{"title":"轮胎-土壤切向力强化学习模型","authors":"Yingchun Qi, Jiaqi Zhao, Ye Zhuang","doi":"10.56884/rhbe9228","DOIUrl":null,"url":null,"abstract":"Tire-soil tangential interaction involves complex terramechanics. When modeling the tire-soil tangential forces, a great amount of experiments is needed to identify the parameters in the currently available models. The efficiency and accuracy of the current models is still time and cost consuming. The control of the terrain vehicle is introduced gradually to the off-road vehicles. Such application requires more accurate and real-time tire-soil force models. Therefore, the machine learning technique, the reinforcement learning algorithm, is introduced to the tire-soil tangential force modelling. First, the tire-soil rolling experiment is carried out under longitudinal and lateral slip condition with the tire-soil test facility. The tire-soil forces vs slip ratios test data is obtained on the sand and mud road surfaces. The reinforcement learning model, which including the physical interpretation and the uncertainty (with Gaussian Process), is proposed. The model parameters is identified through the supervised learning (training) by the model from the acquired experimental data. The model accuracy could be improved gradually with the iterative off-line learning (training). The trained model could calculate the force-vs-slip ratio relationship with high accuracy and efficiency. The proposed model could also be updated with the new data learning.","PeriodicalId":447600,"journal":{"name":"Proceedings of the 11th Asia-Pacific Regional Conference of the ISTVS","volume":"71 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Tire-Soil Tangential Force Reinforcement Learning Modeling\",\"authors\":\"Yingchun Qi, Jiaqi Zhao, Ye Zhuang\",\"doi\":\"10.56884/rhbe9228\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Tire-soil tangential interaction involves complex terramechanics. When modeling the tire-soil tangential forces, a great amount of experiments is needed to identify the parameters in the currently available models. The efficiency and accuracy of the current models is still time and cost consuming. The control of the terrain vehicle is introduced gradually to the off-road vehicles. Such application requires more accurate and real-time tire-soil force models. Therefore, the machine learning technique, the reinforcement learning algorithm, is introduced to the tire-soil tangential force modelling. First, the tire-soil rolling experiment is carried out under longitudinal and lateral slip condition with the tire-soil test facility. The tire-soil forces vs slip ratios test data is obtained on the sand and mud road surfaces. The reinforcement learning model, which including the physical interpretation and the uncertainty (with Gaussian Process), is proposed. The model parameters is identified through the supervised learning (training) by the model from the acquired experimental data. The model accuracy could be improved gradually with the iterative off-line learning (training). The trained model could calculate the force-vs-slip ratio relationship with high accuracy and efficiency. The proposed model could also be updated with the new data learning.\",\"PeriodicalId\":447600,\"journal\":{\"name\":\"Proceedings of the 11th Asia-Pacific Regional Conference of the ISTVS\",\"volume\":\"71 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2022-10-10\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Proceedings of the 11th Asia-Pacific Regional Conference of the ISTVS\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.56884/rhbe9228\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the 11th Asia-Pacific Regional Conference of the ISTVS","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.56884/rhbe9228","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Tire-Soil Tangential Force Reinforcement Learning Modeling
Tire-soil tangential interaction involves complex terramechanics. When modeling the tire-soil tangential forces, a great amount of experiments is needed to identify the parameters in the currently available models. The efficiency and accuracy of the current models is still time and cost consuming. The control of the terrain vehicle is introduced gradually to the off-road vehicles. Such application requires more accurate and real-time tire-soil force models. Therefore, the machine learning technique, the reinforcement learning algorithm, is introduced to the tire-soil tangential force modelling. First, the tire-soil rolling experiment is carried out under longitudinal and lateral slip condition with the tire-soil test facility. The tire-soil forces vs slip ratios test data is obtained on the sand and mud road surfaces. The reinforcement learning model, which including the physical interpretation and the uncertainty (with Gaussian Process), is proposed. The model parameters is identified through the supervised learning (training) by the model from the acquired experimental data. The model accuracy could be improved gradually with the iterative off-line learning (training). The trained model could calculate the force-vs-slip ratio relationship with high accuracy and efficiency. The proposed model could also be updated with the new data learning.