No abstract

No abstract

This study presents an intelligent distribution framework based on edge computing and proposes navigation and obstacle avoidance algorithms for connected logistics vehicles (CLVs) on the basis of Trimble BD982 positioning sensor and tentacle algorithm (TA). An edge computing framework for the distribution of CLVs is established, and the functions of three layers (cloud server, edge equipment, and terminal) are described in detail. The basic functions, hardware, and software systems of the CLV are designed and presented. Focusing on autonomous driving, a Global Positioning System (GPS) navigation algorithm and an obstacle avoidance control strategy on the layer of edge equipment are developed on the basis of the TA. Autonomous GPS navigation is realized by combining the entire road network with the local road network to avoid obstacles. The TA is improved to help the CLV for avoiding obstacles. Experiments show that the hardware system and designed algorithms of the CLV are effective. The tracking error on the straight-line track is within 3 cm, the change rate of longitudinal velocity is within 0.3 g/s, the change rate of tire side deflection angle is less than 1°/s, and the calculation time is shortened by 25% when the calculation time is 30 ms. These results indicate that the vehicle has good stability and performance during obstacle avoidance in real time, and the proposed algorithms are superior to traditional algorithms. The CLV can realize autonomous GPS navigation, with high navigation accuracy, reliable obstacle avoidance performance, and stable vehicle handling.

No abstract

Low-light image enhancement is a preprocessing work for many recognition and tracking tasks for autonomous driving at night. It needs to handle various factors simultaneously including uneven lighting, low contrast, and artifacts. We propose a novel end-to-end Retinex-based illumination attention low-light enhancement network. Specifically, our proposed method adopts multibranch architecture to extract rich features for different depth levels. Meanwhile, we consider the features from different scales in built-in illumination attention module. We encode reflectance features and illumination features into latent space based on Retinex in each submodule, which could cater for highly ill-posed image decomposition tasks. It aims to enhance the desired illumination features under different receptive fields. Subsequently, we propose a memory gate mechanism to learn adaptively long-term and short-term memory. Their weight could control how many high-level and low-level features should be reserved. This method could improve the image quality from both different feature scales and feature levels. Comprehensive experiments on BDD10K and cityscapes datasets demonstrate that our proposed method outperforms various types of methods in terms of visual quality and quantitative metrics. We also show that our proposed method has certain antinoise capability and generalizes well without fine-tuning when dealing with unseen images. Meanwhile, our restoration performance is comparable to that of advanced computationally intensive models.1