As the demand for traffic simulation has shifted to large-scale urban areas, achieving both high accuracy and computational efficiency has become increasingly challenging. Among various traffic simulation levels, microscopic simulation provides the most detailed and accurate representation of traffic dynamics; however, its high computational cost prevents its applicability to large-scale urban networks. Consequently, the need to find a compromise between accuracy and computational cost leads to hybrid or multi-level traffic simulation. Multi-level traffic simulation integrates multiple simulation levels within a single framework, where micro-meso hybrids are generally suitable for urban networks and micro-macro hybrids for highways. To maintain constant simulation performance in terms of accuracy and computational cost, recent studies have emphasized the importance of dynamic properties that adjust the simulation levels of road segments in response to the traffic conditions over time. Accordingly, this study proposes a dynamic multi-level traffic simulation for urban networks by combining microscopic and mesoscopic traffic simulations. A unified simulation framework and data structure were proposed to ensure compatibility and consistency between different simulation levels. Temporal and spatial interfaces were designed and verified for proper functioning, including the preservation of vehicle information and consistency in traffic dynamics. The proposed simulation was evaluated in terms of computational cost and accuracy under various demand scenarios, using multiple methods for dynamically determining micro- and meso-level representations. The dynamic multi-level simulation consistently achieved higher accuracy than the mesoscopic model with reduced computational cost than the microscopic model across all demand scenarios. In addition, the proposed approach showed higher reliability than the fixed multi-level simulations, especially under an ROI-unfocused demand patterns. These outcomes were consistently observed across networks of varying sizes. Finally, an application to a real-world urban network demonstrated the potential of the proposed model for practical use in traffic management systems.

Open source is a driving force behind scientific advancement. However, this openness is also a double-edged sword, with the inherent risk that innovative technologies can be misused for purposes harmful to society. What is the likelihood that an open source AI model or dataset will be used to commit a real-world crime, and if a criminal does exploit it, will the people behind the technology be able to escape legal liability? To address these questions, we explore a legal domain where individual choices can have a significant impact on society. Specifically, we build the EVE-v1 dataset that comprises 200 question-answer pairs related to criminal offenses based on 200 Korean precedents first to explore the possibility of malicious models emerging. We further developed EVE-v2 using 600 fraud-related precedents to confirm the existence of malicious models that can provide harmful advice on a wide range of criminal topics to test the domain generalization ability. Remarkably, widely used open-source large-scale language models (LLMs) provide unethical and detailed information about criminal activities when fine-tuned with \oursall. We also take an in-depth look at the legal issues that malicious language models and their builders could realistically face. Our findings highlight the paradoxical dilemma that open source accelerates scientific progress, but requires great care to minimize the potential for misuse.

Depth information which specifies the distance between objects and current position of the robot is essential for many robot tasks such as navigation. Recently, researchers have proposed depth completion frameworks to provide dense depth maps that offer comprehensive information about the surrounding environment. However, existing methods show significant trade-offs between computational efficiency and accuracy during inference. The substantial memory and computational requirements make them unsuitable for real-time applications, highlighting the need to improve the completeness and accuracy of depth information while improving processing speed to enhance robot performance in various tasks. To address these challenges, in this paper, we propose CHADET (cross-hierarchical-attention depth-completion transformer), a lightweight depth-completion network that can generate accurate dense depth maps from RGB images and sparse depth points. For each pair, its feature is extracted from the depthwise blocks and passed to the equally lightweight transformer-based decoder. In the decoder, we utilize the novel cross-hierarchical-attention module that refines the image features from the depth information. Our approach improves the quality and reduces memory usage of the depth map prediction, as validated in both KITTI, NYUv2, and VOID datasets.

Deep neural network-based image signal processing (ISP-DNN) improves image quality with techniques such as demosaicing, but these models pose substantial computational and memory challenges when implemented on CMOS image sensors, particularly due to the high-resolution inputs that increase memory requirements for activations. Layer fusion reduces memory usage by combining consecutive processing steps, yet it increases computational demands, a critical issue in resource-limited on-sensor environments. To address these challenges, we introduce ISP2DLA, an automated deep learning accelerator design framework that balances computational and memory demands for on-sensor ISP. This framework optimizes hardware designs by adjusting line buffer sizes and the number of MAC units, reducing gate counts by 14-79% across two ISP-DNN models, thus enabling efficient on-sensor ISP model inference within constrained resources.

Contrary to other standard cameras, event cam-eras interpret the world in an entirely different manner; as a collection of asynchronous events. Despite event camera's unique data output, many event feature detection and tracking algorithms have shown significant progress by making detours to frame-based data representations. This paper questions the need to do so and proposes a novel event data-friendly method that achieve simultaneous feature detection and tracking, called event Clustering-based Detection and Tracking (eCDT). Our method employs a novel clustering method, named as k-NN Classifier-based Spatial Clustering and Applications with Noise (KCSCAN), to cluster adjacent polarity events to retrieve event trajectories. With the aid of a Head and Tail Descriptor Matching process, event clusters that reappear in a different polarity are continually tracked, elongating the feature tracks. Thanks to our clustering approach in spatio-temporal space, our method automatically solves feature detection and feature tracking simultaneously. Also, eCDT can extract feature tracks at any frequency with an adjustable time window, which does not corrupt the high temporal resolution of the original event data. Our method achieves 30 % better feature tracking ages compared with the state-of-the-art approach while also having a low error approximately equal to it.

In this paper, a degeneracy avoidance method for a point and line based visual SLAM algorithm is proposed. Visual SLAM predominantly uses point features. However, point features lack robustness in low texture and illuminance variant environments. Therefore, line features are used to compensate the weaknesses of point features. In addition, point features are poor in representing discernable features for the naked eye, meaning mapped point features cannot be recognized. To overcome the limitations above, line features were actively employed in previous studies. However, since degeneracy arises in the process of using line features, this paper attempts to solve this problem. First, a simple method to identify degenerate lines is presented. In addition, a novel structural constraint is proposed to avoid the degeneracy problem. At last, a point and line based monocular SLAM system using a robust optical-flow based lien tracking method is implemented. The results are verified using experiments with the EuRoC dataset and compared with other state-of-the-art algorithms. It is proven that our method yields more accurate localization as well as mapping results.

Numerous car-following models have been developed since the 1950s. However, there still exist many traffic phenomena that cannot be demonstrated using the existing models. Therefore, this research proposed a new car-following model, the Asymmetric car-following (ACF) model based on the understanding of driver's asymmetric behavior, which can explain complex traffic phenomena. We established the asymmetric car-following (ACF) rule under the vehicle's safety constraints using eight parameters that indicate the driver's characteristics and vehicle's performances. To evaluate the ACF model, we performed the simulation for car-following pairs and conducted a comparison analysis with the existing models: Newell, Gipps, GM, and IDM. As a result, the proposed ACF model showed good fitness with the empirical trajectory and the apparent asymmetric behavior compared to others. For further investigation in the congested traffic stream, we simulated a group of vehicles by adding an error term to represent the driver's unexpected behavior. The simulation showed growth, propagation, and dissipation of the stop-and-go traffic. These results proved that the ACF model has the strength to elaborate on various traffic phenomena, such as traffic hysteresis and stop-and-go traffic.

It is believed that life passed through an RNA World stage in which replication was sustained by catalytic RNAs (ribozymes). The two most obvious types of ribozymes are a polymerase, which uses a neighbouring strand as a template to make a complementary sequence to the template, and a nucleotide synthetase, which synthesizes monomers for use by the polymerase. When a chemical source of monomers is available, the polymerase can survive on its own. When the chemical supply of monomers is too low, nucleotide production by the synthetase is essential and the two ribozymes can only survive when they are together. Here we consider a computational model to investigate conditions under which coexistence and cooperation of these two types of ribozymes is possible. The model considers six types of strands: the two functional sequences, the complementary strands to these sequences (which are required as templates), and non-functional mutants of the two sequences (which act as parasites). Strands are distributed on a two-dimensional lattice. Polymerases replicate strands on neighbouring sites and synthetases produce monomers that diffuse in the local neighbourhood. We show that coexistence of unlinked polymerases and synthetases is possible in this spatial model under conditions in which neither sequence could survive alone; hence, there is a selective force for increasing complexity. Coexistence is dependent on the relative lengths of the two functional strands, the strand diffusion rate, the monomer diffusion rate, and the rate of deleterious mutations. The sensitivity of this two-ribozyme system suggests that evolution of a system of many types of ribozymes would be difficult in a purely spatial model with unlinked genes. We therefore speculate that linkage of genes onto mini-chromosomes and encapsulation of strands in protocells would have been important fairly early in the history of life as a means of enabling more complex systems to evolve.