<rss version="2.0" xmlns:atom="https://www.w3.org/2005/Atom">
  <channel>
    <title>Transport Research International Documentation (TRID)</title>
    <link>https://trid.trb.org/</link>
    <atom:link href="https://trid.trb.org/Record/RSS?s=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" rel="self" type="application/rss+xml" />
    <description></description>
    <language>en-us</language>
    <copyright>Copyright © 2026. National Academy of Sciences. All rights reserved.</copyright>
    <docs>http://blogs.law.harvard.edu/tech/rss</docs>
    <managingEditor>tris-trb@nas.edu (Bill McLeod)</managingEditor>
    <webMaster>tris-trb@nas.edu (Bill McLeod)</webMaster>
    <image>
      <title>Transport Research International Documentation (TRID)</title>
      <url>https://trid.trb.org/Images/PageHeader-wTitle.jpg</url>
      <link>https://trid.trb.org/</link>
    </image>
    <item>
      <title>Gaussian enhanced deep reinforcement learning for USV navigation with scarce environmental priors</title>
      <link>https://trid.trb.org/View/2633980</link>
      <description><![CDATA[Currently, deep reinforcement learning (DRL) represents the state-of-the-art approach for autonomous navigation of Unmanned Surface Vessels (USVs). However, most studies do not consider the impact of scarce priors on algorithmic generalization. Based on Gaussian Process, an enhanced Deep Q-learning (DQN) algorithm is proposed for autonomous navigation with scarce priors. Firstly, a Gaussian Process Memory (GPM) module based on a small amount of training trajectory data was designed, and the performance of the algorithm was improved by local generalization. Second, to address the low sample efficiency of DQN, Gaussian Expected Improvement Prioritization (GEP) experience replay under the Bayesian framework is proposed. In addition, a simulation environment for the differential reward function was designed. Experimental results demonstrate that the acceleration ability of GEP is universal. GEP not only improves the convergence speed of USV by 56 % but also improves the convergence speed of CartPole-v1 and LunarLander-V2 by 28 % and 33 %, respectively. GPM can significantly improve the generalization of the algorithm. The fusion algorithm EG-DQN successfully combines the advantages of training acceleration and generalization enhancement, and it can demonstrate outstanding generalization performance in complex and unknown scenarios.]]></description>
      <pubDate>Wed, 04 Mar 2026 09:16:08 GMT</pubDate>
      <guid>https://trid.trb.org/View/2633980</guid>
    </item>
    <item>
      <title>Application Analysis of Time Elastic Band Algorithm in Local Path Planning of Driverless Car</title>
      <link>https://trid.trb.org/View/2642266</link>
      <description><![CDATA[The emergence of driverless cars brings convenience to people’s travel. However, their obstacle avoidance ability during driving is closely related to the occurrence of safety accidents. Therefore, this study proposed using a time elastic band for local path planning design to improve the safety and reliability of driverless cars during operation. Firstly, a planning model was constructed using time elastic band, followed by setting and solving constraints on the model. Experimental simulation showed that in the planned path under a single obstacle and multiple obstacles, the vehicle’s driving speed and angular velocity curves had overall persistence and good smoothness. Compared with other algorithms, the driving time in the simulation was reduced by 17.73%. The average speed was increased by 0.46 s. The proposed path planning method fully considered local path sequences, reducing the time cost in the planned parking path by 44.84% compared to other algorithms. The parking position was more in line with the target position. The TEB algorithm was significantly superior in terms of accuracy and success rate of obstacle avoidance compared to the existing methods. Therefore, using time elastic band for local path planning of driverless cars has ideal reliability and stability. Meanwhile, the obstacle avoidance performance of driverless cars based on the algorithm is better.]]></description>
      <pubDate>Tue, 17 Feb 2026 13:12:51 GMT</pubDate>
      <guid>https://trid.trb.org/View/2642266</guid>
    </item>
    <item>
      <title>Deep learning-based recognition of small maritime targets for obstacle avoidance in visual wave gliders</title>
      <link>https://trid.trb.org/View/2660767</link>
      <description><![CDATA[This paper presents a deep learning-based detection system for small maritime targets using visual wave gliders. The system enhances target detection performance in complex maritime environments by integrating the latest YOLOv9s model with a Convolutional Attention Mixing module. Coupled with the StarNet network, it reduces the model's size and computational demands, facilitating real-time operation on low-performance devices. Ablation study results show that compared to the YOLOv9s model, the proposed algorithm increases accuracy by 3.39% and decreases recall by 4.47%. The algorithm has been extensively trained on an augmented dataset composed of synthetic and small-scale ship images, significantly improving the model's capability to recognize low-resolution maritime targets. Sea trials have demonstrated that the model designed in this study outperforms other advanced algorithms in recognizing small maritime vessels, exhibiting superior robustness and stability, and effectively enhancing the visual wave glider's ability to identify small maritime targets.]]></description>
      <pubDate>Wed, 11 Feb 2026 15:10:46 GMT</pubDate>
      <guid>https://trid.trb.org/View/2660767</guid>
    </item>
    <item>
      <title>Path planning of USV in confined waters based on improved A ∗ and DWA fusion algorithm</title>
      <link>https://trid.trb.org/View/2660764</link>
      <description><![CDATA[In the field of unmanned vessels, path planning in confined and complex environments has become a crucial research focus. Existing methods face issues such as insufficient obstacle avoidance and low planning efficiency. To address these challenges, this paper proposes a hybrid approach combining an improved A∗ algorithm with an optimized Dynamic Window Approach (DWA). The enhanced A∗ algorithm adjusts the weight of the heuristic function by introducing a tuning factor (α), which directly influences the obstacle density. Additionally, a 5-neighborhood search combined with the Floyd algorithm is employed to boost search efficiency and improve path smoothness. The modified DWA algorithm incorporates a path smoothing coefficient and a local target selection strategy, enhancing the safety and stability of local planning. MATLAB simulations demonstrate that the proposed hybrid algorithm generates smooth and safe paths, successfully avoids dynamic obstacles, and shows promising effectiveness and feasibility in unmanned vessel path planning.]]></description>
      <pubDate>Wed, 11 Feb 2026 15:10:46 GMT</pubDate>
      <guid>https://trid.trb.org/View/2660764</guid>
    </item>
    <item>
      <title>ARNOLD annotated repository of navigational obstacles from LiDAR data</title>
      <link>https://trid.trb.org/View/2633451</link>
      <description><![CDATA[The strong interest of the maritime community in developing Marine Autonomous Surface Ships, combined with the new technologies nowadays available, requires a vast amount of data for developing and testing perceptual systems to enhance situational awareness. In the marine field, in contrast to other industrial research fields, the lack of real LiDAR data is a limiting factor in developing new systems. In this paper, the authors aim to fill the gap by introducing a novel LiDAR dataset captured in a real marine environment through multiple field campaigns. The dataset is collected in diverse locations, including ports, marinas, and open waters. It encompasses both static and dynamic objects, providing a foundation for training and testing detection, classification, and tracking algorithms. The recorded dataset is available in full open access, and it comprises raw data and a subset of annotated targets classified into four categories: quay, motor boat, sailing boat and ship. The paper presents the data and provides all the necessary instructions for accessing and using it with dedicated parsing codes.]]></description>
      <pubDate>Tue, 27 Jan 2026 16:16:17 GMT</pubDate>
      <guid>https://trid.trb.org/View/2633451</guid>
    </item>
    <item>
      <title>Optimizing the Path of a Mobile Agent in the Environment with Static Obstacles using Reinforcement Learning</title>
      <link>https://trid.trb.org/View/2624146</link>
      <description><![CDATA[Path planning for a mobile agent concerns the problem of searching for a collision-free and optimal path between the initial and target positions. The space in which the agent moves contains a number of obstacles modeled by ordered grids, each representing the obstacle location in the space of the agent movement. The final boundary of an obstacles is formed by its actual boundary plus a minimum safe distance, taking into account the size of the agent, which allows to treating the obstacles as points in the environment. In the article, reinforcement learning algorithms were used to determine the path, including numerous design methods: dynamic programming (policy iteration algorithm and value iteration algorithm), Monte Carlo method (Monte Carlo control algorithm), temporal-difference (TD) learning (Q-learning algorithm and Sarsa algorithm), eligibility traces (Q(ïZ)) algorithm and Sarsa(ïZ) algorithm), planning and learning (Dyna-Q algorithm), and gradient methods (Q-learning algorithm with Adam optimizer and Sarsa algorithm with Adam optimizer). The reinforcement learning algorithms operate on the principle of determining the agent's policy, which seeks the minimum distance between the initial and target positions of the mobile agent, while avoiding obstacles. These learning procedures differ between, dynamic programming, which requires a good knowledge of the environment model to determine the agent's policy, and other methods which do not require this knowledge. The aim of the reported work was to examine the above named algorithms in terms of their effectiveness and speed of finding the optimal solution. Based on the results of the simulation studies, the most effective methods turned out to be that using gradient methods for optimization, i.e.: Q-learning with Adam optimizer.]]></description>
      <pubDate>Tue, 27 Jan 2026 09:21:21 GMT</pubDate>
      <guid>https://trid.trb.org/View/2624146</guid>
    </item>
    <item>
      <title>Safe Reinforcement Learning for Autonomous Driving by Using Disturbance-Observer-Based Control Barrier Functions</title>
      <link>https://trid.trb.org/View/2610724</link>
      <description><![CDATA[Recently, reinforcement learning (RL) has been increasingly used in autonomous driving (AD) navigation control systems. However, most RL-based AD navigation control systems remain in the simulation stage. Its practical application is limited due to growing safety concerns. The safety of these algorithms remains uncertain when confronted with real-world disturbances and vehicle model uncertainties. To enhance the safety of RL, we propose a disturbance observer (DOB) based safe soft actor-critic (SAC) algorithm that combines the SAC algorithm with a safety constraints filter composed of DOB and control barrier function (CBF). When the SAC agent's action output is unsafe, the safety constraints filter will alter it. We employ a DOB to accurately estimate the difference between the nominal model of the vehicle and the actual model, i.e., the lumped disturbances. Then, a more accurate vehicle model can be obtained. To ensure the safety of DOB-SAC under complex and dynamically changing environmental conditions, a further predictive safety constraint is defined based on model predictive control (MPC) ideas. The safe action will be rendered using safety-critical optimal control according to the DOB compensated vehicle model, CBF, and the predictive safety constraints. We discuss the SAC architecture and training details, and investigate the effectiveness of CBF in modeling safety constraints. Joint simulations are conducted in scenarios with static obstacles and intersection scenes with dynamic obstacles.]]></description>
      <pubDate>Fri, 09 Jan 2026 16:59:07 GMT</pubDate>
      <guid>https://trid.trb.org/View/2610724</guid>
    </item>
    <item>
      <title>Integrating Ensemble Methods for Dynamic Obstacle Avoidance in Autonomous Vehicle Navigation</title>
      <link>https://trid.trb.org/View/2621907</link>
      <description><![CDATA[Autonomous vehicles represent a transformative shift in transportation, promising enhanced mobility and safety. However, the deep learning models currently used in this field frequently fail to achieve the real-time, robust performance required for safe and dependable operation. To overcome these limitations we introduce in this study, a novel real-time obstacle avoidance framework that leverages an ensemble deep learning approach combining the strengths of EfficientNet B4 and DenseNet201 architectures. A comprehensive dataset of 30,336 images was curated using data from three strategically placed cameras in the VSim-AV simulator, capturing diverse environmental conditions and various obstacles. Data augmentation and meticulous pre-processing techniques were employed to enrich the dataset, thereby enhancing the model’s adaptability to real-world driving challenges. The proposed ensemble was rigorously evaluated across multiple scenarios in autonomous mode, achieving a testing accuracy of 97.9% and an average reaction time of 0.15 seconds. These results not only validate the effectiveness of the ensemble model but also surpass the performance metrics of existing obstacle detection and avoidance studies in autonomous vehicles.]]></description>
      <pubDate>Fri, 09 Jan 2026 14:44:06 GMT</pubDate>
      <guid>https://trid.trb.org/View/2621907</guid>
    </item>
    <item>
      <title>Fuzzy adaptive zoning algorithm: integrating LiDAR and fuzzy logic for enhanced autonomous electric vehicle navigation in controlled spaces</title>
      <link>https://trid.trb.org/View/2619173</link>
      <description><![CDATA[This research presents the development and comprehensive evaluation of the fuzzy adaptive zoning algorithm (FAZA), an innovative approach to obstacle avoidance systems for AEVs (autonomous electric vehicles) operating in confined environments such as office complexes, industrial zones, and campus areas. Motivated by the need for precise yet cost-effective navigation systems in controlled environments, this study employs a light detection and ranging (LiDAR) VLP-16 sensor through a strategically designed single-sensor architecture optimized for computational efficiency and cost-effectiveness in controlled environments. The system integrates a type-1 Mamdani fuzzy inference system (FIS) with hybrid membership functions combining rectangular and triangular configurations to achieve optimal decision-making balance. The research methodology is based on dividing the detection area into eight distinctive zones (A–H) within a limited operational domain of -2.5m≤x≤2.0m and 0m<y≤6.25m, each with specific response characteristics: “half braking” for objects at safe distances, “full braking”for critical situations, and left or right maneuvers for dynamic avoidance.System evaluation encompassed 96 test scenarios with 100 % accuracy in decision-making, including 16 “half braking” scenarios, 11 right maneuver scenarios, 24 left maneuver scenarios, and 33 “full braking”scenarios. The system demonstrated significant reliability in handling 7 multi-object scenarios and decisiveness in rejecting inputs outside specified parameters. Validation through field testing under low-light conditions confirmed the system’s effectiveness for practical implementation. The results indicate that FAZA successfully provides an efficient and reliable solution for AEV navigation in confined environments, with significant implications for developing more affordable intelligent transportation systems.]]></description>
      <pubDate>Mon, 29 Dec 2025 09:34:36 GMT</pubDate>
      <guid>https://trid.trb.org/View/2619173</guid>
    </item>
    <item>
      <title>Edge-Guided Multi-Scale Fusion and Importance-Aware Learning for Real-Time Semantic Segmentation in Waterborne Navigation</title>
      <link>https://trid.trb.org/View/2598393</link>
      <description><![CDATA[Effective multi-scale feature representation and focused attention on critical objects are essential for accurate perception of waterborne navigation scenes. To address the insufficient exploitation of multi-scale information in existing methods that leads to imprecise segmentation, this study proposes a real-time semantic segmentation method for waterborne navigation scenes through multi-scale information enhancement and importance-weighted optimization. First, DDRNet-23-slim is selected as the backbone network for feature extraction. An edge-guided branch is embedded into its shallow layers, and a Dynamic Feature Fusion Module (DFFM) is constructed by integrating a lightweight hybrid attention mechanism, effectively enhancing multi-scale feature interaction capabilities. Second, the loss function is improved using an importance-weighted strategy to prioritize critical objects during training. Finally, a parameter-free attention mechanism is introduced in the upsampling stage, maintaining real-time performance while ensuring segmentation stability for key objects under complex background interference. Evaluations on the On_Water and Seaships datasets demonstrate that the proposed method achieves mIoU scores of 83.1% and 73.2%, respectively, with ship segmentation accuracy reaching 88.2% on On_Water. The inference speed attains 69.1 FPS, outperforming mainstream real-time segmentation models (e.g., DDRNet, STDC) in balancing accuracy and efficiency. Notably, it exhibits stronger robustness in complex inland river scenarios with dense shore structures and numerous small targets.]]></description>
      <pubDate>Mon, 22 Dec 2025 17:03:50 GMT</pubDate>
      <guid>https://trid.trb.org/View/2598393</guid>
    </item>
    <item>
      <title>Navigation Framework from a Monocular Camera for Autonomous Mobile Robots</title>
      <link>https://trid.trb.org/View/2610737</link>
      <description><![CDATA[Ground detection plays a critical role in the perception systems of autonomous mobile robots. The ground is typically characterized as a smooth, drivable surface with uniform texture, making it distinguishable from its surroundings. However, it may exhibit imperfections such as shadows and varying light conditions. This paper presents a framework for the detection of vanishing points, drivable road regions, intersections, and obstacles in the context of autonomous mobile robot navigation. The proposed framework leverages Google's DeepLab v3+ for semantic segmentation of the road, employs the Hough line transform to identify vanishing points and drivable areas, utilizes an intersection analyzer to locate intersections linked to drivable areas, and incorporates a free obstacle detector to identify various objects within drivable regions. Our objective is to simplify the perception of ground-related information in recent methodologies and offer a solution to comprehend and harness the capabilities of these frameworks. The primary significance of this study lies in evaluating the performance of these networks in real-world deployment scenarios. The evaluation results demonstrate that our proposed framework achieves high accuracy across diverse and challenging situations. Consequently, the developed framework holds promise for integration into autonomous mobile robots (AMRs).]]></description>
      <pubDate>Mon, 15 Dec 2025 10:32:25 GMT</pubDate>
      <guid>https://trid.trb.org/View/2610737</guid>
    </item>
    <item>
      <title>Smooth Filtering Neural Network for Reinforcement Learning</title>
      <link>https://trid.trb.org/View/2598769</link>
      <description><![CDATA[Reinforcement learning (RL) has demonstrated considerable potential in addressing intricate control and decision problems such as vehicle tracking control and obstacle avoidance. Nonetheless, the control policies acquired through RL often lack smoothness even in the presence of minor noises or disturbances, which may induce oscillations, overheating, and even damage to the system in real-world applications. Existing methods handle this issue from temporal or spatial domains and are often tightly coupled with specific tasks. This paper studies the smoothness of control policy from the frequency domain perspective. We propose a class of neural networks with low-pass filtering ability, named Smonet, to alleviate the non-smooth issue by learning a low-frequency representation within hidden layers. Smonet features with serial filtering layers responsible for low-pass filtering of the input signal. Each filtering layer contains multiple inertia cells, one adaptive cell, and one activation layer. To facilitate the filtering ability of Smonet, we further proposed a Smonet-based RL training method by integrating an extra regularization term relating to filtering factors to standard RL loss. Finally, we assess the efficacy of Smonet through diverse simulated robot control tasks and a real-world mobile robot obstacle avoidance experiment, comparing its performance with two commonly utilized networks, multi-layer perceptron and gated recurrent unit. Results indicate that Smonet consistently enhances policy smoothness under various observation noises without compromising control performance. Notably, it achieves up to a 72.7% reduction in the action fluctuation ratio compared to traditional network structures.]]></description>
      <pubDate>Wed, 26 Nov 2025 16:13:29 GMT</pubDate>
      <guid>https://trid.trb.org/View/2598769</guid>
    </item>
    <item>
      <title>Serious and fatal collisions between mobile equipment and pedestrian workers in Quebec: A critical analysis of contributing factors and the Role of proximity detection devices in risk reduction</title>
      <link>https://trid.trb.org/View/2619188</link>
      <description><![CDATA[Mobile equipment (ME), such as loader, that runs over pedestrian workers (PWs) is still a major cause of fatal work accidents. Proximity detection devices (PDDs) are an avenue to mitigate these risks, yet there is a lack of comprehensive studies to inform their practical implementation. Aim. The aim is to provide a structured analysis of the factors contributing to ME-PW collisions and to evaluate the potential impact of integrating PDDs into the causality chain of these accidents. An exhaustive database was created through analysis of investigation reports on fatal ME-PW collisions in Quebec, Canada, from 2013 to 2023 (n = 34). Contributing factors were structured using fault tree analysis (FTA). ME-PW collisions account for 8 % of all available fatal accident reports. Two accident types stand out: dump trucks backing up to deliver earthwork (8/34) and hydraulic shovels used in earthwork operations (8/34). Reversing collisions were the most frequent (21/34), as were accidents occurring during the starting phase of the ME (21/34). In 88 % of cases, the operator was unaware of the PW’s presence just before the accident. This finding confirms the strong potential of PDD as an additional layer of risk reduction. The FTA provides a novel perspective on the potential benefits of PDDs, while also highlighting the challenges of their effective integration into real-world operations. For instance, (1) ensuring coverage of the actual danger zone according to the movement, (2) providing highly targeted warnings, and (3) accounting for subcontracting and environmental conditions that may affect PDD performance.]]></description>
      <pubDate>Mon, 24 Nov 2025 10:19:55 GMT</pubDate>
      <guid>https://trid.trb.org/View/2619188</guid>
    </item>
    <item>
      <title>Optimal Encirclement Control Along Arbitrary Orbits for Vehicles Considering Safety Constraints: A Safe Learning Approach</title>
      <link>https://trid.trb.org/View/2591686</link>
      <description><![CDATA[This paper studies the optimal encirclement problem for autonomous vehicles along arbitrary patterns subject to safety constraints (i.e., avoidance region). A learning-based safe optimal encircling control scheme is proposed to steer vehicles to pursue the target within arbitrary shapes while minimizing the cost and avoiding obstacles. Specifically, an encirclement orbit generator is explored to produce user-specified reference paths centered on targets, rendering the optimal circling problem converted to an optimal tracking issue with additional safety constraints. Furthermore, by mathematically describing the obstacles as control barrier functions (CBFs), a new Hamilton-Jacobi-Bellman (HJB) equation with CBF constraints is constructed, and then a crucial safety declaration is incorporated into the reinforcement learning (RL) strategy to assure safety. Afterward, an improved critic-only approximator is tailored to synthesize the control policy, in which a novel finite-time learning rule formulated by parametric adaptation with guaranteed convergence is developed via setting the auxiliary variables. Compared with the prevailing enclosing alternatives forming an ordinary circular or elliptical path, the proposed approach can not only deliver an arbitrarily user-defined encirclement pattern but also endow an optimum encircling performance without violating safety constraints via online safe learning. Finally, simulations verify the feasibility and superiority of the suggested algorithm.]]></description>
      <pubDate>Fri, 24 Oct 2025 16:53:57 GMT</pubDate>
      <guid>https://trid.trb.org/View/2591686</guid>
    </item>
    <item>
      <title>An Efficient and Rapidly Adaptable Lightweight Multi-Destination Urban Path Planning Approach for UAVs Using Q-Learning</title>
      <link>https://trid.trb.org/View/2591682</link>
      <description><![CDATA[Advancement in UAV technologies have facilitated the development of lightweight airborne platforms capable of fulfilling a diverse range of tasks due to a varied array of mountable sensing and interaction modules available. To further advance UAVs and widen their application spectrum, providing them with fully autonomous operations capability is necessary. To address this challenge, we present Multiple Q-table Path Planning (MQTPP), a novel method specifically tailored for UAV path planning in urban environments. Unlike a conventional Q-learning approach that necessitates relearning in response to dynamic changes in urban landscapes or targets, MQTPP is designed to adaptively re-plan UAV paths with notable efficiency, utilising a singular learning phase executed prior to take-off. Results obtained through simulation demonstrate the exceptional capability of MQTPP to swiftly generate new paths or modify existing ones during flight. This performance significantly surpasses existing state-of-the-art methods in terms of computational efficiency, while still achieving near-optimal path planning results. Thus, demonstrating MQTPP's potential as a robust solution for real-time, adaptive in-flight UAV navigation in complex urban settings.]]></description>
      <pubDate>Fri, 24 Oct 2025 16:53:57 GMT</pubDate>
      <guid>https://trid.trb.org/View/2591682</guid>
    </item>
  </channel>
</rss>