Showing papers by "Toyota published in 2022"

Structured Hammerstein-Wiener Model Learning for Model Predictive Control

Abstract: This letter aims to improve the reliability of optimal control using models constructed by machine learning methods. Optimal control problems based on such models are generally non-convex and difficult to solve online. In this letter, we propose a model that combines the Hammerstein-Wiener model with input convex neural networks, which have recently been proposed in the field of machine learning. An important feature of the proposed model is that resulting optimal control problems are effectively solvable exploiting their convexity and partial linearity while retaining flexible modeling ability. The practical usefulness of the method is examined through its application to the modeling and control of an engine airpath system.

Large-Scale Three-Dimensional Anisotropic Topology Optimization of Variable-Axial Lightweight Composite Structures

Abstract: Variable-axial fiber-reinforced composites allow for local customization of fiber orientation and thicknesses. Despite their significant potential for performance improvement over the conventional multiaxial composites and metals, they pose challenges in design optimization due to the vastly increased design freedom in material orientations. This paper presents an anisotropic topology optimization method for designing large-scale, 3D variable-axial lightweight composite structures subject to multiple load cases. The computational challenges associated with large-scale 3D anisotropic topology optimization with extremely low volume fraction are addressed by a tensor-based representation of 3D orientation that would avoid the 2π periodicity of angular representations such as Euler angles, and an adaptive meshing scheme, which, in conjunction with PDE regularization of the density variables, refines the mesh where structural members appear and coarsens where there is void. The proposed method is applied to designing a heavy-duty drone frame subject to complex multi-loading conditions. Finally, the manufacturability gaps between the optimized design and the fabrication-ready design for tailored fiber placement (TFP) is discussed, which motivates future work toward a fully automated design synthesis.

Benefits of Feedforward for Model Predictive Airpath Control of Diesel Engines

Abstract: This paper investigates options to complement a diesel engine airpath feedback controller with a feedforward. The control objective is to track the intake manifold pressure and exhaust gas recirculation (EGR) rate targets by manipulating the EGR valve and variable geometry turbine (VGT) while satisfying state and input constraints. The feedback controller is based on rate-based Model Predictive Control (MPC) that provides integral action for tracking. Two options for the feedforward are considered one based on a look-up table that specifies the feedforward as a function of engine speed and fuel injection rate, and another one based on a (non-rate-based) MPC that generates dynamic feedforward trajectories. The controllers are designed and verified using a high-fidelity engine model in GT-Power and exploit a low-order rate-based linear parameter-varying (LPV) model for prediction which is identified from transient response data generated by the GT-Power model. It is shown that the combination of feedforward and feedback MPC has the potential to improve the performance and robustness of the control design. In particular, the feedback MPC without feedforward can lose stability at low engine speeds, while MPC-based feedforward results in the best transient response. Mechanisms by which feedforward is able to assist in stabilization and improve performance are discussed.

Peimine ameliorates pulmonary fibrosis via the inhibition of M2-type macrophage polarization through the suppression of P38/Akt/STAT6 signals

Abstract: Peimine, a bioactive substance isolated from Chinese medicine Fritillaria, can potentially suppress pulmonary fibrosis (PF); however, its therapeutic mechanism remains unclear. Recent evidence suggests the participation of M2-type macrophages in the pathogenesis of PF. The present study aimed to investigate the effect of peimine on a bleomycin (BLM)-induced PF rat model and the underlying mechanism of this effect. After BLM administration, peimine was administered to rats from day 29 to day 42, with pirfenidone (PFD) as a positive control. H&E and Masson's trichrome stain were used to analyze histological changes. Q-PCR and western blotting were used to measure mRNA levels and protein levels, respectively. High-throughput RNA sequencing (RNA-seq) technology detected the differentially expressed genes (DEGs) regulated by peimine. Our results revealed that peimine treatment significantly ameliorated BLM-induced PF by suppressing histological changes and collagen deposition. In addition, peimine decreased the number of M2 macrophages and the expression of profibrotic factors. RNA-seq results showed that DEGs regulated by peimine in IL-4-induced macrophages were mainly associated with immune system processes, the PI3K/Akt pathway, and the MAPKs pathway. Then, immunofluorescence assay and western blot results demonstrated that peimine treatment suppressed the expression of p-p38 MAPK and p-Akt (s473) and also inhibited the nuclear translocation of p-STAT6. In conclusion, the present study demonstrated that peimine has a protective effect on PF through the suppression of M2 polarization of macrophages by inhibiting the STAT6, p38 MAPK, and Akt signals.

Feature Selection Methods for Uplift Modeling and Heterogeneous Treatment Effect

Abstract: Uplift modeling is a causal learning technique that estimates subgroup-level treatment effects. It is commonly used in industry and elsewhere for tasks such as targeting ads. In a typical setting, uplift models can take thousands of features as inputs, which is costly and results in problems such as overfitting and poor model interpretability. Consequently, there is a need to select a subset of the most important features for modeling. However, traditional methods for doing feature selection are not fit for the task because they are designed for standard machine learning models whose target is importantly different from uplift models. To address this, this paper introduces a set of feature selection methods explicitly designed for uplift modeling, drawing inspiration from statistics and information theory. Empirical evaluations are conducted on the proposed methods on publicly available datasets, demonstrating the advantages of the proposed methods compared to traditional feature selection. We make the proposed methods publicly available as a part of the CausalML open-source package.

Periodic switching of acoustic radiation force with beat created by multitone field

Abstract: Acoustic radiation force plays a key role in microfluidic systems for particle and cell manipulation. In this study, we investigate the acoustic radiation force resulting from synthesized ultrasounds that are emitted from multiple sound sources with slightly different oscillation frequencies. Due to the synthesized field, the acoustic radiation force is expressed as the sum of a dc component and harmonics of fundamental frequencies of a few hertz. This induces the beat of the acoustic radiation force. We demonstrate that the synthesized field provides the periodic on/off switching of the acoustic radiation force associated with the one denominational planar standing wave in a straight microfluidic channel. Consequently, our system can temporally manipulate acoustic radiation force without active controls.

The Stability of Generalized Ricci Solitons

Abstract: In this paper, I computed the second variation formula of the generalized Einstein-Hilbert functional and prove that a Bismut-flat, Einstein manifold is linearly stable under some curvature assumption. In the last part of the paper, I prove that dynamical stability and linear stability are equivalent on a steady gradient generalized Ricci soliton $(g, H,f)$ which generalizes the result done by Kr\"oncke, Haslhofer, Sesum, Raffero, and Vezzoni.

The Relationship between Drivers' Hazard Perception Ability and Cognitive Traits of Empathizing-Systemizing: Pilot Study based on Gaze Response Speed

Abstract: Introduction: Previous research (Danno & Taniguchi, 2015) demonstrated that near-miss incident experience was basically reduced by the Empathy Quotient (EQ) and was disturbed by the Systemizing Quotient (SQ) when the Empathy Quotient was low, based on the Empathizing and Systemizing (E-S) model using a web questionnaire survey [1]. It means that drivers with a low EQ and a high SQ had a higher rate of near-miss incidents. It was claimed that drivers with a stronger Empathizing function may have better danger perception ability, despite the fact that when Empathizing is poor, the Systemizing function may weaken hazard perception ability. The D score (standard SQ (T) score minus standard EQ (T) score) that near-miss incident was then discovered to have a considerable impact on the near-miss occurrence experience. Method: Those findings suggested that a D score, which is used to categorise "E-S types," should be related to near-miss incident experience, i.e., hazard perception ability. The D scores were supposed to be related to the cognitive ability to estimate the mental situations of other road users and predict their behaviour, or to recognise stable laws in traffic situations. The purpose of this study was to look directly at the relationship between a driver's physical visual attention ability (gaze movement) and hazard (near-miss incident) perception ability at various D levels. Drivers' Real-time Useful Field of View (rUFOV) [2] was measured in a driving simulator with six traffic scenarios under normal and hasty driving conditions. Results: The results of seven individuals with various D scores revealed that under hasty driving conditions, a driver's visual attention capacity (gaze movement) falls in proportion to their scores. Conclusion: Since a D score is used to categorise "E-S types," this pilot study highlighted the prospect that individual differences in cognitive trait using the E-S model could be a promising tool for understanding the process of hazard perception.

Discovering surface reaction pathways using accelerated molecular dynamics and network analysis tools

Abstract: We present an automated method that maps surface reaction pathways with no experimental data and with minimal human interventions.

Transient small displacement since the end of 2020 at Noto peninsula, Japan, revealed by Sentinel-1 InSAR time series analysis

Abstract: <p>At the end of 2020, anomalous transient surface deformation was observed by an operational GNSS network at the Noto peninsula, Japan. Although the Noto peninsula locates far from the plate boundary, seismic observations recorded that seismic swarms were accompanied with this transient deformation. Nishimura et al. (2021, presentation at the 2021 Geodetic Society of Japan) estimated that this deformation and swarms may be associated with the intrusion of water from the subducting oceanic plate. Here I performed Sentinel-1 InSAR time series analysis to obtain more detailed view of this transient displacement and to investigate the mechanism of this phenomenon.<br>In the analysis, at first I created interferograms from Sentinel-1 IW SLCs using ISCE2 software. Then these interferograms were used for the LiCSBAS time series analysis. Orbital and topographic fringes were modeled and removed based on precise orbit information and SRTM 1-arcsecond DEM. No atmospheric corrections were applied. I used both ascending and descending paths so that I could calculate 2.5 dimensional analysis to derive quasi-horizontal and quasi-vertical displacements.<br>The result of Sentinel-1 time series showed that the transient displacement seems to start since the end of 2020, which is consistent with the result from the GNSS observation. The estimated largest surface velocities became 13 mm/year in ascending and 15 mm/year in descending. The 2.5 dimensional analysis suggested that the uplift was concentrated at the eastern front of the peninsula, which is also consistent with the GNSS observation. The derived displacement fields suggested that there is an inflation source but this need to be further investigation by, for example, using elastic spherical and/or rectangular fault models.<br>By the presentation, I will perform the InSAR atmospheric correction and source modelling and show these results.</p>

Gröbner Bases of Toric Ideals Associated with Matroids

Abstract: In 1980, White conjectured that the toric ideal of a matroid is generated by quadratic binomials corresponding to a symmetric exchange. In this paper, we compute Gröbner bases of toric ideals associated with matroids and show that, for every matroid on ground sets of size at most seven except for two matroids, Gröbner bases of toric ideals consist of quadratic binomials corresponding to a symmetric exchange.

Data-driven Steering Torque Behaviour Modelling with Hidden Markov Models

Abstract: Modern Advanced Driver Assistance Systems (ADAS) are limited in their ability to consider the driver's intention, resulting in unnatural guidance and low customer acceptance. In this research, we focus on a novel data-driven approach to predict driver steering torque. In particular, driver behavior is modeled by learning the parameters of a Hidden Markov Model (HMM) and estimation is performed with Gaussian Mixture Regression (GMR). An extensive parameter selection framework enables us to objectively select the model hyper-parameters and prevents overfitting. The final model behavior is optimized with a cost function balancing between accuracy and smoothness. Naturalistic driving data covering seven participants is obtained using a static driving simulator at Toyota Motor Europe for the training, evaluation, and testing of the proposed model. The results demonstrate that our approach achieved a 92% steering torque accuracy with a 37% increase in signal smoothness and 90% fewer data compared to a baseline. In addition, our model captures the complex and nonlinear human behavior and inter-driver variability from novice to expert drivers, showing an interesting potential to become a steering performance predictor in future user-oriented ADAS.

Technology trends of automotive semiconductors for CASE application

Abstract: CASE is a new word referred to “Connected”, “Autonomous”, “Shared & Service”, “Electric”. In the CASE era, technological advances in CASE are significantly changing the concept of the automobile and semiconductors play an important role. Regarding its technology for automotive applications, many specific requirements should be achieved to implement power devices onto EVs, peripheral monitoring sensors on autonomous driving, and SoCs for driving control. Due to the fact that, they should be developed by automobile makers or auto parts manufacturers themselves. In this paper, the technology trends of automotive semiconductors for CASE application will be reported, and the technology of power devices for electrification and peripheral detection sensors for autonomous driving will be focused on.

Detection of Dangerous Drivers based on their Weak Cognitive Ability using the Empathizing-systemizing Model and the Real-time UFOV Method

Abstract: Introduction: Many traffic accidents are caused by driver-related errors, with cognitive ability being the most important direct cause. Since not all drivers are dangerous (accident-prone), the most effective method to prevent accidents is to detect a “dangerous” driver, which is, driver in weak cognitive ability before a traffic accident occurs and to what extent weak in his cognitive ability during driving. This study is intended to detect drivers’ inherent cognitive styles as the primary cause of traffic accidents. Method: This study utilized a web questionnaire survey [1] to collect examples of both traffic accidents and near misses that occured during ordinary driving, classifying participants into three cognitive style of E-S types (Type E, B, and S) with Empathizing-Systemizing (E-S) model and its Empathy Quotient (EQ) and Systemizing Quotient (SQ) scales as indices of cognitive traits to detect the relationship between traffic accidents/near misses experienced in the past. Results: The outcomes allowed us to discriminate between a "safe" driver, Type E, and a "dangerous" driver, Type S, who scored highest for the number of near misses. We could identify those differences and measure in terms of driver’s visual attention capability (gaze response speed) under the driving simulation environment incorporating with the rUFOV method [2]. Conclusion: The potential of weak cognition style can be detected in advance of driving with E-S type, and to measure “degree of accident-proneness” in advance of driving, since not all drivers are equally “dangerous”. Then, we propose that the E-S model can be incorporated into the driver aptitude test (before a licence provision).

Bayesian optimization with constraints based on explanation-based learning and its application in materials research

Abstract: Abstract The present paper describes an algorithm for improving the convergence of Bayesian optimization (BO) using the constraints derived from symbolic learning and inference for observations by explanation-based learning (EBL). During the BO process, the constraints based on observations generalized by the domain knowledge (theory) of EBL are applied to maximization of the acquisition function, thereby improving the efficiency of exploration. For the maximization, a (1 + 1)-evolutionary strategy is exploited to incorporate the constraints. In addition, knowledge regarding the improvements is provided as a human-readable representation, which is reflected by the domain theory. Using synthetic and real-world data provided by experiments of lithium-ion batteries to maximize ionic conductivity, the effectiveness of the proposed algorithm is demonstrated. The obtained knowledge is reasonable in the domain theory regarding high ionic conductivity. Thus, the effect of improving explainability for the optimization is shown.