Robust LIDAR localization using multiresolution Gaussian mixture maps for autonomous driving

3D LIDARs (Light Detection and Ranging) with 1.5μm nanosecond pulse lasers have been increasingly used in different applications. The main reason for their popularity is that these LIDARs have high performance while at the same time can be made eye-safe. Because the laser hazard effect on eyes or skin at this wavelength region (<1.4μm) is mainly from the thermal effect accumulated from many individual pulses over a period of seconds, scanning can effectively reduce the laser beam hazard effect from the LIDARs. Neptec LIDARs have been used in docking to the International Space Station, military helicopter landing and industrial mining applications. We have incorporated the laser safety requirements in the LIDAR design and conducted laser safety analysis for different operational scenarios. While 1.5μm is normally said to be the eye-safe wavelength, in reality a high performance 3D LIDAR needs high pulse energy, small beam size and high pulse repetition frequency (PRF) to achieve long range, high resolution and high density images. The resulting radiant exposure of its stationary beam could be many times higher than the limit for a Class 1 laser device. Without carefully choosing laser and scanning parameters, including field-of-view, scan speed and pattern, a scanning LIDAR can’t be eye- or skin-safe based only on its wavelength. This paper discusses the laser safety considerations in the design of eye-safe scanning LIDARs, including laser pulse energy, PRF, beam size and scanning parameters in two basic designs of scanning mechanisms, i.e. galvanometer based scanner and Risley prism based scanner. The laser safety is discussed in terms of device classification, nominal ocular hazard distance (NOHD) and safety glasses optical density (OD).

Simultaneous localization and mapping (SLAM) has been investigated in the field of robotics for two decades, as it is considered to be an effective method for solving the positioning and mapping problem in a single framework. In the SLAM community, the Extended Kalman Filter (EKF) based SLAM and particle filter SLAM are the most mature technologies. After years of development, graph-based SLAM is becoming the most promising technology and a lot of progress has been made recently with respect to accuracy and efficiency. No matter which SLAM method is used, loop closure is a vital part for overcoming the accumulated errors. However, in 2D Light Detection and Ranging (LiDAR) SLAM, on one hand, it is relatively difficult to extract distinctive features in LiDAR scans for loop closure detection, as 2D LiDAR scans encode much less information than images; on the other hand, there is also some special mapping scenery, where no loop closure exists. Thereby, in this paper, instead of loop closure detection, we first propose the method to introduce extra control network constraint (CNC) to the back-end optimization of graph-based SLAM, by aligning the LiDAR scan center with the control vertex of the presurveyed control network to optimize all the poses of scans and submaps. Field tests were carried out in a typical urban Global Navigation Satellite System (GNSS) weak outdoor area. The results prove that the position Root Mean Square (RMS) error of the selected key points is 0.3614 m, evaluated with a reference map produced by Terrestrial Laser Scanner (TLS). Mapping accuracy is significantly improved, compared to the mapping RMS of 1.6462 m without control network constraint. Adding distance constraints of the control network to the back-end optimization is an effective and practical method to solve the drift accumulation of LiDAR front-end scan matching.

This paper presents an offline ground truth generation method using LIDAR(Light Detection and Ranging) scans and odometry. The generated ground truth allows quantitative evaluation of the performance of self-localization methods in urban areas where GNSS(Global Navigation Satellite System) cannot be trusted. The proposed method determines the vehicle pose (position and orientation) by aligning the LIDAR input with previously collected point cloud data. However, as alignment convergences are affected by the environment around the vehicle during each LIDAR scan, it can be erroneous. Incorrect estimates are removed and poses are interpolated by relying on odometry; which is locally accurate. A step by step optimization approach is adopted to yield the most accurate result. Experiments performed in a typical urban environment, with many buildings and surrounding obstacles, demonstrated the effectiveness of the proposed method.

In the field of robotics, four tasks of autonomous robots such as navigation, localization, object tracking, and motion planning, play a crucial role to ensure the autonomy and intelligence of system in the complex and diverse environment. All of its operation relies on the working ability of laser scanner or LiDAR (Light Detection and Ranging) sensor which becomes a promising tool for these tasks due to the accurate distance measurement and wide field of view. The objective of this paper is to evaluate and compare the working performance between 2D and 3D LiDAR sensor in the specific tasks. In the different applications, 2D LiDAR sensor scans on a single plane while the 3D one uses lasers to capture precise 3D data of objects and environments. A series of real-world experiments are conducted in a laboratory setting, focusing on three main tasks: object detection and tracking, mapping capabilities, and motion planning. Owing to these practical tests, the results highlight the strengths and weaknesses of both 2D and 3D LiDAR sensors in such tasks. Experimental results show that, compared with 2D LiDAR, the 3D LiDAR reduces measurement error by up to 59% in the Y-axis, improves human detection accuracy by 28.8% under occlusion conditions, and eliminates navigation collisions caused by planar blind spots, albeit at the cost of a lower frame rate.

Routine bridge inspections require labor intensive and highly subjective visual interpretation to determine bridge deck surface condition. Light Detection and Ranging (LiDAR) a relatively new class of survey instrument has become a popular and increasingly used technology for providing as-built and inventory data in civil applications. While an increasing number of private and governmental agencies possess terrestrial and mobile LiDAR systems, an understanding of the technology’s capabilities and potential applications continues to evolve. LiDAR is a line-of-sight instrument and as such, care must be taken when establishing scan locations and resolution to allow the capture of data at an adequate resolution for defining features that contribute to the analysis of bridge deck surface condition. Information such as the location, area, and volume of spalling on deck surfaces, undersides, and support columns can be derived from properly collected LiDAR point clouds. The LiDAR point clouds contain information that can provide quantitative surface condition information, resulting in more accurate structural health monitoring. LiDAR scans were collected at three study bridges, each of which displayed a varying degree of degradation. A variety of commercially available analysis tools and an independently developed algorithm written in ArcGIS Python (ArcPy) were used to locate and quantify surface defects such as location, volume, and area of spalls. The results were visual and numerically displayed in a user-friendly web-based decision support tool integrating prior bridge condition metrics for comparison. LiDAR data processing procedures along with strengths and limitations of point clouds for defining features useful for assessing bridge deck condition are discussed. Point cloud density and incidence angle are two attributes that must be managed carefully to ensure data collected are of high quality and useful for bridge condition evaluation. When collected properly to ensure effective evaluation of bridge surface condition, LiDAR data can be analyzed to provide a useful data set from which to derive bridge deck condition information.

The demand for a reliable and accurate navigation system that can replace Global Positioning System (GPS) in GPS-denied environment has become increasingly imperative. For indoor environment where GPS is almost unavailable or unreliable, the utilization of other sensors such as inertial sensors becomes necessary. However, inertial sensors alone cannot sustain reliable long-term accuracy due to errors accumulation without external periodic corrections. Thus this paper proposes the utilization of Light Detection and Ranging (LiDAR) as an alternative system to provide periodic corrections. In this paper, a tightly-coupled integrated navigation system that integrates LiDAR, a single-axis gyroscope and wheel encoder is introduced. Straight lines detection and extraction algorithm is utilized to estimate the changes in orientation and range from LiDAR to the extracted line. LiDAR-estimated orientation change and range change to the extracted line feature between two consecutive LiDAR scans are first filtered out through a high rate extended Kalman Filter (EKF) to remove the effect of short-term noise associated with LiDAR scans. Then the smoothed orientation and range changes are fused by a low rate EKF with those predicted by gyroscope and wheel encoder. The proposed system is verified through real experiment on a wirelessly controlled Unmanned Ground Vehicle (UGV). Experimental results indicate that navigation accuracy has been improved to sub-meter and gyroscope bias is precisely estimated.

Light Detection and Ranging (LIDAR) scanning can be used to safely and remotely provide intelligence on the interior of dangerous structures for use by first responders that need to enter these structures. By scanning into structures through windows and other openings or moving the LIDAR scanning into the structure, in both cases carried by a remote controlled robotic crawler, the presence of dangerous items or personnel can be confirmed or denied. Entry and egress pathways can be determined in advance, and potential hiding/ambush locations identified. This paper describes an integrated system of a robotic crawler and LIDAR scanner. Both the scanner and the robot are wirelessly remote controlled from a single laptop computer. This includes navigation of the crawler with real-time video, self-leveling of the LIDAR platform, and the ability to raise the scanner up to heights of 2.5 m. Multiple scans can be taken from different angles to fill in detail and provide more complete coverage. These scans can quickly be registered to each other using user defi ned 'pick points', creating a single point cloud from multiple scans. Software has been developed to deconstruct the point clouds, and identify specific objects in the interior of the structure from the point cloud. Software has been developed to interactively visualize and walk through the modeled structures. Floor plans are automatically generated and a data export facility has been developed. Tests have been conducted on multiple structures, simulating many of the contingencies that a first responder would face. Keywords Hazards, laser scanning, LIDAR, model, segmentation Language: en

The two main categories of light detection and ranging (LIDAR) for mapping and monitoring landslides are static and mobile devices. Static devices are commonly named ground-based LIDAR or terrestrial laser scanners (TLSs). Regarding mobile devices, the different instruments are normally classified depending on the platform configuration: either boats, cars or flying devices. There are three main types of LIDAR, corresponding to three different ways of measuring the range, for instance, the distance along the line of sight (LOS) between the sensor and the terrain. The types include: pulse LIDAR measures the time of flight (TOF) of a laser pulse; phase LIDAR uses the phase shift between the emitted and received signals; and triangulation LIDAR uses a camera to locate the laser spot on the scanned surface. Georeferencing is the process of coordinate transformation from an internal referential system relative to the LIDAR device into a real-world system of reference.

In this study, we describe the realization of a time-of-flight scanning LiDAR (Light Detection and Ranging) prototype, that leverages mirrors as micro-electromechanical systems (MEMS) for agile beam steering in a compact size, and a field-programmable gate array (FPGA)-based processing unit for real-time 3D image reconstruction. The proposed 3D LiDAR system is designed to operate within a range of up to 1 meter with a spatial resolution of 400 × 300 pixels at a frame rate of 30 Hz. The LiDAR prototype architecture consists of 3 main parts: an optomechanical system, a digital processing unit (FPGA-based), an analog front-end. Processed 3D depth maps are rendered in real time via a high-definition multimedia interface (HDMI), providing immediate visual feedback. The full system was deeply characterized and tested. The integration of MEMS mirrors, an FPGA-based time-to-digital converter, and an optimized analog front-end resulted in a highly efficient, compact, and real-time depth sensing platform, ready for a final engineering step. The results obtained represent a feasibility study for a potential commercial product, that is low-cost and small-size, with different consumer applications.

Light Detection and Ranging (LiDAR)-based simultaneous localization and mapping (SLAM) is a core technology for autonomous vehicles and robots. One key contribution of this work to 3D LiDAR SLAM and localization is a fierce defense of view-based maps (pose graphs with time-stamped sensor readings) as the fundamental representation of maps. As will be shown, they allow for the greatest flexibility, enabling the posterior generation of arbitrary metric maps optimized for particular tasks, for example, obstacle avoidance and real-time localization. Moreover, this work introduces a new framework in which mapping pipelines can be defined without coding, defining the connections of a network of reusable blocks much like deep-learning networks are designed by connecting layers of standardized elements. We also introduce tightly-coupled estimation of linear and angular velocity vectors within the Iterative Closest Point (ICP)-like optimizer, leading to superior robustness against aggressive motion profiles without the need for an IMU. Extensive experimental validation reveals that the proposal compares well to, or improves, former state-of-the-art (SOTA) LiDAR odometry systems, while also successfully mapping some hard sequences where others diverge. A proposed self-adaptive configuration has been used, without parameter changes, for all 3D LiDAR datasets with sensors between 16 and 128 rings, and has been extensively tested on 83 sequences over more than 250 km of automotive, hand-held, airborne, and quadruped LiDAR datasets, both indoors and outdoors. The system flexibility is demonstrated with additional configurations for 2D LiDARs and for building 3D NDT-like maps. The framework is open-sourced online: https://github.com/MOLAorg/mola.

&lt;span lang="EN-US"&gt;In recent years, there has been a strong demand for self-driving cars. For safe navigation, self-driving cars need both precise localization and robust mapping. While global navigation satellite system (GNSS) can be used to locate vehicles, it has some limitations, such as satellite signal absence (tunnels and caves), which restrict its use in urban scenarios. Simultaneous localization and mapping (SLAM) are an excellent solution for identifying a vehicle’s position while at the same time constructing a representation of the environment. SLAM-based visual and light detection and ranging (LIDAR) refer to using cameras and LIDAR as source of external information. This paper presents an implementation of SLAM algorithm for building a map of environment and obtaining car’s trajectory using LIDAR scans. A detailed overview of current visual and LIDAR SLAM approaches has also been provided and discussed. Simulation results referred to LIDAR scans indicate that SLAM is convenient and helpful in localization and mapping.&lt;/span&gt;

The work conducted in this thesis is a part of an EU Horizon 2020 research initiative project known as DigiArt. This part of the DigiArt project presents and explores the design, formulation and implementation of probabilistically orientated state estimation algorithms with focus towards unmanned system positioning and three-dimensional (3D) mapping. State estimation algorithms are considered an influential aspect of any dynamic system with autonomous capabilities. Possessing the ability to predictively estimate future conditions enables effective decision making and anticipating any possible changes in the environment. Initial experimental procedures utilised a wireless ultra-wide band (UWB) based communication network. This system functioned through statically situated beacon nodes used to localise a dynamically operating node. The simultaneous deployment of this UWB network, an unmanned system and a Robotic Total Station (RTS) with active and remote tracking features enabled the characterisation of the range measurement errors associated with the UWB network. These range error metrics were then integrated into an Range based Extended Kalman Filter (R-EKF) state estimation algorithm with active outlier identification to outperform the native approach used by the UWB system for two-dimensional (2D) pose estimation.The study was then expanded to focus on state estimation in 3D, where a Six Degreeof-Freedom EKF (6DOF-EKF) was designed using Light Detection and Ranging (LiDAR) as its primary observation source. A two step method was proposed which extracted information between consecutive LiDAR scans. Firstly, motion estimation concerning Cartesian states x, y and the unmanned system’s heading (ψ) was achieved through a 2D feature matching process. Secondly, the extraction and alignment of ground planes from the LiDAR scan enabled motion extraction for Cartesian position z and attitude angles roll (θ) and pitch (φ). Results showed that the ground plane alignment failed when two scans were at 10.5◦ offset. Therefore, to overcome this limitation an Error State Kalman Filter (ES-KF) was formulated and deployed as a sub-system within the 6DOF-EKF. This enabled the successful tracking of roll, pitch and the calculation of z. The 6DOF-EKF was seen to outperform the R-EKF and the native UWB approach, as it was much more stable, produced less noise in its position estimations and provided 3D pose estimation.

Recently, light detection and ranging (LiDAR) technology has emerged as a promising approach to enhance sensing systems, particularly in human activity recognition (HAR). Advances in computer vision and LiDAR sensor development are revolutionizing HAR systems. This work proposes a novel method to detect and classify human exercise poses using 3D LiDAR technology by taking advantage of deep learning techniques. We utilize the Point Pillars network for 3D object detection and employ pseudo-image transformation for effective 3D to 2D data conversion, enhancing the applicability of 2D CNNs for 3D point cloud data analysis. To account for temporal dependencies in 3D point cloud data streams, we use the "smush" technique, which considers a few consecutive 3D LiDAR point cloud frames. This method significantly improves the detection accuracy of dynamic exercises, highlighting the potential of our approach in enhancing fitness monitoring technologies. This research not only demonstrates the effective use of 3D LiDAR in capturing complex activities but also lays the groundwork for future advancements in mapping human movement using 3D LiDAR. <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup>

This paper presents a Light Detection and Ranging (LiDAR) data set that targets complex urban environments. Urban environments with high-rise buildings and congested traffic pose a significant challenge for many robotics applications. The presented data set is unique in the sense it is able to capture the genuine features of an urban environment (e.g. metropolitan areas, large building complexes and underground parking lots). Data of two-dimensional (2D) and threedimensional (3D) LiDAR, which are typical types of LiDAR sensors, are provided in the data set. The two 16-ray 3D LiDARs are tilted on both sides for maximal coverage. One 2D LiDAR faces backward while the other faces forwards to collect data of roads and buildings, respectively. Raw sensor data from Fiber Optic Gyro (FOG), Inertial Measurement Unit (IMU), and the Global Positioning System (GPS) are presented in a file format for vehicle pose estimation. The pose information of the vehicle estimated at 100 Hz is also presented after applying the graph simultaneous localization and mapping (SLAM) algorithm. For the convenience of development, the file player and data viewer in Robot Operating System (ROS) environment were also released via the web page. The full data sets are available at: http://irap.kaist.ac.kr/dataset. In this website, 3D preview of each data set is provided using WebGL.

Three-dimensional light detection and ranging (LiDAR) sensors have received much attention in the field of autonomous navigation owing to their accurate, robust, and rich geometric information. Autonomous vehicles are typically equipped with multiple 3D LiDARs because there are many commercially available low-cost 3D LiDARs. Extrinsic calibration of multiple LiDAR sensors is essential in order to obtain consistent geometric information. This paper presents a systematic procedure for the extrinsic calibration of multiple 3D LiDAR sensors using plane objects. At least three independent planes are required within the common field of view of the LiDAR sensors. The planes satisfying the condition can easily be found on objects such as the ground, walls, or columns in indoor and outdoor environments. Therefore, the proposed method does not require environmental modifications such as using artificial calibration objects. Multiple LiDARs typically have different viewpoints to reduce blind spots. This situation increases the difficulty of the extrinsic calibration using conventional registration algorithms. We suggest a plane registration method for cases in which correspondences are not known. The entire calibration process can easily be automated using the proposed registration technique. The presented experimental results clearly show that the proposed method generates more accurate extrinsic parameters than conventional point cloud registration methods.