Pub Date : 2025-11-21DOI: 10.1109/TIV.2025.3635064
Hyungtae Lim;Daebeom Kim;Hyun Myung
As various 3D light detection and ranging (LiDAR) sensors have been introduced to the market, research on multi-session simultaneous localization and mapping (MSS) using heterogeneous LiDAR sensors has been actively conducted. Existing MSS methods mostly rely on loop closure detection for inter-session alignment; however, the performance of loop closure detection can be potentially degraded owing to the differences in the density and field of view (FoV) of the sensors used in different sessions. In this study, we challenge the existing paradigm that relies heavily on loop detection modules and propose a novel MSS framework, called Multi-Mapcher, that employs large-scale map-to-map registration to perform inter-session initial alignment, which is commonly assumed to be infeasible, by leveraging outlier-robust 3D point cloud registration. Next, after finding inter-session loops by radius search based on the assumption that the inter-session initial alignment is sufficiently precise, anchor node-based robust pose graph optimization is employed to build a consistent global map. As demonstrated in our experiments, our approach shows substantially better MSS performance for various LiDAR sensors used to capture the sessions and is faster than state-of-the-art approaches.
{"title":"Multi-Mapcher: Loop Closure Detection-Free Heterogeneous LiDAR Multi-Session SLAM Leveraging Outlier-Robust Registration for Autonomous Vehicles","authors":"Hyungtae Lim;Daebeom Kim;Hyun Myung","doi":"10.1109/TIV.2025.3635064","DOIUrl":"https://doi.org/10.1109/TIV.2025.3635064","url":null,"abstract":"As various 3D light detection and ranging (LiDAR) sensors have been introduced to the market, research on multi-session simultaneous localization and mapping (MSS) using heterogeneous LiDAR sensors has been actively conducted. Existing MSS methods mostly rely on loop closure detection for inter-session alignment; however, the performance of loop closure detection can be potentially degraded owing to the differences in the density and field of view (FoV) of the sensors used in different sessions. In this study, we challenge the existing paradigm that relies heavily on loop detection modules and propose a novel MSS framework, called <italic>Multi-Mapcher</i>, that employs large-scale map-to-map registration to perform inter-session initial alignment, which is commonly assumed to be infeasible, by leveraging outlier-robust 3D point cloud registration. Next, after finding inter-session loops by radius search based on the assumption that the inter-session initial alignment is sufficiently precise, anchor node-based robust pose graph optimization is employed to build a consistent global map. As demonstrated in our experiments, our approach shows substantially better MSS performance for various LiDAR sensors used to capture the sessions and is faster than state-of-the-art approaches.","PeriodicalId":36532,"journal":{"name":"IEEE Transactions on Intelligent Vehicles","volume":"11 2","pages":"338-351"},"PeriodicalIF":14.3,"publicationDate":"2025-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146057658","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-19DOI: 10.1109/TIV.2025.3634557
Lei Wang;Huaming Wu;Keqiu Li;Wei Yu
Multivariate time series forecasting has extensive applications across various domains, including economics, finance, bioinformatics, and intelligent transportation. The inherent spatiotemporal data is characterized by pronounced nonlinearity and stochastic uncertainty. However, current deep learning-based methods all employ deterministic parameters to characterize data features. This approach fails to effectively capture the temporal and spatial uncertainty inherent in data, resulting in limited model capability to extract data features and reduced analytical prediction accuracy. To solve this problem, this paper proposes Spatiotemporal Bayesian Convolution Neural Networks, referred to as ST-BayesianNet, for enhancing multivariate time series forecasting. Specifically, we decompose the uncertainty of spatiotemporal data into space-time dimensions, thus facilitating the prediction of multivariate spatiotemporal sequences. First, we leverage a self-adaptive uncertainty adjacency matrix to model intricate uncertain spatial relationships, while the acquisition of knowledge for this uncertain matrix hinges upon judicious a priori assumptions. Then, a non-deterministic Temporal Bayesian Convolutional Neural Network (TBCN) is constructed to adeptly capture temporal uncertainty. The optimization of model parameters, comprising both deterministic and probabilistic aspects, is achieved through variational inference. Finally, the experimental results obtained from seven real-world datasets confirm that ST-BayesianNet is more accurate than baseline methods at making predictions.
{"title":"ST-BayesianNet: Spatiotemporal Bayesian Convolution Neural Networks for Multivariate Time Series Forecasting","authors":"Lei Wang;Huaming Wu;Keqiu Li;Wei Yu","doi":"10.1109/TIV.2025.3634557","DOIUrl":"https://doi.org/10.1109/TIV.2025.3634557","url":null,"abstract":"Multivariate time series forecasting has extensive applications across various domains, including economics, finance, bioinformatics, and intelligent transportation. The inherent spatiotemporal data is characterized by pronounced nonlinearity and stochastic uncertainty. However, current deep learning-based methods all employ deterministic parameters to characterize data features. This approach fails to effectively capture the temporal and spatial uncertainty inherent in data, resulting in limited model capability to extract data features and reduced analytical prediction accuracy. To solve this problem, this paper proposes Spatiotemporal Bayesian Convolution Neural Networks, referred to as ST-BayesianNet, for enhancing multivariate time series forecasting. Specifically, we decompose the uncertainty of spatiotemporal data into space-time dimensions, thus facilitating the prediction of multivariate spatiotemporal sequences. First, we leverage a self-adaptive uncertainty adjacency matrix to model intricate uncertain spatial relationships, while the acquisition of knowledge for this uncertain matrix hinges upon judicious a priori assumptions. Then, a non-deterministic Temporal Bayesian Convolutional Neural Network (TBCN) is constructed to adeptly capture temporal uncertainty. The optimization of model parameters, comprising both deterministic and probabilistic aspects, is achieved through variational inference. Finally, the experimental results obtained from seven real-world datasets confirm that ST-BayesianNet is more accurate than baseline methods at making predictions.","PeriodicalId":36532,"journal":{"name":"IEEE Transactions on Intelligent Vehicles","volume":"11 2","pages":"325-337"},"PeriodicalIF":14.3,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146057654","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-05DOI: 10.1109/TIV.2025.3629618
Peter Popov
This paper deals with the Bayesian safety assessment of autonomous vehicles (AV) conducted via driving AVs on the public roads, often referred to as “driving to safety.” A key safety measure is the probability of catastrophic failure (i.e., a road accident) per mile of driving (pfm), assumed a random variable. We argue that a Bayesian prediction based on a univariate (“black-box”) probabilistic model has an intrinsic deficiency: it cannot accommodate the variation of pfm due to changing road conditions, which in turn may affect significantly the predicted pfm and may lead to optimistic predictions. A multivariate probabilistic model is developed to overcome this limitation of the univariate model. Using a set of contrived examples the predictions of the multivariate model are compared with those derived with univariate models. Our results provide an intriguing insight that even when AV driving does not lead to accidents at all, the pfm predictions with the multivariate model may be more pessimistic than the assumed prior, and those derived with a black-box model, including the predictions using the recently developed “conservative Bayesian inference”. The multivariate Bayesian safety assessment can be applied to autonomous vehicles and to other complex intelligent systems such as robots, UAVs, etc., where the operating conditions vary.
{"title":"Why Black-Box Bayesian Safety Assessment of Autonomous Vehicles Is Problematic and What Can Be Done About It","authors":"Peter Popov","doi":"10.1109/TIV.2025.3629618","DOIUrl":"https://doi.org/10.1109/TIV.2025.3629618","url":null,"abstract":"This paper deals with the Bayesian safety assessment of autonomous vehicles (AV) conducted via driving AVs on the public roads, often referred to as “driving to safety.” A key safety measure is the probability of <italic>catastrophic failure</i> (i.e., a road accident) per mile of driving (<italic>pfm</i>), assumed a random variable. We argue that a Bayesian prediction based on a univariate (“black-box”) probabilistic model has an <italic>intrinsic deficiency</i>: it cannot accommodate the variation of <italic>pfm</i> due to changing road conditions, which in turn may affect significantly the predicted <italic>pfm</i> and may lead to optimistic predictions. A multivariate probabilistic model is developed to overcome this limitation of the univariate model. Using a set of contrived examples the predictions of the multivariate model are compared with those derived with univariate models. Our results provide an intriguing insight that even when AV driving does not lead to accidents at all, the <italic>pfm</i> predictions with the multivariate model may be <italic>more pessimistic</i> than the assumed prior, and those derived with a black-box model, including the predictions using the recently developed “conservative Bayesian inference”. The multivariate Bayesian safety assessment can be applied to autonomous vehicles and to other complex intelligent systems such as robots, UAVs, etc., where the operating conditions vary.","PeriodicalId":36532,"journal":{"name":"IEEE Transactions on Intelligent Vehicles","volume":"11 2","pages":"311-324"},"PeriodicalIF":14.3,"publicationDate":"2025-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146057616","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","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：