Fast, accurate spatial-temporal cognition is vital for situational awareness and mission success in indoor emergencies. Complex layouts and occlusions hinder first responders' ability to judge distances, scales, and task completion times. This paper explores the enhancement of these cognitive abilities through the application of Augmented Reality (AR) devices. Utilizing mixed reality and reality capturing technologies, a digital twin model of a three-level indoor structure was built based on LiDAR scans of Turlington Hall at the University of Florida. A mixed reality environment was then constructed using the Unity engine and Microsoft HoloLens2. In a user study (N = 30), we systematically compared three visualization conditions (2D Map, 3D Map, and Egocentric Map) across spatial and temporal estimation tasks. Non-parametric statistical analysis with correction for multiple comparisons revealed a task-dependent pattern. The Egocentric Map showed the lowest error for ETA estimation (MAPE = 17.2% vs. 23.3% for 2D maps), suggesting a trend toward improved temporal cognition, though this pairwise difference did not reach statistical significance after correction. Conversely, for distance-between-objects tasks, 2D and 3D maps yielded significantly lower error than Egocentric views (MAPE = 24.3% and 17.7% vs. 36.8%, $p_{adj} < 0.01$padj<0.01). These findings reveal a task-dependent trade-off, indicating that no single visualization mode is universally optimal: AR overlays may support action-oriented temporal planning, while exocentric views remain superior for precise metric spatial judgments in emergency response.

Traffic congestion remains a significant challenge in modern urban networks. Autonomous driving technologies have emerged as a potential solution. Among traffic control methods, reinforcement learning has shown superior performance over traffic signals in various scenarios. However, prior research has largely focused on small-scale networks or isolated intersections, leaving large-scale mixed traffic control largely unexplored. This study presents the first attempt to use decentralized multi-agent reinforcement learning for large-scale mixed traffic control in which some intersections are managed by traffic signals and others by robot vehicles. Evaluating a real-world network in Colorado Springs, CO, USA with 14 intersections, we measure traffic efficiency via average waiting time of vehicles at intersections and the number of vehicles reaching their destinations within a time window (i.e., throughput). At 80% RV penetration rate, our method reduces waiting time from 6.17 s to 5.09 s and increases throughput from 454 vehicles per 500 seconds to 493 vehicles per 500 seconds, outperforming the baseline of fully signalized intersections. These findings suggest that integrating reinforcement learning-based control large-scale traffic can improve overall efficiency and may inform future urban planning strategies.

We present an approach to automatically discovering explicit rules for software process evaluation from evaluation histories. Each rule is a conjunction of a subset of attributes in a process execution, characterizing why the execution is normal or anomalous. The discovered rules can be used for stakeholder as expertise to avoid mistakes in the future, thus improving software process quality; it can also be used to compose a classifier to automatically evaluate future process execution. We formulate this problem as a contrasting itemset mining task, and employ the branch-and-bound technique to speed up mining by pruning search space. We have applied the proposed approach to four real industrial projects in a commercial bank. Our empirical studies show that the discovered rules can precisely pinpoint the cause of all anomalous executions, and the classifier built on the rules is able to accurately classify unknown process executions into the normal or anomalous class.

Software maintenance effort constitutes a major portion of the software lifecycle effort. Its estimation is vital for successful project planning and strategic resource allocation. In this paper, we conduct and report an industrial case study in this field. The data set was collected from an industrial software process management tool QONE (formerly SoftPM). The methodology proposed provides corresponding guidance for effort estimation in software evolutionary projects that employ use-cases in capturing maintenance requirements. And the model, constructed using the linear regression analysis and validated by the leave-one-out cross-validation, provides an effort prediction for the future maintenance of the project. The analysis results indicate that the methodology can be applied at an early stage of the project life cycle and provides a good tradeoff among simplicity, early-estimating and accuracy in one estimate.

Nowadays, commercial or in-house customized process management tools have been prevalently adopted for supporting software project management and process improvement. In this paper we report a case study to empirically investigate and evaluate the usage status and implications of a supporting tool named QONE in industrial environment in China. Decision theory is adopted in study design. The analysis is mainly based on the usage data from a typical industrial project. Further questionnaires and follow-up interviews with the end-users are also conducted. The analysis results reveal that 1) the effects of such supporting tool vary with respect to different task types, 2) tasks with smaller granularity are comparatively easier to predict and control, 3) missing data reporting analysis helps to reveal opportunities for further process improvement and tool enhancement. This investigation aims to help us take advantages of such supporting tools and benefit software development eventually.

Although a lot of attention has been paid to software cost estimation since 1960, making accurate effort and schedule estimation is still a challenge. To collect evidence and identify potential areas of improvement in software cost estimation, it is important to investigate the estimation accuracy, the estimation method used, and the factors influencing the adoption of estimation methods in current industry. This paper analyzed 112 projects from the Chinese software project benchmarking dataset and conducted questionnaire survey on 116 organizations to investigate the above information. The paper presents the current situations related to software project estimation in China and provides evidence-based suggestions on how to improve software project estimation. Our survey results suggest, e.g., that large projects were more prone to cost and schedule overruns, that most computing managers and professionals were neither satisfied nor dissatisfied with the project estimation, that very few organizations (15%) used model-based methods, and that the high adoption cost and insignificant benefit after adoption were the main causes for low use of model-based methods.