Target-Driven Visual Navigation Using Causal Intervention

doi:10.1109/CCDC58219.2023.10327097

	Target-Driven Visual Navigation Using Causal Intervention
	Zhao, Xinzhou 1; Wang, Tian 2; Liu, Kexin 3; Zhang, Baochang 4; Li, Ce5 ; Snoussi, Hichem 6
	2023
会议名称	35th Chinese Control and Decision Conference (CCDC)
会议录名称	IEEE
页码	3508-3513
会议日期	MAY 20-22, 2023
会议地点	Yichang, PEOPLES R CHINA
出版地	NEW YORK
摘要	Target-driven visual navigation has gained significance and presents great potentials in scientific and industrial fields. However, how to achieve faster convergence and better generalization is a challenging problem. One of the most critical hurdles is the neglect of confounders, which often leads to spurious correlations. Confounders make it difficult to discover the real causality and therefore are taken into consideration in some other fields. In this paper, we introduce a Causal Intervention Visual Navigation (CIVN) method, based on deep reinforcement learning and causal inference. We propose to realize causal intervention in navigation via front-door adjustment as most confounders are unobservable. Specifically, CIVN is implemented by Target-Related Shortcut, which serves as an approximation of causal intervention. To eliminate the confounding effect, we adapt cross-sampling and strengthen the target information. It is worth mentioning that causal intervention is for the first time applied by us in solving the confounding effect in target-driven visual navigation. Navigation results on AI2-THOR demonstrate that CIVN converges faster and achieves better evaluation performance than prior arts. Moreover, the generalization for unknown targets and scenes is also improved.
关键词	target-driven visual navigation causal intervention front-door adjustment
DOI	10.1109/CCDC58219.2023.10327097
收录类别	CPCI-S
语种	英语
WOS研究方向	Automation & Control Systems ; Operations Research & Management Science
WOS类目	Automation & Control Systems ; Operations Research & Management Science
WOS记录号	WOS:001116704303127
原始文献类型	Proceedings Paper
引用统计	无
文献类型	会议论文
条目标识符	https://ir.lut.edu.cn/handle/2XXMBERH/170212
专题	电气工程与信息工程学院
通讯作者	Wang, Tian
作者单位	1.Beihang Univ, Sch Automat Sci & Elect Engn, Beijing, Peoples R China; 2.Beihang Univ, Zhongguancun Lab, Inst Artificial Intelligence, SKLSDE, Beijing, Peoples R China; 3.Beihang Univ, Sch Automat Sci & Elect Engn, Zhongguancun Lab, Beijing, Peoples R China; 4.Beihang Univ, Inst Artificial Intelligence, Zhongguancun Lab, Beijing, Peoples R China; 5.Lanzhou Univ Technol, Coll Elect & Informat Engn, Lanzhou, Peoples R China; 6.Univ Technol Troyes, Inst Charles Delaunay LM2S, Troyes, France
推荐引用方式 GB/T 7714	Zhao, Xinzhou,Wang, Tian,Liu, Kexin,et al. Target-Driven Visual Navigation Using Causal Intervention[C]. NEW YORK,2023:3508-3513.