Navigation of Autonomous Vehicles Utilizing Hybrid Sensor Fusion Techniques

Engineering notes

Key topics: autonomous driving, autonomous vehicle, lidar, sensor fusion, control. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Numerous sensor integration frameworks have been presented using multiple setups, sensor combinations, and fusion methodologies. Most studies focused on optimizing accuracy and efficiency, but research is still lacking in incorporating these methods into autonomous vehicles (AVs). Some fusion frameworks might perform exceptionally well in lab environments with abundant computing power; however, their use in integrated edge computing (EC) poses severe challenges due to cost and high computational needs. This study proposes a hybrid sensor fusion technique (HSFT) to enhance autonomous vehicle navigation systems based on deep Q networks (DQN). The approach features the control of flow for camera and LiDAR data and the synthesis of an independent algorithm for object identification based on sensor image data. Noise suppression has been implemented using a Federated Kalman Filter (FKF). Contrast enhancement has been achieved with the help of Homomorphic Filtering (HMF) and adaptive thresholding. Recognition and classification have been performed using the YOLO V7 model. The proposed work is evaluated based on requirements such as speed and accuracy rate.

Links and sources

• Official / arXiv page

Comments