In the problem of calculating hull elements using the table of offsets, the sparsity between hull slices will bring uncertainty and error to the calculation. Therefore, this paper proposes a refinement algorithm of the ship hull based on the table of offsets: Firstly, the NURBS curve for the hull is constructed based on the table of offsets, and the hull's NURBS surface is obtained through the skinning algorithm. Secondly, the IR-BFS algorithm is used to inverse the knot parameters of the stations of the target station in the hull's NURBS surface. Thirdly, based on the knot parameters and the hull NURBS surface expression, the hull section, after refinement of the target station, is obtained. In constructing the hull's NURBS surface, the hull section is first expressed using the NURBS interpolation algorithm and the flattening algorithm of the NURBS based on the IR-BFS algorithm. Then the skinning algorithm is improved by fixing the -direction knot parameters to express the expressed hull NURBS cross-section as a hull's NURBS surface, which improves the computational efficiency. The effectiveness of the improved skinning algorithm is judged by comparing the increase in the number of control points and the computational time consumption in the expression of the hull NURBS surface before and after the improved skinning algorithm. The usability of the refinement algorithm of the hull section is verified by comparing the hull section based on the table of offsets with the refined hull section. The experimental results show that the improved skinning algorithm can effectively improve the speed of NURBS surface generation; The proposed refinement algorithm of the hull section can effectively generate refined sections through refinement intervals.
{"title":"Fast NURBS Skinning Algorithm and Ship Hull Section Refinement Model","authors":"Kaige Zhu, Guoyou Shi, Jiao Liu, Jiahui Shi, Yuchuang Wang, Xing Jiang","doi":"10.1145/3583788.3583792","DOIUrl":"https://doi.org/10.1145/3583788.3583792","url":null,"abstract":"In the problem of calculating hull elements using the table of offsets, the sparsity between hull slices will bring uncertainty and error to the calculation. Therefore, this paper proposes a refinement algorithm of the ship hull based on the table of offsets: Firstly, the NURBS curve for the hull is constructed based on the table of offsets, and the hull's NURBS surface is obtained through the skinning algorithm. Secondly, the IR-BFS algorithm is used to inverse the knot parameters of the stations of the target station in the hull's NURBS surface. Thirdly, based on the knot parameters and the hull NURBS surface expression, the hull section, after refinement of the target station, is obtained. In constructing the hull's NURBS surface, the hull section is first expressed using the NURBS interpolation algorithm and the flattening algorithm of the NURBS based on the IR-BFS algorithm. Then the skinning algorithm is improved by fixing the -direction knot parameters to express the expressed hull NURBS cross-section as a hull's NURBS surface, which improves the computational efficiency. The effectiveness of the improved skinning algorithm is judged by comparing the increase in the number of control points and the computational time consumption in the expression of the hull NURBS surface before and after the improved skinning algorithm. The usability of the refinement algorithm of the hull section is verified by comparing the hull section based on the table of offsets with the refined hull section. The experimental results show that the improved skinning algorithm can effectively improve the speed of NURBS surface generation; The proposed refinement algorithm of the hull section can effectively generate refined sections through refinement intervals.","PeriodicalId":292167,"journal":{"name":"Proceedings of the 2023 7th International Conference on Machine Learning and Soft Computing","volume":"58 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126453086","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Changhao Wang, Jin Xi, Changqing Xia, Chi Xu, Yong Duan
Indoor positioning services are being used more and more widely. However, existing indoor positioning techniques cannot simultaneously take into account low cost, ease of use, high precision, and seamless switching between indoor and outdoor positioning. With the maturity of 5G techniques, 5G-based indoor positioning is gradually being paid attention to. 5G-based indoor positioning does not require additional equipment, and supports flexible indoor and outdoor switching under the same system. However, the 5G-related information used in existing research on 5G indoor positioning is not open to users. Therefore, in this paper, we propose an indoor fingerprint positioning method based on measured 5G signals. This method first collects 5G signals in the positioning area, and processes them to form a fingerprint database. Then, a machine learning algorithm is used to match the signal to be located with the fingerprint database to obtain the positioning result. Finally, we conduct experiments in real field, and the experimental result demonstrates that the positioning accuracy of our proposed method can reach 96%.
{"title":"Indoor Fingerprint Positioning Method Based on Real 5G Signals","authors":"Changhao Wang, Jin Xi, Changqing Xia, Chi Xu, Yong Duan","doi":"10.1145/3583788.3583819","DOIUrl":"https://doi.org/10.1145/3583788.3583819","url":null,"abstract":"Indoor positioning services are being used more and more widely. However, existing indoor positioning techniques cannot simultaneously take into account low cost, ease of use, high precision, and seamless switching between indoor and outdoor positioning. With the maturity of 5G techniques, 5G-based indoor positioning is gradually being paid attention to. 5G-based indoor positioning does not require additional equipment, and supports flexible indoor and outdoor switching under the same system. However, the 5G-related information used in existing research on 5G indoor positioning is not open to users. Therefore, in this paper, we propose an indoor fingerprint positioning method based on measured 5G signals. This method first collects 5G signals in the positioning area, and processes them to form a fingerprint database. Then, a machine learning algorithm is used to match the signal to be located with the fingerprint database to obtain the positioning result. Finally, we conduct experiments in real field, and the experimental result demonstrates that the positioning accuracy of our proposed method can reach 96%.","PeriodicalId":292167,"journal":{"name":"Proceedings of the 2023 7th International Conference on Machine Learning and Soft Computing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129176854","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Deep learning algorithms mostly have network parameters that can affect their training results, and the combination of neural network architectures also has a significant impact on the algorithm performance. The performance of deep learning algorithms is usually proportional to the overall number of network parameters, leading to excessive resource consumption for exploring neural network architectures with a large number of hyper-parameters. To solve this problem, a vector representation is proposed which for neural network architectures, and a multi-objective optimization model is established based on genetic algorithms in this paper, and it is short for “NNOO Vector Representation based on GA and Its Optimization Method”. The multi-objective optimization model can automatically optimize the neural network architecture and hyper-parameters in the network, improve the network accuracy, and reduce the overall number of network parameters. It is shown in the test results with the MNIST data set, and the accuracy is 95.61% for the traditional empirical setting network, and the average accuracy is 86.2% for the network optimized by TensorFlow’s optimization algorithm. While the network accuracy is improved to 96.86% with the proposed optimization method in this paper and the network parameters are reduced by 32.6% compared with the traditional empirical network, and the network parameters are reduced by13.2% compared with the network by TensorFlow’s optimization algorithm. Therefore, the method is presented which has obvious practical application value in neural network optimization problems and provides a new way of thinking for large and deep network optimization problems.
{"title":"Neural Network Optimization Objective Vector Representation based on Genetic Algorithm and Its Multi-objective Optimization Method","authors":"Yunke Xiong, Qun Hou, Xin Liu","doi":"10.1145/3583788.3583796","DOIUrl":"https://doi.org/10.1145/3583788.3583796","url":null,"abstract":"Deep learning algorithms mostly have network parameters that can affect their training results, and the combination of neural network architectures also has a significant impact on the algorithm performance. The performance of deep learning algorithms is usually proportional to the overall number of network parameters, leading to excessive resource consumption for exploring neural network architectures with a large number of hyper-parameters. To solve this problem, a vector representation is proposed which for neural network architectures, and a multi-objective optimization model is established based on genetic algorithms in this paper, and it is short for “NNOO Vector Representation based on GA and Its Optimization Method”. The multi-objective optimization model can automatically optimize the neural network architecture and hyper-parameters in the network, improve the network accuracy, and reduce the overall number of network parameters. It is shown in the test results with the MNIST data set, and the accuracy is 95.61% for the traditional empirical setting network, and the average accuracy is 86.2% for the network optimized by TensorFlow’s optimization algorithm. While the network accuracy is improved to 96.86% with the proposed optimization method in this paper and the network parameters are reduced by 32.6% compared with the traditional empirical network, and the network parameters are reduced by13.2% compared with the network by TensorFlow’s optimization algorithm. Therefore, the method is presented which has obvious practical application value in neural network optimization problems and provides a new way of thinking for large and deep network optimization problems.","PeriodicalId":292167,"journal":{"name":"Proceedings of the 2023 7th International Conference on Machine Learning and Soft Computing","volume":"503 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115221176","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Chatbots trained on large corpus generate fluent responses, but often suffer from the problem of generating responses that contradict past utterances. Recent research treats dialogue contradiction detection as a task of natural language inference (NLI), and a method to remove contradiction from responses has been proposed and has shown high performance. However, these datasets do not provide explicit information about emotions, and these models cannot capture changes in emotions. In this work, we create a new dataset by explicitly labeling emotional information on an existing contradiction detection dataset and use this dataset to train an NLI model. Furthermore, we train the NLI model on the original dataset as well and compare the accuracy of both in dialogue contradiction detection.
{"title":"Resolving Context Contradictions in the Neural Dialogue System based on Sentiment Information","authors":"Shingo Hanahira, Xin Kang","doi":"10.1145/3583788.3583816","DOIUrl":"https://doi.org/10.1145/3583788.3583816","url":null,"abstract":"Chatbots trained on large corpus generate fluent responses, but often suffer from the problem of generating responses that contradict past utterances. Recent research treats dialogue contradiction detection as a task of natural language inference (NLI), and a method to remove contradiction from responses has been proposed and has shown high performance. However, these datasets do not provide explicit information about emotions, and these models cannot capture changes in emotions. In this work, we create a new dataset by explicitly labeling emotional information on an existing contradiction detection dataset and use this dataset to train an NLI model. Furthermore, we train the NLI model on the original dataset as well and compare the accuracy of both in dialogue contradiction detection.","PeriodicalId":292167,"journal":{"name":"Proceedings of the 2023 7th International Conference on Machine Learning and Soft Computing","volume":"113 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128107619","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Semantic segmentation has been a core learning task in the autonomous driving technology stack. However, current deep learning-based models do not perform well at nighttime due to the low illumination. In this study, we present an instance-level data augmentation method to increase the quantity and diversity for the low-resource classes to feed more instances of these classes to the training algorithm, with an aim to encourage the model to learn more features and patterns to better distinguish the low-resource classes presented in the original training set. We validate the method on the Dark Zurich dataset, a typical dataset that contains driving scene images taking at daytime e, twilight, and nighttime. We take the ``person'' class as an example and apply the instance-level data augmentation method. Experimental results have shown significant improvement compared to the SOTA, lifting the IoU by 4.52%. The results demonstrate the efficacy of the proposed method, indicating that the augmenting low-resource classes at the instance level is a promising strategy and can be an effective complement alongside other performance boosting methods.
{"title":"Nighttime Semantic Segmentation with Instance-level Data Augmentation: a Case Study of the Dark Zurich Benchmark","authors":"Alex Liu, Zhifeng Xiao","doi":"10.1145/3583788.3583814","DOIUrl":"https://doi.org/10.1145/3583788.3583814","url":null,"abstract":"Semantic segmentation has been a core learning task in the autonomous driving technology stack. However, current deep learning-based models do not perform well at nighttime due to the low illumination. In this study, we present an instance-level data augmentation method to increase the quantity and diversity for the low-resource classes to feed more instances of these classes to the training algorithm, with an aim to encourage the model to learn more features and patterns to better distinguish the low-resource classes presented in the original training set. We validate the method on the Dark Zurich dataset, a typical dataset that contains driving scene images taking at daytime e, twilight, and nighttime. We take the ``person'' class as an example and apply the instance-level data augmentation method. Experimental results have shown significant improvement compared to the SOTA, lifting the IoU by 4.52%. The results demonstrate the efficacy of the proposed method, indicating that the augmenting low-resource classes at the instance level is a promising strategy and can be an effective complement alongside other performance boosting methods.","PeriodicalId":292167,"journal":{"name":"Proceedings of the 2023 7th International Conference on Machine Learning and Soft Computing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133376453","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper mainly explores the method of Chinese cursive character recognition, establishes the Standard Cursive database in the process of research, and puts forward the similarity distance to measure the similarity between the cursive font to be recognized and the character set, and improves the calculation method of the similarity distance. Through experimental comparison, pH algorithm performs best in cursive character recognition.
{"title":"Research on Cursive Font Recognition Based on Improved Hash Algorithm","authors":"Benguo Yu, Yinqing Tang, Yang Yang","doi":"10.1145/3583788.3583808","DOIUrl":"https://doi.org/10.1145/3583788.3583808","url":null,"abstract":"This paper mainly explores the method of Chinese cursive character recognition, establishes the Standard Cursive database in the process of research, and puts forward the similarity distance to measure the similarity between the cursive font to be recognized and the character set, and improves the calculation method of the similarity distance. Through experimental comparison, pH algorithm performs best in cursive character recognition.","PeriodicalId":292167,"journal":{"name":"Proceedings of the 2023 7th International Conference on Machine Learning and Soft Computing","volume":"2016 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127462746","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Chronic diseases are serious threats to human safety and major public health problems worldwide. Many chronic diseases tend to have co-morbidities. Most machine learning techniques nowadays tend to focus on predicting a single disease while ignoring the study of co-morbidities. It is urgent to develop an artificial intelligence-based multi-label classification model based on patients' physical data, which is useful for the early detection and treatment of patients' diseases. In this study, we proposed a layer-by-layer processing structure, termed CascadeTransformer, that applies the Transformer architecture as weak classifiers, to solve the multi-label prediction problem of chronic diseases. We built a chronic diseases dataset using real-world data from West China Hospital, which consists of 1174 anonymous instances and 131 features. Systematic experiments show that our method shows better experimental performance compared to other methods on our chronic disease dataset.
{"title":"CascadeTransformer: Multi-label Classification with Transformer in Chronic Disease Prediction","authors":"Bo Zeng, Donghai Zhai, Bo Peng, Y. Yao","doi":"10.1145/3583788.3583817","DOIUrl":"https://doi.org/10.1145/3583788.3583817","url":null,"abstract":"Chronic diseases are serious threats to human safety and major public health problems worldwide. Many chronic diseases tend to have co-morbidities. Most machine learning techniques nowadays tend to focus on predicting a single disease while ignoring the study of co-morbidities. It is urgent to develop an artificial intelligence-based multi-label classification model based on patients' physical data, which is useful for the early detection and treatment of patients' diseases. In this study, we proposed a layer-by-layer processing structure, termed CascadeTransformer, that applies the Transformer architecture as weak classifiers, to solve the multi-label prediction problem of chronic diseases. We built a chronic diseases dataset using real-world data from West China Hospital, which consists of 1174 anonymous instances and 131 features. Systematic experiments show that our method shows better experimental performance compared to other methods on our chronic disease dataset.","PeriodicalId":292167,"journal":{"name":"Proceedings of the 2023 7th International Conference on Machine Learning and Soft Computing","volume":"182 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117170056","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Fayas Asharindavida, O. Nibouche, J. Uhomoibhi, Jun Liu, Hui Wang
Food quality analysis can be carried out by spectral data acquired from spectrometers with its advantage of non-destructive way of testing. Portable and miniature spectroscopy can be a suitable solution when it meets the specifications such as portability, cost, and short processing time requirements, to enable ordinary citizens to use such a device in the fight against food fraud. Compared to more expensive, bulky, and non-portable devices, the data collected using miniature and portable spectrometers is of a lower quality and thus adversely affect the quality of the analysis. Research have been carried out to use machine learning (ML) classifiers on spectral data analysis for food quality assessment. The present work focuses on two aspects: firstly, preliminary exploratory statistical analysis is conducted on the real spectral data on different food products including oils, fruits and spices acquired from such miniature devices, which aims to evaluate and illustrate the distinctive characteristics of such spectral data, data distribution and difference in the spectra across multiple data acquisitions etc. along with a summary of the key challenges to face and explore. Secondly, a case study for the differentiation of extra virgin olive from adulterated with vegetable oil is provided to analyze and evaluate how some commonly used ML classifiers can be used for classification, while the impact of different preprocessing methods to improve the accuracy and efficiency is also provided. The case study demonstrates the good potential of using data analytics for spectral data from miniature device, although the overall performance of those ML classifiers is not exceptional (the classification rates of up to 83.32%) which is partially due to the quality of data, and partially due to limiting to only some classifiers. More elaborate data pre-processing and cleaning methods can be used to address the key challenges of the spectral data from miniature device, and other types of classifiers can be also explored further in future work.
{"title":"Machine Learning on Spectral Data from Miniature Devices for Food Quality Analysis - A Case Study","authors":"Fayas Asharindavida, O. Nibouche, J. Uhomoibhi, Jun Liu, Hui Wang","doi":"10.1145/3583788.3583801","DOIUrl":"https://doi.org/10.1145/3583788.3583801","url":null,"abstract":"Food quality analysis can be carried out by spectral data acquired from spectrometers with its advantage of non-destructive way of testing. Portable and miniature spectroscopy can be a suitable solution when it meets the specifications such as portability, cost, and short processing time requirements, to enable ordinary citizens to use such a device in the fight against food fraud. Compared to more expensive, bulky, and non-portable devices, the data collected using miniature and portable spectrometers is of a lower quality and thus adversely affect the quality of the analysis. Research have been carried out to use machine learning (ML) classifiers on spectral data analysis for food quality assessment. The present work focuses on two aspects: firstly, preliminary exploratory statistical analysis is conducted on the real spectral data on different food products including oils, fruits and spices acquired from such miniature devices, which aims to evaluate and illustrate the distinctive characteristics of such spectral data, data distribution and difference in the spectra across multiple data acquisitions etc. along with a summary of the key challenges to face and explore. Secondly, a case study for the differentiation of extra virgin olive from adulterated with vegetable oil is provided to analyze and evaluate how some commonly used ML classifiers can be used for classification, while the impact of different preprocessing methods to improve the accuracy and efficiency is also provided. The case study demonstrates the good potential of using data analytics for spectral data from miniature device, although the overall performance of those ML classifiers is not exceptional (the classification rates of up to 83.32%) which is partially due to the quality of data, and partially due to limiting to only some classifiers. More elaborate data pre-processing and cleaning methods can be used to address the key challenges of the spectral data from miniature device, and other types of classifiers can be also explored further in future work.","PeriodicalId":292167,"journal":{"name":"Proceedings of the 2023 7th International Conference on Machine Learning and Soft Computing","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128145241","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Long-tail problem is one of the major challenges in distantly supervised relation extraction. Some recent works on the long-tail problem attempt to transfer knowledge from data-rich and semantically similar head classes to data-poor tail classes using a relation hierarchical tree. These methods, however, are based on the assumption that long-tail and head relations have a strong correlation, which does not always hold true, and the model’s ability to learn long-tail relations is essentially not improved. In this paper, a novel joint learning framework that combines relation extraction and contrastive learning is proposed, allowing the model to directly learn the subtle differences between different categories to improve long-tail relation extraction. Experimental results show that our proposed model outperforms the current state-of-the-art (SOTA) model on various mainstream datasets.
{"title":"LDRC: Long-tail Distantly Supervised Relation Extraction via Contrastive Learning","authors":"Tingwei Li, Zhi Wang","doi":"10.1145/3583788.3583804","DOIUrl":"https://doi.org/10.1145/3583788.3583804","url":null,"abstract":"Long-tail problem is one of the major challenges in distantly supervised relation extraction. Some recent works on the long-tail problem attempt to transfer knowledge from data-rich and semantically similar head classes to data-poor tail classes using a relation hierarchical tree. These methods, however, are based on the assumption that long-tail and head relations have a strong correlation, which does not always hold true, and the model’s ability to learn long-tail relations is essentially not improved. In this paper, a novel joint learning framework that combines relation extraction and contrastive learning is proposed, allowing the model to directly learn the subtle differences between different categories to improve long-tail relation extraction. Experimental results show that our proposed model outperforms the current state-of-the-art (SOTA) model on various mainstream datasets.","PeriodicalId":292167,"journal":{"name":"Proceedings of the 2023 7th International Conference on Machine Learning and Soft Computing","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124765204","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jiao Liu, Guoyou Shi, Kaige Zhu, Jiahui Shi, Yuchuang Wang
Aiming at the problems that the current decision-making model of ship collision avoidance does not consider International Regulations for Preventing Collisions at Sea (COLREGS), ship maneuverability, and the need for a lot of training time, combined with the advantages of reinforcement learning and imitation learning, a ship intelligent collision avoidance decision-making model based on Generic Adversary Imitation Learning (GAIL) is proposed: Firstly, the collision avoidance data in Automatic Information System (AIS) data is extracted as expert data; Secondly, in the generator part, the environment model is established based on Mathematical Model Group (MMG) and S-57 chart rendering, and the state space, behaviour space and reward function of reinforcement learning are constructed. The deep deterministic policy gradient (DDPG) is used to interact with the environment model to generate ship trajectory data. At the same time, the generator can constantly learn expert data; Finally, a discriminator can distinguish the expert data from the data generated by the generator is constructed and trained. The model training is completed when the discriminator cannot distinguish the two. In order to verify the performance of the model, AIS data near the South China Sea is used to process and extract collision avoidance decision data, and a ship intelligent collision avoidance decision model based on GAIL is established. After the model converges, the final generated data is compared with the expert data. The experimental results verify that the model proposed in this paper can reproduce the expert collision avoidance trajectory and is a practical decision model of ship collision avoidance.
{"title":"Decision Model of Ship Intelligent Collision Avoidance Based on Automatic Information System Data and Generic Adversary Imitation Learning-Deep Deterministic Policy Gradient","authors":"Jiao Liu, Guoyou Shi, Kaige Zhu, Jiahui Shi, Yuchuang Wang","doi":"10.1145/3583788.3583790","DOIUrl":"https://doi.org/10.1145/3583788.3583790","url":null,"abstract":"Aiming at the problems that the current decision-making model of ship collision avoidance does not consider International Regulations for Preventing Collisions at Sea (COLREGS), ship maneuverability, and the need for a lot of training time, combined with the advantages of reinforcement learning and imitation learning, a ship intelligent collision avoidance decision-making model based on Generic Adversary Imitation Learning (GAIL) is proposed: Firstly, the collision avoidance data in Automatic Information System (AIS) data is extracted as expert data; Secondly, in the generator part, the environment model is established based on Mathematical Model Group (MMG) and S-57 chart rendering, and the state space, behaviour space and reward function of reinforcement learning are constructed. The deep deterministic policy gradient (DDPG) is used to interact with the environment model to generate ship trajectory data. At the same time, the generator can constantly learn expert data; Finally, a discriminator can distinguish the expert data from the data generated by the generator is constructed and trained. The model training is completed when the discriminator cannot distinguish the two. In order to verify the performance of the model, AIS data near the South China Sea is used to process and extract collision avoidance decision data, and a ship intelligent collision avoidance decision model based on GAIL is established. After the model converges, the final generated data is compared with the expert data. The experimental results verify that the model proposed in this paper can reproduce the expert collision avoidance trajectory and is a practical decision model of ship collision avoidance.","PeriodicalId":292167,"journal":{"name":"Proceedings of the 2023 7th International Conference on Machine Learning and Soft Computing","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114453801","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：