DiLO: Direct light detection and ranging odometry based on spherical range images for autonomous driving

Abstract

Over the last few years, autonomous vehicles have progressed very rapidly. The odometry technique that estimates displacement from consecutive sensor inputs is an essential technique for autonomous driving. In this article, we propose a fast, robust, and accurate odometry technique. The proposed technique is light detection and ranging (LiDAR)‐based direct odometry, which uses a spherical range image (SRI) that projects a three‐dimensional point cloud onto a two‐dimensional spherical image plane. Direct odometry is developed in a vision‐based method, and a fast execution speed can be expected. However, applying LiDAR data is difficult because of the sparsity. To solve this problem, we propose an SRI generation method and mathematical analysis, two key point sampling methods using SRI to increase precision and robustness, and a fast optimization method. The proposed technique was tested with the KITTI dataset and real environments. Evaluation results yielded a translation error of 0.69%, a rotation error of 0.0031°/m in the KITTI training dataset, and an execution time of 17 ms. The results demonstrated high precision comparable with state‐of‐the‐art and remarkably higher speed than conventional techniques.

EMOS: Enhanced moving object detection and classification via sensor fusion and noise filtering

“…Further improvements in speed were necessary to integrate this approach into AVs. Therefore, we adopted the method implemented in SECOND [16] and adjusted the stride values of the upsampling network blocks from [1,2,4] to [0.5, 1, 2]. After this adjustment, the output feature map had half the width and height of the input, whereas the number of channels increased from 64 to 384.…”

LOCUS 2.0: Robust and Computationally Efficient Lidar Odometry for Real-Time 3D Mapping

“…The work [3] proposes DILO, a lidar odometry technique that projects a three-dimensional point cloud onto a twodimensional spherical image plane and exploits image-based odometry techniques to recover the robot ego-motion in a frame-to-frame fashion without requiring map generation. This results in dramatic speed improvements, however, the method does not fuse additional sensing modalities and is not opensource.…”

PathGAN: Local path planning with attentive generative adversarial networks

“…[35], in which the intentions hidden in the generated words are captured in a sequential latent space through a variational autoencoder (VAE). Our approach is different from that of [35] in the following aspects: (1) We train our model to estimate the intentions hidden in the positions in the paths directly and learn the distribution of the sequential intentions simultaneously through a generative adversarial network (GAN) framework [36]; (2) we design our model to utilize the visual contexts for effective path generation and intention estimation via a novel NN architecture.…”

“…HD maps enable AVs to see beyond the coverage of the mounted sensors by providing an accurate representation of the road and information on the surrounding environment [1,2]. Furthermore, the HD map data ensure the precise localization of AVs using the surrounding lane lines or landmarks as reference positions.…”

“…Figure 1 illustrates the overview of the proposed model. To the best of our knowledge, our model is the first that meets all the following properties: (1) the model generates multiple paths corresponding to nine maneuvers; (2) to generate paths consistent with the nine maneuvers, only one neural network (NN) model is trained; and (3) geometrical information (e.g., HD map and goal position) is not required for path generation. As verified in Section 5, simply modifying existing models [13,14] to take a target maneuver as an input does not work well.…”