Recently, three-dimensional CAD systems based on feature-based solid modeling techniques have been widely used for product design. However, when part models associated with features are used in various downstream applications, simplified models at various levels of detail (LODs) are frequently more desirable than the full details of the parts. One challenge is to generate valid models at various LODs after an arbitrary rearrangement of features using a certain LOD criterion, because composite Boolean operations consisting of union and subtraction are not commutative. This article proposes an algorithm for feature-based multiresolution solid modeling based on the effective volumes of features. This algorithm guarantees the same resulting shape and the reasonable intermediate LOD models for an arbitrary rearrangement of the features, regardless of whether feature types are additive or subtractive. This characteristic enables various LOD criteria to be used for a wide range of applications including computer-aided design and analysis.

Vehicle re-identification aims to obtain the same vehicles from vehicle images. It is challenging but essential for analyzing and predicting traffic flow in the city. Although deep learning methods have achieved enormous progress in this task, requiring a large amount of data is a critical shortcoming. To tackle this problem, we propose a novel framework called Synthetic-to-Real Domain Adaptation Network (StRDAN), which is trained with inexpensive large-scale synthetic data as well as real data to improve performance. The training method for StRDAN is combined with domain adaptation and semi-supervised learning methods and their associated losses. StRDAN shows a significant improvement over the baseline model, which is trained using only real data, in two main datasets: VeRi and CityFlow-ReID. Evaluating with the mean average precision (mAP) metric, our model outperforms the reference model by 12.87% in CityFlow-ReID and 3.1% in VeRi.

A majority of recently developed advanced vehicles have been equipped with various automated driver assistance systems, such as adaptive cruise control (ACC) and lane keeping assistance systems. ACC systems have several operational modes, and drivers can be unaware of the mode in which they are operating. Because mode confusion is a significant human error factor that contributes to traffic accidents, it is necessary to develop user interfaces for ACC systems that can reduce mode confusion. To meet this requirement, this paper presents a new human-automation interaction design methodology in which the compatibility of the machine and interface models is determined using the proposed criteria, and if the models are incompatible, one or both of the models is/are modified to make them compatible. To investigate the effectiveness of our methodology, we designed two new interfaces by separately modifying the machine model and the interface model and then performed driver-in-the-loop experiments. The results showed that modifying the machine model provides a more compact, acceptable, effective, and safe interface than modifying the interface model.