sanchez-Ibanez et al., 2021 ) . One such technique, Artificial Potential Fields (APF), generates obstacle-free paths by assigning positive “charges” to distant goals represented as attractive forces ( \vec{F}_{\text{att}} ), and negative charges to nearby obstacles represented as repulsive forces ( \vec{F}_{\text{rep}} ). APF are an attractive solution for local path planning, as they represent a computationally efficient solution for many navigation scenarios, they are intuitive for remote supervisors of the UGVs to understand, and they work really well most of the time ( Szczepanski et al., 2022 ; Lazarowska, 2019 ; Yao et al., 2020 ) . However, unmanned ground vehicles (UGVs) using APF encounter three significant challenges: becoming trapped in local minima, difficulty navigating between obstacles, and undesirable oscillations ( Szczepanski et al., 2021 ; Orozco-Rosas et al., 2019 ) . Examples of environments where local minimum problems can occur in the real world include irregular terrain, dense obstacle fields, or complex topographical features (e.g., valleys, cliffs, or narrow passages) ( Gong et al., 2009 ) . Without an effective strategy to avoid or bypass local minima, the UGV may become stuck, not knowing where to move, and fail to reach its goal. In order to effectively predict local minima, some studies have used well-defined environments, such as fully mapped or convex settings ( Loizou, 2011 ; Paternain et al., 2017 ; Farivarnejad and Berman, 2020 ) , or pre-processed environments by transforming into simplified and convex shapes ( Sawant et al., 2023 ) , which might result in distortion of the environment. Other research has focused on adapting to non-convex environments with specific shapes, such as triangular or U-shaped, which can be challenging to apply in a general context ( Szczepanski et al., 2021 ; Szczepanski et al., 2022 ) . Nonetheless, these approaches can be difficult to generalize to the more complex and varied conditions found in real-world settings. This paper proposes a novel approach to predict local minima in advance using Dynamic Bayesian filtering. Our aim is to leverage the significant computational and logistical advantages of APF as a local-path planning method, while addressing the local minimum challenge through advance prediction. We make the following two assumptions: A1) the UGV relies on a sensor that provides the distance to local obstacles. This limited perception framework lays the groundwork for our study, aiming to predict the occurrence of local minima in advance within an environment that is only partially known, which is used to mimic real-world off-road environments ( Nowakowski and Kurylo, 2023 ) . A2) the UGV navigates towards the global goal using its local view and a fixed step size within an APF field based on its current position and the goal location. There exists an attractive force ( \vec{F}_{\text{att}} ) towards the goal and repulsive forces ( \vec{F}_{\text{rep}} ) from locally observed obstacles. For prediction, we have defined a state transition model that uses raw sensor data to indicate obstacle positions or free space within the sensor range. We also consider the uncertainty of the information in its current state to refine the prediction. We demonstrate the superiority of our methodology by comparing its prediction results with existing methods in two common situations where local minima arise. The main contributions of this paper are as follows: • We propose a novel method that estimates the probability of the UGV becoming trapped in local minima in unstructured environments using local sensor data; • We demonstrate that the performance of the proposed method is better than two existing methods; The paper is organized as follows. A brief background of previous work done in APF local path planning is discussed in Section 2 . Our proposed method for local minima prediction is presented in Section 3 . The simulation setup and results are detailed in Section 4 . Finally, the discussion and conclusion are provided in Sections 5 and 6 , respectively, where we briefly discuss how future work can use these predictions of local minima to improve UGV navigation in unstructured and poorly-known environments. 2 Backgrounds and Related Works 2.1 Artificial Potential Field (APF) The Artificial Potential Field (APF) algorithm operates by considering the sum of virtual forces generated by external elements, namely the global goal position and local obstacles, to determine the motion of a UGV ( Khatib, 1986 ) . The algorithm computes an \vec{F}_{\text{att}} from the goal position, as well as \vec{F}_{\text{rep}} from nearby obstacles. \begin{array}[]{ll}\vec{F}_{\text{att}}(t)&=\xi(X(t)-X_{g})\\ \vec{F}_{\text{rep}}(t)&=\sum_{i=1}^{S(t)}\eta\frac{1}{||d_{i}(t)||^{2}}\left(\frac{1}{||d_{i}(t)||}-\frac{1}{\rho_{0}}\right)\frac{d_{i}(t)}{||d_{i}(t)||}\end{array} (1) where X(t) is the UGV’s current position at time t , X_{g} is the goal position, \rho_{0} is the sensor range, S(t) is the set of obstacles in sensor range at time t , d_{i}(t) is the vector from UGV to the i th obstacle in S(t) , \xi is the coefficient of \vec{F}_{\text{att}}(t) , and \eta is the coefficient of \vec{F}_{\text{rep}}(t) . These virtual forces enable the UGV to reach the goal position by steering it away from obstacles, thereby preventing collisions. The magnitude and direction of these forces depend on the locations of both the goal position and the obstacles. By combining these forces, the UGV is guided away from obstacles while maintaining a safe distance. However, the APF algorithm faces a local minima challenge that occurs when the forces \vec{F}_{\text{att}}(t) and \vec{F}_{\text{rep}}(t) sum to zero. 2.2 Discrete Time Control The APF method simplifies the