Deep Learning-Based Stopped Vehicle Detection Method Utilizing In-Vehicle Dashcams

In this paper, a hierarchical framework for task assignment and path planning of multiple unmanned aerial vehicles (UAVs) in a dynamic environment is presented. For multi-agent scenarios in dynamic environments, a candidate algorithm should be able to replan for a new path to perform the updated tasks without any collision with obstacles or other agents during the mission. In this paper, we propose an intersection-based algorithm for path generation and a negotiation-based algorithm for task assignment since these algorithms are able to generate admissible paths at a smaller computing cost. The path planning algorithm is also augmented with a potential field-based trajectory replanner, which solves for a detouring trajectory around other agents or pop-up obstacles. For validation, test scenarios for multiple UAVs to perform cooperative missions in dynamic environments are considered. The proposed algorithms are implemented on a fixed-wing UAVs testbed in outdoor environment and showed satisfactory performance to accomplish the mission in the presence of static and pop-up obstacles and other agents.

This paper introduces an accurate nonlinear model predictive control-based algorithm for trajectory following. For accuracy, the algorithm incorporates both the planned state and control trajectories into its cost functional. Current following algorithms do not incorporate control trajectories into their cost functionals. Comparisons are made against two trajectory following algorithms, where the trajectory planning problem is to safely follow a person using an automated ATV with control delays in a dynamic environment while simultaneously optimizing speed and steering, minimizing control effort, and minimizing the time-to-goal. Results indicate that the proposed algorithm reduces collisions, tracking error, orientation error, and time-to-goal. Therefore, tracking the control trajectories with the trajectory following algorithm helps the vehicle follow the planned state trajectories more accurately, which ultimately improves safety, especially in dynamic environments.

The evaluation of collision avoidance strategies for special-purpose unmanned ground vehicles (UGVs) operating in dynamic environments with human presence is presented in this study. The autonomous wheeled vehicle equipped with an armament module for combat and reconnaissance tasks is described considering mission specifications and off-road usage scenarios. The main operational parameters are outlined, and the obstacle detection features integrated into the PERUN vehicle are introduced. The control system architecture is presented, along with installed sensors and their perception capabilities of surrounding object detection. Safety considerations in operational conditions where human presence force appropriate collision avoidance actions are discussed. Real field tests conducted in diverse weather and environmental conditions, such as industrial area and unstructured terrains, are shown to demonstrate the system's performance. Carried out tests show that the collision avoidance system enables obstacle detection and proving the effectiveness of steering strategies for the platform's navigation. The safe operation of unmanned systems in environments with human presence is crucial for the successful deployment and scalability of UGV technologies.

To generate safe and efficient trajectories for an automated vehicle in dynamic environments, a layered approach is usually considered, which separates path planning and speed profile planning. This paper is focused on speed profile planning for a given path that is represented by a set of waypoints. The speed profile will be generated using temporal optimization which optimizes the time stamps for all waypoints along the given path. The formulation of the problem under urban driving scenarios is discussed. To speed up the computation, the non-convex temporal optimization is approximated by a set of quadratic programs which are solved iteratively using the slack convex feasible set (SCFS) algorithm. The simulations in various urban driving scenarios validate the effectiveness of the method.

Connected and automated vehicles (CAVs) have shown the potential to improve safety, increase throughput, and optimize energy efficiency and emissions in complicated traffic scenarios. This paper presents a mixed-integer linear programming (MILP) method for scheduling and coordination of CAVs in a highly dynamic environment that consists of multiple human-driven vehicles and multiple conflict zones, such as merging points and intersections. The proposed approach ensures safety, high throughput and energy efficiency by solving a centralized high-level decision making problem. The solution provides a feasible and optimal time schedule through road segments and conflict zones for the automated vehicles, by using information from the position, velocity, and destination of the manual vehicles, which cannot be directly controlled. The performance and computational load of the proposed method are assessed in closed-loop simulations on an illustrative scenario. Despite MILP having combinatorial complexity, the proposed formulation appears feasible for real-time implementation, e.g., in mobile edge computers (MECs).

An adaptive Monte Carlo tree search (MCTS) framework is presented for long-range, energy-aware trajectory planning and online re-planning of fixed-wing uncrewed aerial vehicles in dynamic environments. Mission planning is posed as a finite-horizon Markov decision process with a four-degree-of-freedom kinematic and energy model, forecast wind fields, terrain, and time-varying no-fly zones. The MCTS planner searches over waypoint references generated by pre-stabilizing feedback controllers and a constraint-aware navigation field, yielding dynamically feasible trajectories that approximately satisfy altitude, obstacle, and airspace constraints. Simulations for a coastal medical delivery mission show that, relative to a Batch Informed Trees baseline, the method trades modest path-length increases for reduced energy consumption and, under emerging storm-front no-fly regions, rapidly computes safe diversion trajectories that preserve most nominal energy performance under fixed computational budgets.

The original article to which this Erratum refers was published in IEEJ Transactions on Electrical and Electronic Engineering Vol.4, No. 4, 2009, pp.561‐569.

Efficient trajectory planning for intelligent vehicles in dynamic environments is a non-trivial task due to the diversity and complexity of driving scenarios. It requires the planner to be capable of responding to the changes in driving scenarios in real-time. This paper proposes a hybrid trajectory planning framework by combining the sampling- and numerical optimization-based approaches to cope with the complex driving scenarios. First, a risk field model is introduced to assess the risks with the static and moving obstacles. Then, the sampling-based approach is used to generate collision-free trajectory candidates via the Path Velocity Decomposition method. Thus, the optimal behavior trajectory can be obtained by considering curve smoothness, collision risk, and travel time. The optimization-based method is adopted to optimize the behavior trajectory to guarantee safety, vehicle dynamics stability, and driving comfort using the Sequential Quadratic Programming within the spatio-temporal boundaries. Finally, the proposed framework is examined in typical dynamic driving scenarios through simulation, and the results verify its competency in generating high-quality trajectories in real-time.

Formation flight holds significant potential for various applications involving aerial robot swarms. However, current methodologies lack the capability to autonomously execute large-scale formation flights in densely populated environments. To bridge the gap, a decentralized and asynchronous formation flight planner is proposed based on a graph-based formation metric and tight representations of kinodynamically feasible trajectories for collision avoidance. The planner handles formation path planning along with dynamic obstacles and intervehicle collision avoidance using minimum volume convex hulls for agent trajectories. The employed formation metric is invariant to rotation, translation, and scaling, granting greater flexibility in formation coordination. A decoupled and distributed trajectory optimization framework is proposed to enhance the computational feasibility of large-scale formation flights. Moreover, to mitigate issues relating to communication delays between agents, asynchronous execution of a finite horizon navigation framework with usage of sparse trajectory control points for trajectory segments is employed. Simulations with multiple agents, static and dynamic obstacles support the robustness of the planner to formation flights in real world. The planner demonstrates goal/waypoints achievement and formation adherence capabilities that are assessed and compared using a popular quantifiable formation similarity metric. Furthermore, the paper also serves as a guideline to build upon trajectory planning frameworks for tight formation control in cluttered, dynamic environments.

Current vehicle trajectory prediction methods often fail to capture the complex spatial and temporal interactions among surrounding vehicles in dynamic environments, leading to suboptimal prediction accuracy. In response to these challenges, we propose a novel Transformer-based model for vehicle trajectory prediction, denoted as the Dynamic Spatial-Temporal Interaction-aware Transformer Network (DSTIFormer). The three main parts of the DSTIFormer architecture are a spatial-temporal embedding module, a spatial-temporal dynamics interaction-aware module, and a trajectory encoderdecoder module. The spatial-temporal embedding module aims to encode vehicular trajectory interactions into latent representations that simultaneously preserve spatial dependencies and temporal features. The second module utilizes a temporal transformer network (TTN) along with a spatial graph transformer network (STN) to capture the dynamic spatial-temporal interactions between vehicles. To make trajectory predictions more accurate, the third module uses a sequence-to-sequence (Seq2Seq) architecture as the encoder and an attention-based decoder layer to effectively extract the long-range sequential dependencies in the process. The model's prediction accuracy with the root mean squared error (RMSE) metric on two publicly available datasets: NGSIM and HighD, and the results demonstrate that our proposed model outperforms several benchmark methods in long-term prediction.

Although the number of automobiles on the road currently is more than it was a few years ago, the infrastructure of roads and traffic systems has not kept up with this expansion, which makes handling traffic challenging. This mismatch causes hence a lot of traffic congestion, crowding, and pollution. All throughout the globe, handling increasing traffic is a major challenge. Operations depend on traffic management in great part. It has to do with planning, running, and purchasing the necessary transportation services—cars, trucks, railroads, boats—that carry commodities and vehicles. Intelligent Traffic Systems (ITS) might be a fantastic approach to address these types of issues by means of modern technology. Given this, traffic control philosophy has to change to accommodate smart cities—which use computer vision techniques to run an adaptive traffic management system depending on CCTV image stream analytics. The adaptive traffic management system detects and counts vehicles accurately. In smart cities, where traffic lights must be automated based on vehicle density, this technique is particularly helpful. This study investigated intelligent traffic control, gathering and monitoring traffic data through using CCTV images. It embodies possible risks in the implementation state, benefits, challenges, as well as the future prospect of ITS.

With the development of urban traffic and the massive increase in data collection, there is a problem in the transit time prediction process of the intelligent transportation system, that is, the real-time traffic information contained in the new sample data causes the original prediction model failed. Therefore, the article implements adaptive update of traffic collection data based on time decay factor, thus ensuring effective prediction of short-term transit time. First, data collection is carried out on urban road conditions, and the missing data of traffic flow is complemented by a combination of time and space. Subsequently, the road condition is divided according to the traffic state index, and the semi-supervised k-Nearest Neighbor Means (k-NNM) clustering algorithm is used to obtain the typical similar road segments, for the to-be-predicted samples and the historical collection samples. Finally, based on the time decay factor, the weighted value of the historical typical similar road segment sample data is weighted, so that the validity of the data value is measured by the time distance, and the transit time prediction model is satisfied to meet the requirements of the road condition update. Experiments show the effectiveness and applicability of the method, so as to provide real-time updated road condition data for the prediction of road transit time.

Information regarding the conditions of roads is a safety concern when driving. In Bangkok, public weather sensors such as weather stations and rain sensors are insufficiently available to provide such information. On the other hand, a number of existing CCTV cameras have been deployed recently in various places for surveillance and traffic monitoring. Instead of deploying new sensors designed specifically for monitoring road conditions, images and location information from existing cameras can be used to obtain precise environmental information. Therefore, we propose a road environment extraction framework that covers different situations, such as raining and non-raining scenes, daylight and night-time scenes, crowded and non-crowded traffic, and wet and dry roads. The framework is based on CCTV images from a Bangkok metropolitan dataset, provided by the Bangkok Metropolitan Administration. To obtain information from CCTV image sequences, multi-label classification was considered by applying a convolutional neural network. We also compared various models, including transfer learning techniques, and developed new models in order to obtain optimum results in terms of performance and efficiency. By adding dropout and batch normalization techniques, our model could acceptably perform classification with only a few convolutional layers. Our evaluation showed a Hamming loss and exact match ratio of 0.039 and 0.84, respectively. Finally, a road environment monitoring system was implemented to test the proposed framework.

This research paper focuses on the development and analysis of a web-based system which is designed to study road traffic conditions and their routing behaviour in a dynamic environment. Nowadays Traffic congestion has become a major issue in urban areas due to the increasing number of vehicles and frequent changes in road conditions such as congestion, accidents, and peak-hour delays. Efficient route planning during those conditions requires routing approaches that can calculate traffic variations in real time. In this thesis, the road network is modelled as a weighted graph, where intersections are represented as nodes and road segments are represented as edges with weights corresponding to traffic conditions. The proposed web application allows users to interact with the road network by selecting source and destination points. They can also modify traffic conditions, and routing techniques and observe how routes change dynamically. Both static shortest path computation and dynamic routing approaches are implemented to study their behaviour and performance under varying traffic scenarios. The study provides visual representation of road networks and routing paths and also their performance metrics such as computation time and route cost. Thus it helps understand the clear comparison between static and dynamic routing methods and the impact of traffic changes depending on routing decisions. The developed web application serves as an analytical and educational tool for studying road traffic analysis and routing algorithms. The study shows the usefulness of web-based platforms in visualizing traffic behaviour and evaluating routing performance. But the current approach is limited to simulated data. This leaves room for future work that could improve the system by integrating real-world traffic data and explore more advanced routing strategies.

Aim: The study investigates how deep learning, particularly YOLOv4, might be included into the CCTV systems of the Federal Polytechnic Ile-Oluji Library for proactive monitoring, real-time anomaly detection, and resource optimization. Selected for real-time surveillance in dynamic environments, YOLOv4 balances speed and accuracy against alternatives like R-CNN. Sample: Examining CCTV footage from the Federal Polytechnic Ile-Oluji Library assisted one to identify prospective security issues, heavy usage hours, and activity trends. Study Design: A study leveraging YOLOv4's capacity for activity tracking, anomaly identification, and efficient use of library grounds resources. Conducted in the Federal Polytechnic Ile-Oluji Library, eight months of research spanning January 2024 through October 2024. Methodology: YOLOv4's deep learning architecture was integrated with library CCTV footage to enable real-time video analysis. The evaluation of its performance was conducted using the F1 score, recall, and precision metrics. As privacy and environmental adaptation emerged as focal points, heatmaps revealed areas of significant activity. R-CNN and similar methodologies were not utilized due to their slower processing speeds and higher computational demands, which limit their applicability in real-time settings. Results: The system attained accuracy, recall, and F1 ratings above 90%. The Peak library activity was noted between 10 AM and 2 PM, which helped to guide resources. YOLOv4 improved operating efficiency and library security by detecting questionable activity rather well. Conclusion: YOLOv4 has been demonstrated to be a powerful instrument for real-time surveillance, surpassing the efficiency and practicality of conventional methods and other deep learning models such as R-CNN in this application. This work shows how deep learning could turn CCTV systems into intelligent monitoring systems, therefore opening the path for safer academic surroundings.