Tianyu Wang, Yuhang Ye, Zihan Zhang, Haoran Zhang, Zonghan Yang
{"title":"焦点层-将注意力吸引到必要的障碍上","authors":"Tianyu Wang, Yuhang Ye, Zihan Zhang, Haoran Zhang, Zonghan Yang","doi":"10.1109/wsai55384.2022.9836381","DOIUrl":null,"url":null,"abstract":"With the development of automatic driving, fast and safe motion planning algorithms are in need. However, data transferred to the motion planning part may be noisy, and some obstacles are unnecessary for later processing. This paper proposes a focus layer and a DQN agent to select necessary barriers and submit them to the motion planning algorithms. The Focus layer ignores some obstacles that are not likely to impact the ego vehicle and focuses attention on those critical obstacles. Note to Practitioners: This paper is motivated by the heavy computation time in automatic driving when planning a trajectory. Constraints such as obstacles along the road affect the efficiency of the planning methodology. Existing research conducts experiments on capturing drivers' facial expressions or eye contact when driving on the road. However, such research cannot fit into the automatic driving algorithms. Thus, we propose a method to reduce unnecessary obstacles in a simulation environment, which is similar to focusing on the essential elements for drivers. Our process generates a layer to focus ego vehicles' attention on critical obstacles before the trajectory planning algorithm and can easily fit in all trajectory planning algorithms.","PeriodicalId":402449,"journal":{"name":"2022 4th World Symposium on Artificial Intelligence (WSAI)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Focus Layer - Drawing Attention to Necessary Obstacles\",\"authors\":\"Tianyu Wang, Yuhang Ye, Zihan Zhang, Haoran Zhang, Zonghan Yang\",\"doi\":\"10.1109/wsai55384.2022.9836381\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"With the development of automatic driving, fast and safe motion planning algorithms are in need. However, data transferred to the motion planning part may be noisy, and some obstacles are unnecessary for later processing. This paper proposes a focus layer and a DQN agent to select necessary barriers and submit them to the motion planning algorithms. The Focus layer ignores some obstacles that are not likely to impact the ego vehicle and focuses attention on those critical obstacles. Note to Practitioners: This paper is motivated by the heavy computation time in automatic driving when planning a trajectory. Constraints such as obstacles along the road affect the efficiency of the planning methodology. Existing research conducts experiments on capturing drivers' facial expressions or eye contact when driving on the road. However, such research cannot fit into the automatic driving algorithms. Thus, we propose a method to reduce unnecessary obstacles in a simulation environment, which is similar to focusing on the essential elements for drivers. Our process generates a layer to focus ego vehicles' attention on critical obstacles before the trajectory planning algorithm and can easily fit in all trajectory planning algorithms.\",\"PeriodicalId\":402449,\"journal\":{\"name\":\"2022 4th World Symposium on Artificial Intelligence (WSAI)\",\"volume\":\"47 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2022-06-23\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2022 4th World Symposium on Artificial Intelligence (WSAI)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/wsai55384.2022.9836381\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 4th World Symposium on Artificial Intelligence (WSAI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/wsai55384.2022.9836381","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Focus Layer - Drawing Attention to Necessary Obstacles
With the development of automatic driving, fast and safe motion planning algorithms are in need. However, data transferred to the motion planning part may be noisy, and some obstacles are unnecessary for later processing. This paper proposes a focus layer and a DQN agent to select necessary barriers and submit them to the motion planning algorithms. The Focus layer ignores some obstacles that are not likely to impact the ego vehicle and focuses attention on those critical obstacles. Note to Practitioners: This paper is motivated by the heavy computation time in automatic driving when planning a trajectory. Constraints such as obstacles along the road affect the efficiency of the planning methodology. Existing research conducts experiments on capturing drivers' facial expressions or eye contact when driving on the road. However, such research cannot fit into the automatic driving algorithms. Thus, we propose a method to reduce unnecessary obstacles in a simulation environment, which is similar to focusing on the essential elements for drivers. Our process generates a layer to focus ego vehicles' attention on critical obstacles before the trajectory planning algorithm and can easily fit in all trajectory planning algorithms.