Deep Learning Sensor Fusion for Autonomous Vehicle Perception and Localization: A Review

Abstract

Autonomous vehicles (AV) are expected to improve, reshape, and revolutionize the future of ground transportation. It is anticipated that ordinary vehicles will one day be replaced with smart vehicles that are able to make decisions and perform driving tasks on their own. In order to achieve this objective, self-driving vehicles are equipped with sensors that are used to sense and perceive both their surroundings and the faraway environment, using further advances in communication technologies, such as 5G. In the meantime, local perception, as with human beings, will continue to be an effective means for controlling the vehicle at short range. In the other hand, extended perception allows for anticipation of distant events and produces smarter behavior to guide the vehicle to its destination while respecting a set of criteria (safety, energy management, traffic optimization, comfort). In spite of the remarkable advancements of sensor technologies in terms of their effectiveness and applicability for AV systems in recent years, sensors can still fail because of noise, ambient conditions, or manufacturing defects, among other factors; hence, it is not advisable to rely on a single sensor for any of the autonomous driving tasks. The practical solution is to incorporate multiple competitive and complementary sensors that work synergistically to overcome their individual shortcomings. This article provides a comprehensive review of the state-of-the-art methods utilized to improve the performance of AV systems in short-range or local vehicle environments. Specifically, it focuses on recent studies that use deep learning sensor fusion algorithms for perception, localization, and mapping. The article concludes by highlighting some of the current trends and possible future research directions.

Sensor and Sensor Fusion Technology in Autonomous Vehicles: A Review

“…The research on multi-sensor fusion systems in AVs for environment perception and object detection is well-established in the literature [ 19 , 21 , 30 , 167 , 176 , 177 , 178 ]. Presently, three primary sensor combinations for obstacle detection are prevalent in the literature, including camera-LiDAR (CL); camera-radar (CR); and camera-LiDAR-radar (CLR) sensor combinations.…”

“…Ultimately, a sensor is only considered “smart” when the computer resources is an integral part of the physical sensor design [ 18 ]. Invariably, the overall performance of an AV system is greatly enhanced with multiple sensors of different types (smart/non-smart) and modalities (visual, infrared and radio waves) operating at different range and bandwidth (data rate) and with the data of each being incorporated to produce a fused output [ 17 , 18 , 19 ]. Multi-sensor fusion is effectively now a requisite process in all AD systems to overcome the shortcomings of individual sensor types, improving the efficiency and reliability of the overall AD system.…”

“…Several reviews have been published recently on the topic of multi-sensor fusion, some of them describing the architectural structure and sensor technologies in AVs [ 15 , 20 , 21 ], or focusing on the processing stages like sensor calibration, state estimation, object and tracking [ 22 , 23 , 24 ], or detailing techniques used for multi-sensor fusion, like deep learning-based approaches [ 19 , 25 , 26 ]. Table 1 below summarizes some of the recent studies in sensor and sensor fusion technologies in AD systems.…”

Machine Learning Methods in Drug Discovery

“…The basis of algorithm-based techniques, such as ML, requires a heavy mathematical and computational theory. ML models have been used in many promising technologies, such as deep learning (DL) assisted self-driving cars, advanced speech recognition, and support vector machine-based smarter search engines [ 1 , 2 , 3 , 4 ]. The advent of these computer-assisted computational techniques, first explored in the 1950s, has already been used in drug discovery, bioinformatics, cheminformatics, etc.…”

Development of an Improved Rapidly Exploring Random Trees Algorithm for Static Obstacle Avoidance in Autonomous Vehicles

“…The methods of path planning can be summarized into four categories: the graph search method, the sampling method, the interpolating method, and the numerical optimization method. Among them, numerical optimization using deep learning neural networks (DNN) has been the recent focus [ 10 ]. The Markov random field model was applied to path planning [ 11 ], but as in any numerical optimization, the computational load is often too heavy for real-time applications at present.…”