-
Akira Obuchi, Junko Uchisawa, Shunsuke Suzuki
2023 Volume 54 Issue 2 Pages
218-223
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
Measurements and modeling studies were conducted on the evaporation and decomposition rates of urea adsorbed on a SCR catalyst using a lab-scale atmospheric pressure flow reactor system. For the evaporation behavior, an equilibrium concentration equation for urea in the vapor phase was derived. For the decomposition behavior, a first-order decomposition rate constant equation for the adsorbed urea was obtained. Using these experimental equations, an evaporation/decomposition model for urea based on the BET adsorption equation has been developed.
View full abstract
-
Keisuke Nakamura, Junji Sakai, Hisashi Ozawa, Naoya Ishikawa
2023 Volume 54 Issue 2 Pages
224-229
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
Optimizing of 3D DPF model on active regeneration for helping DPF design process was investigated. DPF inlet temperature and flow velocity distribution of experimental results were entered into 3D DPF model. Optimized 3D DPF model can simulate trend of temperature distribution during active regeneration and amount of remained PM after active regeneration.
View full abstract
-
Masaaki Kato, Tomohiro Nakayama, Hiroya Okada, Yoshirou Shiina
2023 Volume 54 Issue 2 Pages
230-235
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
The diffuse back-illumination extinction imaging method was applied to measure soot formation, accumulation and oxidation processes in a optical access engine with gasoline direct injection system. The relationships between burned gas temperatures which calculated from the twozone model and soot extinction images were verified. The soot formation was affected to the burnt gas temperatures, pressures and exposing timings of rich hydrocarbon. Fuel wetting areas in a combustion chamber have effects not only amounts of fuel films, but also the exposure timings to the hot burned gas. Ignition timings change the burned gas temperature and durations of it's staying a soot generation region. The dilution ratio has an effect to oxidation process of soot particles after nucleation and accumulation processes.
View full abstract
-
- Modeling Oxygen Storage and NOx trap property of LNT -
Teppei Ohori, Hironori Inomata, Yutaka Itagaki, Hisashi Ozawa, Naoya I ...
2023 Volume 54 Issue 2 Pages
236-241
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
Numerous studies have been carried out on the modeling of Lean NOx Trap (LNT) in diesel engine exhaust aftertreatment systems. In these studies, for the purpose of highly accurate simulation of the properties of oxygen storage materials and NOx storage materials that constitute LNT, catalyst models composed of detail kinetics and/or complex overall kinetics have been proposed. In this report, we tried to develop a simple and highly accurate LNT model as a tool for Model-based development (MBD) with the goal of achieving both simulation accuracy and reducing the man-hours and work period required for modeling.
View full abstract
-
Maki Nakamura, Koji Yokota, Kengo Okai, Masakuni Ozawa
2023 Volume 54 Issue 2 Pages
242-247
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
Experiments by powder catalyst are insufficient to evaluate the PM oxidation performance of catalyzed DPFs directly unless experiments of DPFs fabrication, which are prepared and coated with CeO2 catalyst, are performed using real DPF substrate. In addition, the DPF after being deposited by PM, must be submitted for regeneration test at high temperatures. There are many issues, such as experimental costs and catalyst support techniques, if we use novel powder catalysts as actual coat layers on DPF parts. Therefore, few studies have conducted detailed verification and evaluation of the PM oxidation performance of catalyst-loaded DPFs, and their performance is not yet clear. This study aimed to accelerate the development of a novel PM oxidation catalyst by using our short-path technique of the converter model, which can predict the PM purification performance of a DPF coated with a powder catalyst. In this work, by evaluating PM oxidation performance using nano-CeO2 in powder form, a model was developed to convert the catalytic reaction activity of catalyzed DPF from powder data. With this model, one can obtain the fundamental catalytic activity and behaviour when the catalyst is loaded in the DPF, not doing a laboratory-scale fabrication procedure of DPF. Our model will give preliminary knowledge on whether the powder catalyst is useful or not when a novel PM catalyst is reported.
View full abstract
-
Takeshi Yamamoto
2023 Volume 54 Issue 2 Pages
248-252
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
A two-way roller clutch was developed as a connecting element combining the advantages of a frictional clutch that can be connected even when there is a difference in rotational speed, and a dog clutch that does not require external power to maintain the connection during transmission. Equations for the forces acting on the transmission part, and the power required for connection and disconnection, were derived, and a design method based on the specifications was constructed. A cam profile that compensates for geometry changes due to elastic deformation and provides the desired contact force even at high torque was also developed.
View full abstract
-
Toshiteru Nakamura, Ryota Kawamata, Keiichi Betsui, Kazuyoshi Yamazaki ...
2023 Volume 54 Issue 2 Pages
288-293
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
For an advanced driver assistance system (ADAS) that automatically controls and operates the vehicle system in order to realize a safe and secure society in which vehicles and people coexist, we examined a method to realize stereo vision with an inexpensive configuration in a monocular camera. Generally, a monocular camera can construct an ADAS system at low cost, but there is a problem in terms of distance measurement accuracy. On the other hand, a stereo camera that can measure the distance using two monocular cameras with the same mechanism as the human eye has high distance measurement performance, which is advantageous for the sophistication of the ADAS system, but the cost increases. It becomes an issue. Therefore, in this research, we investigated an inexpensive stereo camera and devised a method to realize stereoscopic vision by dividing the field of view of a monocular camera using two reflective mirrors with different postures. We built a prototype and evaluated this method, demonstrating that it can measure a distance of 30 m with an error of 12%.
View full abstract
-
Hozumi Oshika, Shin-ichi Arimoto, Kosho Kawahara, Ki-ichi Furuhashi, Y ...
2023 Volume 54 Issue 2 Pages
294-299
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
We establish computational schemes for shape optimization of structural member in Isogeometric analysis (IGA). The target structure is a dynamic three-point bending analysis of a hat member. In conventional finite element analysis, drastic shape change and arriving at the true optimal shape is extensively difficult. In this study, by searching optimal shape utilizing IGA data characteristics, compatibility with digital data and convergence to the optimum solution is investigated.
View full abstract
-
Hironori Suzuki, Shouma Ohkubo
2023 Volume 54 Issue 2 Pages
300-305
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
This paper aims to develop an online driving abnormality detection system based on a data-driven approach. It assumes that driving abnormalities are reflected in vehicle acceleration and attempts to estimate the parameters of a non-linear intelligent driver model (IDM) by using a dual particle filter (DPF). For normal driving, DPF can accurately reproduce acceleration by estimating only two of the five model parameters and assuming the remaining parameters are constant. However, if there were situations that could not be reproduced accurately with only two parameter estimates, it can be defined that the driving behavior was so unusual and far from normal that the remaining three parameters cannot be treated as constants. Numerical analysis of the data collected in the driving simulator experiment showed that abnormal driving was significantly detected regardless of the hazardous situation and that the detection sensitivity was high and accurate.
View full abstract
-
- Reproduction Method of Real Road Conditions -
Nobunori Okui
2023 Volume 54 Issue 2 Pages
306-311
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
Real driving emission (RDE) testing method is adopted for Light-duty vehicles. In the future, RDE testing method will be also adopted for Heavy-duty vehicles. However, it is difficult for Heavy-duty vehicles to estimate by using RDE testing method. The reason is the Heavy-duty vehicles have a lot of vehicle families and powertrain types. Therefore, we considered the test method of RDE for heavy-duty vehicles by using new approach of Model Based Testing (MBT), which combinates HILS and engine bench system. In this report, in order to construct the MBT method of RDE, we considered the reproduction method of the real road conditions. We could obtain the road data easily and high precision.
View full abstract
-
Toshinori Okajima, Takehiro Yamagishi, Kyohei Yamaguchi, Makoto Nagata ...
2023 Volume 54 Issue 2 Pages
312-318
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
Emission reduction at cold start is a common challenge on gasoline engine vehicles even in recent years with advanced technologies. Focusing on an emission purification system used in series hybrid vehicles, optimal measures that can reduce emissions with less energy were investigated by numerical simulations, where a three-way catalyst purification reaction model that was verified to reproduce experimental results accurately was adopted and functionalized as an electrically heated catalyst.
View full abstract
-
Kenta Tsutsui, Hiroki Murakami, Shigetaka Okano, Masahito Mochizuki
2023 Volume 54 Issue 2 Pages
319-325
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
We constructed a detailed welding phenomenon analysis model and a simplified analysis model using the thermal shrinkage technique capable of predicting welding deformation for thin plate box-shaped lap-welded joints assuming automotive components. In the thermal shrinkage technique, a new method for setting the thermal shrinkage parameter was proposed to improve the accuracy of predicting welding deformation for thin plate structure. The accuracy of each model was confirmed by comparison with the experimental results, and the calculation cost of each model was compared and evaluated.
View full abstract
-
- Design Optimization of Patchwork Blank and Manufacturing Method Exploration-
Dongyong Shi, Kazuya Yasui, Kenichi Watanabe, Liang Chen
2023 Volume 54 Issue 2 Pages
326-332
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
Hot stamping patchwork technology is gathering attention because it can achieve both crashworthiness performance and lightweight by locally increasing strength or thickness. In this study, an optimization method for determining the minimum patchwork blank was developed. Then it was applied to the design of patchwork structures of two hot-stamped parts, a hat-shaped part and a B pillar, and the effects of material strength and thickness on patchwork blank and lightweight were investigated. In addition, taking a hot-stamped hat-shaped part as an example, two manufacturing methods (arc welded overlay patchwork and laser welded patchwork) for realizing the optimized patchwork structure were examined through trial-manufacture, performance evaluation and feature comparison.
View full abstract
-
Kenji Takada, Shinnosuke Nishi, Takashi Matsuno
2023 Volume 54 Issue 2 Pages
333-338
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
In order to reproduce the local deformation of automotive body structures during crash with high accuracy, the use of Isogeometric analysis is an efficient approach. In this paper, crack propagation analyses of high-strength steel were performed using the Isogeometric BEXT solid element composed of third-order basis function and fracture thresholds at high stress triaxiality obtained by small round bar tensile experiment. The results of the BEXT solid element showed the high correlation for crack propagation with the experiment and behavior after fracture initiation.
View full abstract
-
Toshitaka Nakamura, Fuyuki Aida, Shingo Matsuki, Mari Iino, Shinji Has ...
2023 Volume 54 Issue 2 Pages
339-344
Published: 2023
Released on J-STAGE: February 06, 2023
JOURNAL
FREE ACCESS
In this study, we report on a basic study using oil film forming lubricant additives. With the aim of developing drivetrain lubricants having high efficiency for electric vehicles, we confirmed the effects of our newly developed additive and validated the working mechanism. The additive showed the effect of reducing friction in the mixed lubrication and improved efficiency of the reduction gear box and durability of parts. Furthermore, we confirmed that the friction reduction in the mixed lubrication is due to the prevention of metal contact between parts accompanying the improvement of the oil film thickness.
View full abstract
-
- Modeling of Lambda Sensor at Rich Condition -
Kenji Fujii, Teppei Ohori, Hironori Inomata, Daiji Nagaoka, Yutaka Ita ...
2023 Volume 54 Issue 2 Pages
345-350
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
For the purpose of Model-Based Development of Lean NOx Trap (LNT), which is one of the NOx purification devices for diesel exhaust gas, we developed lambda sensor model and LNT model. In this report, we tried the simple model creation method of lambda sensor during LNT regeneration by using oxidized and reducing gas concentration. There are some issues such as to get accurate gas concentration, but simulation results of lambda sensor model showed relatively high accuracy in LNT regeneration. Furthermore, the inverse model that estimates the diesel exhaust gas concentration from the lambda sensor value also has relatively high accuracy.
View full abstract
-
Tsuyoshi Katayama, Takahiko Yoshino
2023 Volume 54 Issue 2 Pages
351-356
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
The eigenvalues when the stiffness value is set to infinity do not coincide the eigenvalues of the four-degree-of-freedom model, in the rear frame flexibility model that has been used conventionally. In this paper, a new modeling is performed that makes it easy to understand the problems of frame model used in Japan. It becomes clear that Sharp's model had no problems, in the course of this research. The new model shows no problems that occurred when the frame stiffness value is set to infinity. The inertial product with 5 degrees of freedom can be calculated with the new model. And using this value solves the problem that has been included by conventional model.
View full abstract
-
Kazuo Higuchi
2023 Volume 54 Issue 2 Pages
357-362
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
Near side crashes make up a large component of fatal crashes. Side airbags protect vehicle occupant in near side crashes. In previous reports, the author analyzed the effect of side airbag and found that the effect of the torso protection side air bag is less than the roll over curtain air bag, and most injured location was the torso, but the actual injury was not analyzed. In this report, the actual injuries in the fatal side crashes are analyzed and found that the most frequent head/neck injury is brain stem injury followed by skull fracture, and the most frequent torso injury is severe rib fracture followed by aorta laceration.
View full abstract
-
Yuki Manabe, Toru Kojima, Tsunetoshi Iwase
2023 Volume 54 Issue 2 Pages
363-369
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
In this report, a scene with a high risk of collision when a lane change is performed is reproduced in a driving simulator. The evasive action of general drivers who encounter the dangerous scene is investigated and evaluated quantitatively. The driver who took better avoidance behavior was assumed to be a C & C driver, and the characteristic driving behavior was summarized.
View full abstract
-
Hideaki Kido, Felipe Gomez Caballero, Takashi Nomura, Naoki Mori
2023 Volume 54 Issue 2 Pages
370-375
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
The understanding of the surrounding environment is a fundamental and important factor in the autonomous driving system and advanced driving assist system. In order to realize these systems, the accurate map information is important to understand external environment of the vehicle as well as the sensors on the vehicle. This paper proposes the method extracting the difference of lanes between a map information which has the roads and lanes location and an aerial image such as a satellite image or an image from an airplane, to ensure the accuracy of the information on the map. The semantic segmentation for extracting road area from an aerial image and the loss optimization of map information for adjusting the location are employed. The result shows the proposed method is effective when the map has lacked information of the lanes.
View full abstract
-
Yoshimitsu Kitada, Tetsuya Nishimoto, Tomokazu Motomura, Yoshiaki Hara ...
2023 Volume 54 Issue 2 Pages
376-381
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
According to the field data in Japan, pelvis and extremities account for more than 50% among those seriously injured in crashes. In this study, based on medical data such as treatment costs and treatment periods, hip joint injuries, resulting in residual disorders were focused. The goal of this study is to analyze the factors that cause hip injuries under various loading conditions using human body models with improved anatomical structures and material properties around the hip joint. As a result, it was suggested that the sitting posture of the occupant and impact direction affect the hip joint injury.
View full abstract
-
Masaya Ando, Akira Saito, Motoki Maekawa, Hideki Kobayashi
2023 Volume 54 Issue 2 Pages
382-389
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
Concept testing is an evaluation method for presenting concepts created at the planning stage of new products and services to assumed customers and grasping their acceptability and improvement points. The concepts usually presented are often a mixture of abstractions and concretes, such as conceptual product concepts (concept sentences) and specific product characteristics, or prototypes. However, survey cooperators have little experience in evaluating products and services at a conceptual level and are often biased toward concrete presentations. In this study, therefore, attention is paid to the point that products and services presented by concept tests are future events, and construal level theory is noticed. The construal level theory is a theory to explain the phenomenon that there is a strong tendency to make abstract and conceptual ideas and judgements when the psychological distance is far away, and that there is a strong tendency to make concrete and context-dependent ideas and judgements when the psychological distance is close. In this study, we prepared a psychological distance scale on the basis of the construal level theory for technology and service in the automobile and examined to use it for verification of the validity and selection of investigation cooperators of the concept test.
View full abstract
-
- Effects of the drivers' knowledge and External HMIs on their Anxiety about Automated Vehicles -
Akira Ohtani, Yoshinori Egami, Azusa Kuriyama, Kenji Sato, Keisuke Ish ...
2023 Volume 54 Issue 2 Pages
390-395
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
This study investigated effects of other drivers' knowledge and external HMIs on their acceptances of the automated vehicles (AVs). A driving simulator experiment was conducted, in which 29 drivers experienced 3 different situations where they need to interact with the AVs on motorway, namely, merging, following and cutin scenes. Drivers were instructed to answer their anxiety when they interacted with the AV in each scene. The results showed that in the following scene, drivers' knowledge and external HMI influenced their anxiety toward the interaction with the AV. Drivers' knowledge of the AVs is important to promote the extensive use of them.
View full abstract
-
Shoichiro Teranishi, Hiroki Kawashima, Genya Abe, Mitsuru Nagatsuka
2023 Volume 54 Issue 2 Pages
396-402
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
In autonomous truck platoon(SAE level 3), the driver is released from driving operations and constant monitoring of systems under certain driving conditions, so it is expected that driver's mental workload will be reduced compared to manual driving. However, it is not fully known whether driver's mental workload is reduced in autonomous truck platoon. We conducted a driving simulator experiment to evaluate changes in driver's mental workload based on pupil diameter. Results showed that driver's mental workload may be reduced by autonomous truck platoon.
View full abstract
-
- Effects of after Automated Driving Guidance on Driver Avoidance of a Parked Vehicle -
Ryo Hasegawa, Joohyeong Lee, Hiroki Nakamura, Genya Abe, Nobuyuki Uchi ...
2023 Volume 54 Issue 2 Pages
403-409
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
During level 3 automated driving on urban roads, the occurrence of take-over caused by urban domain-specific factors is possible. In this research, we analyzed driver behavior by conducting a take-over experiment. The experimental scenario consisted of the driver passing through an intersection and having to avoid a parked car after take-over. Results show that suggesting to the driver to operate the pedal and steering wheel, in addition to issuing the request to intervene, may facilitate a safe and smooth take-over.
View full abstract
-
- Effect of Information Content Including Uncertainty on Whether a Driver can Passing a Signalized Intersection before Yellow Light-
Ryo Yamaguchi, Tatsuru Daimon, Koichiro Iwaoka, Susumu Asada, Toshihik ...
2023 Volume 54 Issue 2 Pages
410-416
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
This paper examined the information content of signal passing support to improve driver behavior, especially in the dilemma zone when approaching or entering a signalized intersection. The experiments were set up to examine information content when the signal passage support was not used, when it was used, and in addition when it was uncertain whether a driver could pass the intersection before the yellow light. The data obtained in the experiment were analyzed in terms of safety and psychological aspects. The results of these experiments were reported
View full abstract
-
Tomokatsu Okuya, Nobuhiro Mizuno, Akira Yoshizawa
2023 Volume 54 Issue 2 Pages
417-422
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
Veteran drivers can develop bad driving habits. Some drivers don’t stop or check left and right at intersections where a stop is required. We’re developing a system that corrects such dangerous habits and encourages drivers to behave safely. In this study, we built a system that provides an audio alert after the driver has crossed an intersection. The system rewards drivers for good driving behavior and reprimands them for bad driving behavior. Experiments showed that the driver's driving behavior was modified so that he continued to drive safely even when the system was used only once a week.
View full abstract
-
Yuya Homma, Takeshi Kohama
2023 Volume 54 Issue 2 Pages
423-429
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
In order to predict where drivers direct their attention and gaze while driving, the previous study has reported a saliency estimation model for self-motion images that consider receptive field properties in the early visual system and higher-order motion selection mechanisms. However, this model needs to improve its prediction accuracy at low driving speeds and the responsiveness of the receptive field model to extract features in dark scenes. In this study, we added a process that dynamically changes the frame interval when calculating the motion vector to construct a mathematical model that can more accurately predict the driver’s gaze position. We also modified the response characteristics of the on-center and off-center receptive field models. The simulation results indicate the effectiveness of the proposed model by showing that the moving normalized scan-path saliency (M-NSS) score, which indicates the accuracy of gaze prediction, increased at intersections and narrow roads.
View full abstract
-
- In the Case of Small Overlap Rate -
Takashi Suzuki, Kazunori Kikuchi, Takashi Wakasugi, Masaaki Senga, Hir ...
2023 Volume 54 Issue 2 Pages
430-435
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
With the social demand to reduce traffic accidents, to understand driving behavior in traffic accidents is important not only for the development of automobiles but also for discussion on the standardization of ADAS (Advanced Driver-Assistance Systems). In the previous report, we conducted an experiment using a driving simulator and investigated the characteristics of avoidance operations in emergencies at a large overlap rate for general drivers. We hypothesized that different overlap rates would be in different driver avoidance operations. In this study, we investigated the characteristics of avoidance operations in emergencies at a small overlap rate for general drivers. As a result, it was found that most drivers avoid collisions with steering in small overlap situations compared to large overlap situations. Moreover, we found that some drivers may be able to be assisted in avoidance by ESF (Emergency Steering Function).
View full abstract
-
- HMI Specification to Maintain Driver’s Attention Ahead by Anticipatory Information from Road Traffic Information Systems -
Taiga Mizuta, Koshiro Sumi, Daimon Daimon, Nobuhiro Mizuno, Hirofumi D ...
2023 Volume 54 Issue 2 Pages
436-442
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
In this study, we focus on maintaining driver’s situation awareness of the surroundings in performing a non-driving related task (NDRT) such as watching videos during automated driving and moreover extracting the design factors of human machine interface (HMI) assist smooth takeover performance from automated driving to manual driving when providing a request to intervene (RtI). We designed two types of HMI in this experiment, one type was to display the text message of anticipatory information on the upper edge area of the display for NDRT after providing anticipatory information and another type was to display traffic conditions ahead as a video image after providing anticipatory information. As a result, it was shown that the HMI with text message on the upper edge of display made drivers aware of the possibility of RtI occurrence during automated driving.In addition, as an element of the HMI to support the driver's taking over during NDRT execution during automatic driving, it was considered that, in order to maintain awareness or attention to the traffic situation even during NDRT execution, the HMI should not interfere too much with the attention resources required for NDRT execution, and by using the display on the HMI, the driver can substitute his or her attention to the traffic situation without relying on memory.
View full abstract
-
Koyo Horie, Yudai Yamasaki
2023 Volume 54 Issue 2 Pages
443-448
Published: 2023
Released on J-STAGE: March 25, 2023
JOURNAL
FREE ACCESS
The mechanism of carbon dioxide emissions from real driving, which remains unclear, was examined through the analysis of real driving data of a hybrid vehicle. A predictive model for instantaneous carbon dioxide emissions was developed, utilizing a gradient boosting machine learning library called LightGBM (Light Gradient Boosting Machine). The novel Shapley additive explanation (SHAP) measures were employed to reveal the importance of various features. We tracked over time the features that turned out to be highly important.
View full abstract