For further improvement of thermal efficiency in super-lean burn spark ignition (SI) engine by reducing unburned loss, it is important to clarify unburnt hydrocarbon emission mechanism. In this study, time-resolved total hydrocarbon (THC) mole fraction measurement by a fast response flame ionization detector (FFID) was conducted in super-lean burn single cylinder SI engine with port fuel injection, Gas sampling probes were mounted at about 47 mm (upstream measurement position) and 209 mm (downstream measurement position) downstream from the exhaust valve. This study investigated the relationship between THC emissions and combustion characteristics in each cycle of a super-lean burn spark ignition engine, in order to clarify the mechanism of unburned hydrocarbon (UHC) formation and emissions. The THC in the cylinder were measured by placing the tip of the FFID probe at two locations with the distances from wall surface of 0 mm and 5 mm. The results showed that the THC mole fraction at 5 mm from the wall was strongly affected by CA90, while the effect of CA90 on THC was weak at 0mm from the wall. Furthermore, the measured THC in exhaust pipe only in stable combustion cycles were extracted and related with the operation conditions of the engine. Even by excluding the measured data of misfire cycles, the results showed THC emissions in each cycle increase with the increase in excess air ratio between 1.0 and 2.0.

抄録全体を表示

The progressive depletion of petroleum-based energy reserves, coupled with the intensifying threat of global climate change, has catalyzed an urgent global imperative to explore alternative energy sources to conventional fossil fuels. Although electric vehicles offer a promising solution by eliminating tailpipe emissions, their widespread adoption is hindered by range anxiety and insufficient charging infrastructure. In contrast, Hybrid Electric Vehicles can serve as a practical transitional technology, offering improved fuel economy with reduced emissions by operating both the IC engine and electric motor in their best efficiency regions. The process of developing HEVs involves various modeling techniques, including Model-in-Loop (MiL), Software-in-Loop (SiL), and Hardware-in-Loop (HiL). In MiL, kinematic, quasi-static, and dynamic modeling techniques are employed to develop accurate digital twins and predict vehicle behavior across various driving conditions. This study introduces a novel hybrid simulation framework that integrates forward and backward simulation techniques to develop a full-parallel P3-type hybrid electric vehicle architecture. The input data for the developed model are generated from actual power plant testing to improve the system's accuracy. The proposed framework facilitates a comprehensive evaluation of energy management strategies and fuel consumption metrics. A deterministic, rule-based control algorithm incorporating PID tuning and engine start-stop functionality is developed to optimize power distribution between the internal combustion engine and the electric motor. Simulation results demonstrate a fuel-efficiency improvement of approximately 30% relative to a conventional baseline vehicle, with strong correlation with experimental data, with deviations of 4–6%.

抄録全体を表示

This study proposes a novel scheme for the simultaneous optimization of degradation parameters and the selection of degradation scenarios for Li-ion battery cells based on capacity retention data. The proposed method identifies the most probable degradation mechanisms through posterior probability. Validation using synthetic data confirmed the scheme's ability to accurately select intended models, such as thin film growth in anode and structural transitions in cathode. Furthermore, the scheme demonstrated robust extrapolation performance, predicting future capacity fade. This framework provides a powerful tool for identifying battery health states and predicting long-term reliability in practical automotive applications.

抄録全体を表示

Personal mobility supports social participation and well-being, yet highway merging remains stressful and risky. Driving behavior is influenced by prediction errors between actual traffic and drivers' predictions. We hypothesized that designing traffic environments to improve driver prediction accuracy can effectively assist drivers. We previously proposed position-triggered speed management using Cooperative Adaptive Cruise Control (CACC) vehicles near highway merging sections to smooth merging by reducing speed differentials between merging and mainline vehicles. However, empirical validation of speed differentials' impact on drivers' merging experience was lacking due to no methods for evaluating human interactions in mixed traffic involving human-driven vehicles (HDVs) and CACC vehicles. To address this, we developed a networked multi-driver driving simulator (NMDDS) allowing ten participants to interact in a virtual merging environment, essential for studying complex driver dynamics in real traffic. Experimental results showed CACC speed management effectively moderated mainline HDV speeds near merging sections, with effectiveness depending on CACC penetration rate. We clarified how personality traits relate to subjective evaluations like perceived danger and merging difficulty. Structural Equation Modeling (SEM) also indicated relative speed between merging and mainline vehicles significantly influences these factors. These findings empirically support that reducing speed differentials improves merging comfort and safety. Future research will implement methods to further minimize speed differentials, including managing merging vehicle speeds and refining adaptive speed control strategies, aiming to enhance safety and driver intuitiveness during highway merging.

抄録全体を表示

In Japan, the prevention of traffic accidents on community roads, which frequently have poor visibility at intersections, is an ongoing issue. As a solution to this issue, we focus on Vehicle-to-Network-to-Vehicle (V2N2V) safety systems, which use cellular networks and allow vehicles to share information over a wider area. In V2N2V safety systems, vehicles provide their self-localization data and local observations of surrounding traffic participants to a central server, and then the system predicts potential collisions for other vehicles based on the provided information. However, in the coordinate transformation of information from vehicles, the system must consider two sources of uncertainty: (1) self-localization uncertainty of the vehicle and (2) local observation uncertainty of the surrounding participants. Based on this motivation, we propose a framework that integrates these two uncertainties in coordinate transformation and then estimates the global state of the traffic participant observed by the vehicle. We validate this framework through both simulation and real-world experiments under conditions that assume on-board sensors of widespread commercial vehicles. In the simulation, we compare the proposed framework with a baseline that does not consider self-localization uncertainty. The results demonstrate that incorporating self-localization uncertainty is crucial for appropriately estimating the overall uncertainty after the coordinate transformation under the assumed on-board sensor setup. In addition, results of real-world experiments demonstrate that the framework can integrate states with appropriate uncertainty. The proposed framework will enable a broader range of vehicles to provide information in V2N2V-based safety systems and contribute to more comprehensive road safety.

抄録全体を表示

This study focused on the driver's brain activity in the prediction phase, which is the stage before recognition, and analyzed the effects of superior or inferior brain activity on recognition and driving performance on curved roads through a driving simulator experiment. A circuit course was set up, and changes in cerebral blood flow and gaze distribution were analyzed across 10 corners, together with the average of lane deviation in the first third of each curve. The results showed that activation of the frontal lobe during the prediction phase led to faster curve recognition and reduced lane departure in the curve sections. Overall, it was found that activation of the frontal lobe in the prediction phase improves both curve recognition and driving performance on curved roads.

抄録全体を表示

One of the longest and most complex quality control procedures during high-voltage battery production is the leak test that must be performed on all batteries. The test consists of applying a controlled pressure variation inside the battery. To improve this process, a CFD model was developed, and the different phases of the test were numerically reproduced and then compared with experimental data. The numerical model was used in a simplified setup for validation and then implemented to the battery. The flow inside the battery was studied to identify impacting factors like geometry to increase the test accuracy in future work.

抄録全体を表示

The integration of Advanced Driver Assistance Systems (ADAS) within modular robotic frameworks presents significant opportunities for scalable autonomous mobility research. This paper introduces a comprehensive ROS 2 Humble-based simulation framework for implementing and evaluating core ADAS functionalities-Lane Keeping Assist (LKA), Adaptive Cruise Control (ACC), Forward Collision Warning (FCW), and Automatic Emergency Braking (AEB)-on a differential-drive robotic platform modeled in Gazebo. Our architecture uniquely combines multi-sensor perception, a dual-SLAM strategy (2D mapping via SLAM Toolbox and dense 3D reconstruction/localization via RTAB-Map), and real-time navigation control using Nav2 within a unified, extensible ROS 2 node graph. The system integrates (i) a URDF/Xacro vehicle equipped with LiDAR, RGB and depth cameras for multimodal perception; (ii) modular AI perception pipelines for lane detection, object recognition and obstacle classification; (iii) control-theoretic ADAS implementations (PID and MPC) for lateral stabilization and adaptive longitudinal regulation; (iv) a Time-to-Collision (TTC) safety module enabling predictive FCW/AEB; and (v) a hierarchical arbitration mechanism that dynamically interfaces with and-under safety-critical conditions-overrides the navigation stack. The architecture is designed with clearly defined topic interfaces, deterministic message flows, and fail-safe prioritization logic to meet real-time constraints; simulation experiments in the Gazebo environment demonstrate robust behavior across dynamic obstacle scenarios, sensor noise profiles and timing stresses. Finally, a computational feasibility analysis on the Raspberry Pi 4 highlights trade-offs between modular and monolithic architectural strategies in latency, scalability and maintainability.

抄録全体を表示