ParkingE2E: Camera-based End-to-end Parking Network, from Images to Planning

Changze Li,Ziheng Ji,Zhe Chen,Tong Qin,Ming Yang

2024-08-04

Abstract:Autonomous parking is a crucial task in the intelligent driving field. Traditional parking algorithms are usually implemented using rule-based schemes. However, these methods are less effective in complex parking scenarios due to the intricate design of the algorithms. In contrast, neural-network-based methods tend to be more intuitive and versatile than the rule-based methods. By collecting a large number of expert parking trajectory data and emulating human strategy via learning-based methods, the parking task can be effectively addressed. In this paper, we employ imitation learning to perform end-to-end planning from RGB images to path planning by imitating human driving trajectories. The proposed end-to-end approach utilizes a target query encoder to fuse images and target features, and a transformer-based decoder to autoregressively predict future waypoints. We conducted extensive experiments in real-world scenarios, and the results demonstrate that the proposed method achieved an average parking success rate of 87.8% across four different real-world garages. Real-vehicle experiments further validate the feasibility and effectiveness of the method proposed in this paper.

Computer Vision and Pattern Recognition,Artificial Intelligence,Robotics

What problem does this paper attempt to address?

The paper aims to address the problem of automatic parking in the field of autonomous driving. Traditional parking algorithms are usually implemented based on rule-based schemes, but they perform poorly in complex parking scenarios because these methods are very cumbersome to design and prone to errors. In contrast, neural network-based methods are more intuitive and flexible. By collecting a large amount of expert parking trajectory data and using learning-based methods to imitate human driving strategies, the parking task can be effectively solved.

The paper proposes an end-to-end parking network called ParkingE2E, which performs path planning directly from RGB images and completes the parking task by imitating human driving trajectories. Specifically, this method uses a target query encoder to fuse image and target features and employs a Transformer-based decoder to autoregressively predict future waypoints. Experimental results show that in four different real-world garages, the method achieved an average parking success rate of 87.8% and validated its feasibility and effectiveness in real vehicle experiments. Overall, this research is dedicated to developing a more reliable and general end-to-end parking solution.