Road segmentation for autonomous vehicle: A review

In recent years, with the progress of society and the development of the economy, the number of cars in China has been increasing. In wireless communication networks, the choice of wireless nodes has a greater impact on the improvement of system performance (such as channel capacity, coverage area, etc.). The intelligent vehicle target detection system can perceive and recognize the surrounding objects such as pedestrians and vehicles through sensors, which is the basis for realizing the unmanned driving of intelligent vehicles. In a wireless environment where multiple wireless nodes coexist, current research focuses on how long it takes to re-plan the selection of wireless nodes (for mobile environments) and how to allocate and manage wireless nodes, such as where under the conditions of wireless, who will wireless with whom, and in what way (such as auto focus, digital fusion, or other) wireless, etc.; the dedicated centralized controller (such as the base station of the infrastructure-based wireless access network) determines the wireless partner (i.e., centralized type selection), or the wireless node decides by itself (i.e., distributed selection); select the appropriate number of wireless nodes to take into account the system performance gain and implementation complexity. The popularity of automobiles has brought great convenience to people's lives, but it has also brought greater traffic pressure. At present, the amount of domestic automobile traffic has increased exponentially, and urban traffic congestion has become more serious, and even caused serious traffic accidents, directly affecting people's quality of life. The emergence of intelligent transportation systems has effectively alleviated road traffic pressure and reduced the incidence of traffic accidents. As an important part of the intelligent transportation system, the research of intelligent vehicles has received extensive attention. Intelligent vehicles use a variety of sensors installed on the body to sense the environment, and realize intelligent driving functions such as lane line detection, obstacle detection, dynamic cruise control, and unmanned driving, which is conducive to reducing the incidence of traffic accidents and improving the safety of vehicle driving. At present, in the research of target recognition and tracking of intelligent vehicles, the traditional target detection method is mainly based on artificial feature extraction, which is difficult to describe more complex or higher-order image features, and the tracking effect is not good, which limits the target detection. An intelligent vehicle is an intelligent system with functions such as environment perception, planning decision-making, and operation control, and is an important part of the intelligent transportation system. Identify the effect. In response to these problems, we present a study of the target recognition and tracking of intelligent vehicles based on grid map and lidar sensor technology. To verify the proposed method in the actual scene, we identify and track the vehicle ahead on the road through a car equipped with a lidar sensor. The meaning of target detection: target detection, also called target extraction, is an image segmentation based on target geometric and statistical features. The results show that the proposed research method can accurately identify the vehicle in front and detect the moving targets such as pedestrians. The tracking trajectory is also consistent with the expected route. The results show that the laser radar-based vehicle target recognition and tracking in this study is effective. In the 10 frames of images, the average time taken for the recognition of one frame of image is 9.8 ms, while the traditional method is 12.9 ms. Technology can improve reliable road information for intelligent vehicle driving and improve the transportation efficiency and safety performance of intelligent vehicles. There are two main target detection methods for lidar: detection methods based on feature extraction and detection methods based on grid maps.

Segmenting or detecting objects in sparse Lidar point clouds are two important tasks in autonomous driving to allow a vehicle to act safely in its 3D environment. The best performing methods in 3D semantic segmentation or object detection rely on a large amount of annotated data. Yet annotating 3D Lidar data for these tasks is tedious and costly. In this context, we propose a self-supervised pretraining method for 3D perception models that is tailored to autonomous driving data. Specifically, we leverage the availability of synchronized and calibrated image and Lidar sensors in autonomous driving setups for distilling self-supervised pre-trained image representations into 3D models. Hence, our method does not require any point cloud nor image annotations. The keyingredient of our method is the use of superpixels which are used to pool 3D point features and 2D pixel features in visually similar regions. We then train a 3D network on the self-supervised task of matching these pooled point features with the corresponding pooled image pixel features. The advantages of contrasting regions obtained by superpixels are that: (1) grouping together pixels and points of visually coherent regions leads to a more meaningful contrastive task that produces features well adapted to 3D semantic segmentation and 3D object detection; (2) all the different regions have the same weight in the contrastive loss regardless of the number of 3D points sampled in these regions; (3) it mitigates the noise produced by incorrect matching of points and pixels due to occlusions between the different sensors. Extensive experiments on autonomous driving datasets demonstrate the ability of our image-to-Lidar distillation strategy to produce 3D representations that transfer well on semantic segmentation and object detection tasks.

LiDAR sensors are critical for 3D scene perception in autonomous driving, providing high spatial accuracy but suffering from limited temporal resolution. Point cloud interpolation, which reconstructs intermediate frames from sparsely captured inputs, effectively enhances temporal continuity. However, the challenges posed by non-linear and non-rigid motion in dynamic scenarios remain unresolved by existing methods, which often sacrifice accuracy for efficiency. To solve the above problems, we propose MSTSINet (Multiscale Transformer and SSM Interpolation Network), a novel framework that combines multiscale forward flow prediction with non-linear trajectory compensation for fast and accurate interpolation. Unlike conventional bidirectional flow-based methods, our forward flow-guided approach reduces inference time while improving robustness to complex motion. MSTSINet integrates multiscale motion modeling through a novel hierarchical feature extraction and semantic-aware cross-frame attention, progressively improving forward flow estimation for accurate non-linear trajectory prediction. To further improve motion fidelity and temporal coherence, we introduce a point-wise compensation mechanism during inference and design a novel multiscale loss function that supervises forward flow learning across varying temporal resolutions. Experiments on autonomous driving benchmarks demonstrate that MSTSINet significantly outperforms existing state-of-the-art methods, achieving up to a 26.39% reduction in Chamfer Distance and substantially improving inference speed. These advantages make it well-suited for real-time deployment in autonomous driving scenarios.

The article discusses the importance of selfdriving cars to improve road safety and reduce the number of accidents caused by human error. Self-driving cars not only reduce human error but also help reduce driver fatigue. We further explore the use of computer vision in autonomous cars, with previous research relying on deep learning algorithms with LiDAR sensors which can be expensive. The authors propose a more cost-effective approach using simple computer vision algorithms such as color space transformation, Canny edge detection, and Hough line transformation to detect lane lines and steer the car accordingly. This approach requires less operational hardware and can be implemented using affordable boards like Raspberry Pi and Nvidia Jetson Nano. The article also highlights the reconstruction of a remote-controlled car that had a 95% accuracy using a certain set of parameters was a tool for understanding autonomous cars better.

Nowadays, closed-set perception methods for autonomous driving perform well on datasets containing normal scenes. However, they still struggle to handle anomalies in the real world, such as unknown objects that have never been seen while training. The lack of public datasets to evaluate the model performance on anomaly and corner cases has hindered the development of reliable autonomous driving systems. Therefore, we propose a multimodal Synthetic Dataset for Anomaly and Corner case detection, called SDAC, which encompasses anomalies captured from multi-view cameras and the LiDAR sensor, providing a rich set of annotations for multiple mainstream perception tasks. SDAC is the first public dataset for autonomous driving that categorizes anomalies into object, scene, and scenario levels, allowing the evaluation under different anomalous conditions. Experiments show that closed-set models suffer significant performance drops on anomaly subsets in SDAC. Existing anomaly detection methods fail to achieve satisfactory performance, suggesting that anomaly detection remains a challenging problem. We anticipate that our SDAC dataset could foster the development of safe and reliable systems for autonomous driving.

In the recent era, a lot of interest is attracted by the autonomous vehicles which can sense surroundings and navigate without human intervention. Object detection and recognition form a major part of autonomous driving systems. Lidar sensors can be used to capture point clouds of driving environment. Detecting multiple 3D objects in point clouds in real time and defining their boundaries with the help of 3D bounding boxes are critical in motion planning by self driving systems. This paper proposes a LiDAR-based 3D object detection system that operates in real-time, with emphasis on autonomous driving scenarios. A state-of-the-art 2D standard object detector for RGB images, YOLOv4, is used as the base for object detection. The multi-class 3D bounding boxes are generated using a complex regression approach. An Euler-Region-Proposal Network (E-RPN) is used to predict the pose of the object. The proposed model receives point cloud data as input and outputs 3D bounding boxes with classes in real-time. The experiments done on the KITTI benchmark dataset proves that the proposed system outperforms existing methods in terms of accuracy and performance.

As an important part of intelligent vehicle, environmental perception system mainly refers to the detection of the surrounding environment of the vehicle by the sensors attached on the vehicle. In order to ensure the accuracy and stability of the intelligent vehicle environmental perception system, it is necessary to use intelligent vehicle sensors to detect and track objects in the passable area. In this paper, an object detection and tracking algorithm based on the LiDAR and camera information fusion is proposed. The algorithm uses the point cloud data clustering method of LiDAR to detect the objects in the passable area and project them onto the image to determine the tracking objects. After the objects are determined, the algorithm uses color information to track objects in the image sequence. Since the object tracking algorithm based on image is easily affected by light, shadow and background interference, the algorithm uses LiDAR point cloud to modify the tracking results. This paper uses KITTI data set to verify and test this algorithm and experiments show that the target area detection overlap of the proposed target detection and tracking algorithm is 83.10% on average and the tracking success rate is 80.57%. Compared with particle filtering algorithm, the average region overlap increased by 29.47% and the tracking success rate increased by 19.96%.

Autonomous vehicles are expected to arrive sooner than expected. Autonomous vehicles of higher automation rely on both on-board and on-road deployed sensory data. Advanced approaches to improve the situational awareness of autonomous vehicle suggest to implement federated learning, where the raw data need to be transferred from vehicles to edges or clouds and vice-versa. This consequently generates dynamically varying communication link demands. The envisioned new era of autonomous driving demands the strong interplay of key stakeholders like: city authorities and communication network providers. In this paper, we study this relationship, where the traffic efficiency on different road segments may be achieved by incentivizing the autonomous vehicles through better communication resources on alternate routes. We model profit functions of the involved stakeholder. To carryout experiments, we use real traffic data of 8 months, which were collected through sensors deployed at Ernst-Reuter-Platz, Berlin, Germany. We developed an extensive validation framework to validate the approach, which comprises of SUMO, network simulator, and contributed modules. Results show that proposed approach achieves the traffic efficiency and help network operators to use the under-utilized network resources on the alternate paths.

Semantic computing addresses the transformation of data, both structured and unstructured into information that is useful in application domains. One domain where semantic computing would be extremely effective is evacuation route planning, an area of critical importance in disaster emergency management and homeland defense preparation. Evacuation route planning, which identifies paths in a given transportation network to minimize the time needed to move vulnerable populations to safe destinations, is computationally challenging because the number of evacuees often far exceeds the capacity, i.e. the number of people that can move along the road segments in a unit time. A semantic computing framework would help further the design and development of effective tools in this domain, by providing a better understanding of the underlying data and its interactions with various design techniques. Traditional Linear Programming(LP) based methods using time expanded networks can take hours to days of computation for metropolitan sized problems. In this paper, we propose a new approach, namely a capacity constrained routing planner for evacuation route planning which models capacity as a time series and generalizes shortest path algorithms to incorporate capacity constraints. We describe the building blocks and discuss the implementation of the system. Analytical and experimental evaluations that compare the performance of the proposed system with existing route planners show that the capacity constrained route planner produces solutions that are comparable to those produced by LP based algorithms while significantly reducing the computational cost.

The prevalence of GPS applications and other mobile devices has led to the accumulation of a large amount of trajectory data that contains valuable information for intelligent transportation, route planning, city computing etc. However, massive data not only brings new challenges to data storage and retrieval but also leads to serious privacy risks because of the abundant spatiotemporal information. In this paper, we propose a storage scheme that strikes a balance between the compression ratio and precision. We then introduce a road segment generalization method to address privacy issues stemming from sensitive places. Next, we design a two-layer index mechanism to provide an effective retrieval. Furthermore, a privacy preserving storage system PP-TrajStore is implemented. It provides efficient storage based on a road segment compression scheme, preserves privacy by employing sensitive segment generalization technologies, and achieves rapid retrieval by a two-layer index strategy. Finally, a realworld dataset is utilized to demonstrate the performance of PP-TrajStore

Traffic congestion wastes time, energy, and causes pollution. In this paper, We propose an application taking advantage of vehicular communications, namely dynamic route planning. In dynamic route planning, the route from a travel origin to its destination, instead of being statically determined at the travel starting time, is periodically recomputed according to real-time traffic information. Estimating future travel time is indeed central to the dynamic route planning problem. Since knowing future travel times cannot be achieved, a major challenge its estimation. In this paper, we consider three estimates for the future travel time: the latest travel time heuristic commonly used in the literature, an improvement of the latest travel time heuristic, and a novel approach based on exploiting the observed correlation between vehicle density in a road segment and travel time. We show through accurate simulation that all these heuristics are able to considerably improve traffic efficiency, with up to 60% traveling time reduction with respect to the case of the static route planning. Among the considered heuristics, the one based on vehicle density is consistently outperforming the others, especially in presence of traffic build up/decongestion situations (e.g., accidents).

Emergency evacuation on freeways is a process aiming to transfer people from dangerous area to the safe area as quickly as possible. Route planning, therefore, plays an important role during this process. This paper proposes a systematic method to generate optimized routes for rescue vehicles step by step. First, the free flow traveling time and historical traffic volume is calculated from large regional toll collection data. Then the Temporal Convolutional Network (TCN) is adopted to generate real-time ratio between road segments and toll gates. Finally, the Dynamic Bureau of Public Road (DBPR) function and Dijkstra algorithm are used to obtain real-time optimized routes for rescue vehicles. The proposed algorithms are tested on a hypothetical emergency event taking place on the Shantou-Kunming expressway in Xingyi, Anhui Province. The computational results show that the generated rescue routes are helpful for rescue vehicles and can save plenty of time. Generate rescue routes rapidly and accurately may provide a practical method for emergency evacuation without expensive facilities and can be a guide for further rescue operations.

Estimated time of arrival (ETA) prediction, also known as travel time estimation, is a fundamental task for a wide range of intelligent transportation applications, such as navigation, route planning, and ride-hailing services. To accurately predict the travel time of a route, it is essential to take into account both contextual and predictive factors, such as spatial-temporal interaction, driving behavior, and traffic congestion propagation inference. The ETA prediction models previously deployed at Baidu Maps have addressed the factors of spatial-temporal interaction (ConSTGAT) and driving behavior (SSML). In this work, we focus on modeling traffic congestion propagation patterns to improve ETA performance. Traffic congestion propagation pattern modeling is challenging, and it requires accounting for impact regions over time and cumulative effect of delay variations over time caused by traffic events on the road network. In this paper, we present a practical industrial-grade ETA prediction framework named DuETA. Specifically, we construct a congestion-sensitive graph based on the correlations of traffic patterns, and we develop a route-aware graph transformer to directly learn the long-distance correlations of the road segments. This design enables DuETA to capture the interactions between the road segment pairs that are spatially distant but highly correlated with traffic conditions. Extensive experiments are conducted on large-scale, real-world datasets collected from Baidu Maps. Experimental results show that ETA prediction can significantly benefit from the learned traffic congestion propagation patterns. In addition, DuETA has already been deployed in production at Baidu Maps, serving billions of requests every day. This demonstrates that DuETA is an industrial-grade and robust solution for large-scale ETA prediction services.

<title>ABSTRACT</title> <p>Route planning plays an integral role in mission planning for ground vehicle operations in urban areas. Determining the optimum path through an urban area is a well understood problem for traditional ground vehicles; however, in the case of autonomous unmanned ground vehicles (UGVs), additional factors must be considered. For a UGV, perception, rather than mobility, will be the limiting factor in determining operational areas. Current ground vehicle route planning techniques do not take perception concerns into account, and these techniques are not suited for route planning for UGVs. For this study, perception was incorporated into the route planning process by including expected sensor accuracy for GPS, LIDAR, and inertial sensors into the path planning algorithm. The path planner also accounts for additional factors related to UGV performance capabilities that affect autonomous navigation.</p>

The demand for high precision navigation services for human drivers and even intelligent vehicles cannot be satisfied by traditional road-level maps, and on the other hand, current enhanced digital maps with abundant details for vehicle localization and navigation simultaneously bring about practical difficulties for lane-level route planning. In this paper, aiming at achieving lane-level route planning, a three-layer map model combining the traditional road-level data with high precision lane-level data is created, in which an intermediate layer serves as a bridge between road and lane level. Then, based on the lane-level map model, a hierarchical route planning framework is established, including road-level route planning, node contraction according to the intermediate layer, and lane-level route planning with the cost model in lane level which takes the lane change and turn into account. A simulation result is implemented to reveal the validation of this hierarchical routing algorithm based on this novel lane-level map model.