Perception and Decision-Making of Autonomous Systems in the Era of Learning: An Overview
share
The ability of inferring its own ego-motion, autonomous understanding the surroundings and planning trajectory are the features of autonomous robot systems. Therefore, in order to improve the perception and decision-making ability of autonomous robot systems, Simultaneous localization and mapping (SLAM) and autonomous trajectory planning are widely researched, especially combining learning-based algorithms. In this paper, we mainly focus on the application of learning-based approaches to perception and decision-making, and survey the current state-of-the-art SLAM and trajectory planning algorithms. Firstly, we delineate the existing classical SLAM solutions and review deep learning-based methods of environment perception and understanding, including deep learning-based monocular depth estimation, ego-motion prediction, image enhancement, object detection, semantic segmentation and their combinations with previous SLAM frameworks. Secondly, we summarize the existing path planning and trajectory planning methods and survey the navigation methods based on reinforcement learning. Finally, several challenges and promising directions are concluded to related researchers for future work in both autonomous perception and decision-making.
READ FULL TEXT
