Redirigiendo al acceso original de articulo en 15 segundos...
Inicio  /  Applied Sciences  /  Vol: 12 Par: 22 (2022)  /  Artículo
ARTÍCULO
TITULO

Imitative Reinforcement Learning Fusing Mask R-CNN Perception Algorithms

Lei He    
Jian Ou    
Mingyue Ba    
Guohong Deng and Echuan Yang    

Resumen

Autonomous urban driving navigation is still an open problem and has ample room for improvement in unknown complex environments. This paper proposes an end-to-end autonomous driving approach that combines Conditional Imitation Learning (CIL), Mask R-CNN with DDPG. In the first stage, data acquisition is first performed by using CARLA, a high-fidelity simulation software. Data collected by CARLA is used to train the Mask R-CNN network, which is used for object detection and segmentation. The segmented images are transformed into the backbone of CIL to perform supervised Imitation Learning (IL). DDPG means using Reinforcement Learning for further training in the second stage, which shares the learned weights from the pre-trained CIL model. The combination of the two methods is an innovative way of considering. The benefit is that it is possible to speed up training considerably and obtain super-high levels of performance beyond humans. We conduct experiments on the CARLA driving benchmark of urban driving. In the final experiments, our algorithm outperforms the original MP by 30%, CIL by 33%, and CIRL by 10% in the most difficult tasks, dynamic navigation tasks, and in new environments and new weather, demonstrating that the two-stage framework proposed in this paper shows remarkable generalization capability in unknown environments on navigation tasks.

 Artículos similares

       
 
Woo Chul Choi and Kyu Soo Chong    
With the increasing popularity of digital navigation systems and smartphones, the role of road signs during driving is gradually diminishing. However, owing to the inaccessibility of the technology to certain portions of the population, e.g., the elderly... ver más
Revista: Applied Sciences

 
Wei Xiao, Mingxia Liu and Xubing Chen    
The underground intelligent load-haul-dump vehicle (LHD) is a product of the deep integration of traditional LHD with information network technology, automatic controlling and artificial intelligence technology. It gathers the functions of environmental ... ver más
Revista: Applied Sciences

 
Marco Antonio Chunab-Rodríguez, Alfredo Santana-Díaz, Jorge Rodríguez-Arce, Emilio Sánchez-Tapia and Carlos Alberto Balbuena-Campuzano    
In recent years, engineering degree programs have become fundamental to the teaching of robotics and incorporate many fundamental STEM concepts. Some authors have proposed different platforms for teaching different topics related to robotics, but most of... ver más
Revista: Applied Sciences

 
Chuanwei Wang, Saisai Wang, Hongwei Ma, Heng Zhang, Xusheng Xue, Haibo Tian and Lei Zhang    
Tracked inspection robots possess prominent advantages in dealing with severe environment rescue, safety inspection, and other important tasks, and have been used widely. However, tracked robots are affected by skidding and slipping, so it is difficult t... ver más
Revista: Applied Sciences

 
Marta Ribeiro, Joost Ellerbroek and Jacco Hoekstra    
Current predictions for future operations with drones estimate traffic densities orders of magnitude higher than any observed in manned aviation. Such densities call for further research and innovation, in particular, into conflict detection and resoluti... ver más
Revista: Aerospace