Eric Hsueh-Chan Lu

With the rapid development of smart buildings and Internet of Things (IoT) applications, indoor positioning technology plays an essential role in smart environments. However, wireless fingerprinting often suffers from signal distribution shifts caused by dynamic changes and device heterogeneity in real-world settings, which significantly affect localization accuracy and generalization ability, limiting its practical use. To address the challenge of adapting to new environments without the need for additional data labeling, this study proposes multimodal adaptation with signal consistency for localization (MASCLoc), an innovative indoor localization framework based on wireless fingerprinting that combines multimodal feature representation and regression-based domain adaptation. MASCLoc integrates multiple modalities, including received signal strength (RSS) vectors, relative fingerprint gradients (RFGs), and radio maps, and employs a spatial embedding strategy to enhance cross-modal semantic integration. It further extends the adversarial learning framework by incorporating cosine similarity and Pearson correlation losses. This design not only improves feature alignment across domains but also ensures greater physical consistency and stability in the predicted results. The experiments conducted on a long-term, cross-month dataset validate MASCLoc’s performance under various data scales and environmental changes. Results show that MASCLoc maintains an average localization error (ALE) below 3 m, demonstrating strong adaptability and generalization capability. Overall, this research presents a highly practical unsupervised indoor localization solution that reduces labeling costs and supports efficient long-term deployment in real-world applications.

Visual localization is essential for autonomous driving, robotics, and augmented reality, yet remains challenging under domain shifts caused by illumination, weather, and seasonal changes. Such variations create significant gaps between training and testing data, leading to degraded performance in existing methods. We propose an improved cross-domain localization framework based on Hu et al., where the original triplet loss is replaced by self-supervised contrastive learning strategies (MoCo and BYOL). This modification removes the reliance on manually defined image pairs, simplifies data preprocessing, and enhances generalization. In addition, multi-task supervision on semantic segmentation and depth prediction enriches scene understanding, while adversarial domain adaptation aligns real and virtual domains. Experiments on Virtual KITTI, KITTI, and the Extended CMU-Seasons datasets show consistent improvements over the baseline, with notable gains under challenging conditions such as mixed foliage and snow. To the best of our knowledge, this is the first study to apply self-supervised contrastive learning to cross-domain visual localization.

Lane detection is a pivotal technology of the intelligent driving system. By identifying the position and shape of the lane, the vehicle can stay in the correct lane and avoid accidents. Image-based deep learning is currently the most advanced method for lane detection. Models using this method already have a very good recognition ability in general daytime scenes, and can almost achieve real-time detection. However, these models often fail to accurately identify lanes in challenging scenarios such as night, dazzle, or shadows. Furthermore, the lack of diversity in the training data restricts the capacity of the models to handle different environments. This paper proposes a novel method to train CycleGAN with existing daytime and nighttime datasets. This method can extract features of different styles and multi-scales, thereby increasing the richness of model input. We use CycleGAN as a domain adaptation model combined with an image segmentation model to boost the model’s performance in different styles of scenes. The proposed consistent loss function is employed to mitigate performance disparities of the model in different scenarios. Experimental results indicate that our method enhances the detection performance of original lane detection models in challenging scenarios. This research helps improve the dependability and robustness of intelligent driving systems, ultimately making roads safer and enhancing the driving experience.

With the rapid development of the autonomous vehicles industry, there has been a dramatic proliferation of research concerned with related works, where road markings detection is an important issue. When there is no public open data in a field, we must collect road markings data and label them by ourselves manually, which is huge labor work and takes lots of time. Moreover, object detection often encounters the problem of small object detection. The detection accuracy often decreases when the detection distance increases. This is primarily because distant objects on the road take up few pixels in the image and object scales vary depending on different distances and perspectives. For the sake of solving the issues mentioned above, this paper utilizes a virtual dataset and open dataset to train the object detection model and cross-field testing in the field of Taiwan roads. In order to make the model more robust and stable, the data augmentation method is employed to generate more data. Therefore, the data are increased through the data augmentation method and homography transformation of images in the limited dataset. Additionally, Inverse Perspective Mapping is performed on the input images to transform them into the bird’s eye view, which solves the “small objects at far distance” problem and the “perspective distortion of objects” problem so that the model can clearly recognize the objects on the road. The model testing on the front-view images and bird’s eye view images also shows a remarkable improvement of accuracy by 18.62%.

With the rise in the Internet of Things (IOT), mobile devices and Location-Based Social Network (LBSN), abundant trajectory data have made research on location prediction more popular. The check-in data shared through LBSN hide information related to life patterns, and obtaining this information is helpful for location prediction. However, the trajectory data recorded by mobile devices are different from check-in data that have semantic information. In order to obtain the user’s semantic, relevant studies match the stay point to the nearest Point of Interest (POI), but location error may lead to wrong semantic matching. Therefore, we propose a Self-Attention model for next location prediction based on semantic mining to predict the next location. When calculating the semantic feature of a stay point, the first step is to search for the k-nearest POI, and then use the reciprocal of the distance from the stay point to the k-nearest POI and the number of categories as weights. Finally, we use the probability to express the semantic without losing other important semantic information. Furthermore, this research, combined with sequential pattern mining, can result in richer semantic features. In order to better perceive the trajectory, temporal features learn the periodicity of time series by the sine function. In terms of location features, we build a directed weighted graph and regard the frequency of users visiting locations as the weight, so the location features are rich in contextual information. We then adopt the Self-Attention model to capture long-term dependencies in long trajectory sequences. Experiments in Geolife show that the semantic matching of this study improved by 45.78% in TOP@1 compared with the closest distance search for POI. Compared with the baseline, the model proposed in this study improved by 2.5% in TOP@1.

With the rise of the Internet of Things (IOT), mobile devices and Location-Based Social Network (LBSN), an abundant amount of trajectory data has made research on location prediction more popular. The check-in data shared through LBSN hides information related to life patterns, obtaining this information is helpful for location prediction. However, the trajectory data recorded by mobile devices is different from check-in data that has semantic information. In order to obtain the user’s semantic, relevant studies match the stay point to the nearest Point of Interest (POI), but location error may lead to wrong semantic matching. Therefore, we propose A Self-Attention for Next Location Prediction based on Semantic Mining to predict the next location. When calculating the semantic feature of a stay point, the first step is to search for the k nearest POI, then use the reciprocal of the distance from the stay point to the k nearest POI and the number of categories as weights. Finally, we use the probability to express semantic without losing other important semantic information. Furthermore, this research combined with sequential pattern mining can bring richer semantic features. In order to better perceive the trajectory, temporal features learn periodicity of time series by sine function. In terms of location features, we build a directed weighted graph and regard the frequency of users visiting locations as the weight, so the location features are rich in contextual information. We then adopt the Self-Attention model to capture long-term dependencies in long trajectory sequences. Experiments in Geolife show that the semantic matching of this study improved by 45.78% in TOP@1 compared with the closest distance search for POI. Compared with the baseline, the model proposed in this study improved by 2.5% in TOP@1.