To solve the issue that ensuring the trajectory tracking accuracy of intelligent vehicles could compromise the yaw stability, the four-wheel steering system control strategy was proposed to achieve integrated control of trajectory tracking and yaw stability. The trajectory tracking controller was designed based on model predictive control, and the four-wheel steering angle allocation method was developed based on the Ackermann steering principle. According to the proposed hierarchical control strategy for yaw stability, in the upper layer, the required additional yaw moment was calculated by sliding mode control, while in the lower layer, the torque was allocated with the objective of minimizing the tire adhesion utilization rate. The co-simulation model was established to conduct path tracking simulations. The results demonstrate that compared to the conventional active front steering, under low speed conditions on dry roads, the maximum lateral displacement deviation is approximately reduced by 60.0%, and the peak values of vehicle sideslip angle and yaw rate are respectively decreased by about 91.2% and 15.4%. Under high speed conditions on dry roads, the lateral displacement deviation is reduced by 0.130 m, and the peak values of sideslip angle and yaw rate are respectively decreased by 61.8% and 27.8%. Under extreme conditions on slippery roads at a speed of 120 km/h, the maximum lateral displacement deviation is only 1.810 m with reduction of roughly 83.6%, and the peak values of sideslip angle and yaw rate are respectively reduced by about 99.0% and 90.7%. The proposed strategy significantly improves the driving stability and safety of intelligent vehicles.

The coordinated control strategy of active front-wheel steering (AFS) and direct yaw moment control (DYC) was proposed to enhance the lateral stability of in-wheel motor distributed drive vehicles. Based on the two-degree-of-freedom theoretical model, the AFS system was designed using sliding mode control to correct the front wheel steering angle, while the DYC controller was developed based on model predictive control (MPC) with the objective of maintaining vehicle stability and minimizing tire load through optimal torque distribution across all four wheels. The coordination between AFS and DYC controllers was achieved by the  phase-plane analysis. The results show that under the double lane-change maneuver, the maximum error between vehicle yaw rate and desired value is only 12.8%, and the peak sideslip angles are reduced by 3.80% and 9.40% compared to those of AFS-only and DYC-only control, respectively. Under the sinusoidal lag maneuver, the maximum yaw rate error is 15.1%, and the peak sideslip angles are reduced by 61.40% and 64.01% compared to those of AFS-only and DYC-only control, respectively. The proposed strategy significantly improves the vehicle handling stability and exerts less intervention on longitudinal speed with better speed maintenance and improved driving comfort.

To solve the problem of estimation accuracy decrease or even divergence due to the degradation of particle weights in vehicle state estimation by particle filtering (PF), the improved artificial fish swarm algorithm-particle filter (AFSA-PF) vehicle state estimation method was proposed. To improve the global search capability of AFSA and reduce the risk of falling into local extrema, the attenuation function was introduced to dynamically adjust the field of view to achieve global search and local search using larger and smaller fields of view in the early and late stages of the iteration， respectively. The random step strategy in the movement of artificial fish was changed, and the adaptive step size was used to achieve dynamic switching between large and small step sizes in different situations. The foraging and clustering behaviors of the above improved AFSA were used to optimize the particle weight calculation and particle set resampling in PF state estimation. The improved algorithm was verified by the joint simulation of Carsim and Simulink. The results show that compared with AFSA-PF, by the improved AFSA-PF, the MAE and RMSE of the yaw rate estimation are reduced by respective 40.1% and 34.9% under the dual-line shifting condition, and the MAE and RMSE of the sideslip angle estimation are reduced by respective 35.1% and 33.5%. Under the step conditions, the MAE and RMSE of the estimated yaw rate are reduced by respective 52.7% and 36.3%, and the MAE and RMSE of the estimated sideslip angle are reduced by respective 51.5% and 24.0%.

The active flipping cushion protection strategy was proposed to improve the collision safety in autonomous vehicles. The flipping cushion actively to different angles before collision was used to reduce the occupant injuries. The proposed strategy was verified through simulation experiments，and the collision model was validated using cadaver test data. The active cushion flipping model was adjusted, and the simulations were conducted with the THUMS human body model for  100% frontal collision to determine the cushion flipping angle with the minimum overall injury risk in 300 ms timeframe. The results show that the active flipping cushion measure effectively reduces occupant injury risk in frontal collisions. Specifically, when the cushion is flipped to 35°, the occupant's overall injury risk is minimized. Without the flipping measure, the occupants experience submarining due to the inclined seating posture, which increases the risk of injury to internal organs and head. With the active flipping cushion measure, the submarining is effectively suppressed, which reduces the risk of internal organ injuries, but it can increase the risk of lumbar injury. The active flipping cushion measure unavoidably causes the compression in the chest region. However, compared to the scenario without the measure, the risk of chest injuries is significantly reduced.

To effectively coordinate forward and reverse logistics and respond to the national call for developing low-carbon economy, the vehicle routing problem with simultaneous pickup-delivery and soft time windows (VRPSPDSTW) was investigated. Considering the effects of vehicle speed, load and other factors on carbon emission, the mathematical model was constructed with the sum of vehicle fixed cost, transport cost, carbon emission cost and time window penalty cost as objective function, and the improved ant colony algorithm (IACO) was proposed. The algorithm and the model were verified through the test cases. The calculating results demonstrate that in some cases, compared to the other algorithms, the delivery schemes obtained by IACO can achieve shorter travel distances with the maximum reduction of 10.73%. In the modified RCdp5001 case, when the simultaneous pickup-delivery mode is compared with the separate pickup-delivery mode, the former achieves the total delivery cost reduction of 50.20%. Compared to the model without considering carbon emission factors, the delivery plan generated by the VRPSPDSTW model incorporating carbon emission factors achieves 3.16% reduction in travel distance, 1.53% decrease in total delivery cost and 4.56% reduction in carbon emissions.

As the main pungent ingredient in chili products, the content of capsaicin is one of the important markers for evaluating the quality of chili peppers. Developing straightforward, rapid and sensitive method to determine capsaicin levels that aligns with human sensory capabilities is highly significant. The A549 cells were selected as sensitive element， and the electrochemical bio-sensors with cell/PPy/PEDOT:PSS/AuNPs modified ITO was successfully prepared. To improve the sensitivity of sensor, the electrochemical deposition of AuNPs was performed for 120 s with the concentration of PEDOT:PSS in the PPy/PEDOT:PSS conductive gel of 3.0% and the density of A549 cells of 5×105 cells/mL, and the electrochemical cellular sensors constructed under these conditions showed the best sensing performance. The results show that the detection of capsaicin is realized by differential pulse voltammetry, and the limit of detection is 2.53 μmol/L. The linear range of the assay is 30 to 210 μmol/L. The relative standard deviation of the five times of the assay is 3.25%, and the stability of the assay on the 10th day is 91.19%. The bio-sensor provides new method for the detection of capsaicin.

As a typical low-temperature meat product, Yao-meat has high requirements on the temperature and oxygen content of the storage environment and is prone to deterioration. It is particularly important to develop technology for the real-time monitoring of Yao-meat freshness and storage environment. The method was proposed to monitor the freshness and storage conditions of Yao-meat in real time based on the indication of three-dimensional code of colorimetric sensing. The freshness, oxygen, temperature indicating inks and screen printing were used to prepare the three-dimensional codes. The reliance of the three-dimensional codes on ambient light and camera parameters during the scanning process was reduced by developing smartphone application that included black and white correction programme to read the three-dimensional codes and the colorimetric sensor data. The results show that the recognition accuracy of the three-dimensional code is more than 90% under different storage conditions. The prepared colorimetric three-dimensional code achieves the simultaneous monitoring of changes in storage conditions and freshness during the storage process of Yao-meat, which provides new idea for the freshness monitoring of low-temperature meat products.

To solve the issue of noise samples causing bias in AdaBoost training and the difficulty of achieving stable pruning results with single-objective particle swarm optimization for pruning ensemble models, the AdaBoost method was proposed based on the adaptive penalty and multi-objective particle swarm optimization. The adaptive penalty strategy was proposed to apply weighted penalties to suspicious noise for reducing impact on the subsequent training process. The two-stage multi-strategy multi-objective particle swarm optimization algorithm (TSMSMOPSO) was introduced for ensemble pruning. During the search phase, the particles were accelerated toward non-dominated particles to avoid searching worthless space. The global optimum was selected by considering the trade-off between diversity and convergence. To prevent getting stuck in local optima, the reference particles were randomly mutated to generate elite particles for enhancing population diversity in the exploitation phase. To validate the performance of the proposed algorithm, seven comparison algorithms were evaluated on 16 datasets. The adaptive penalty strategy and TSMSMOPSO were verified through the ablation tests and the pruning comparison experiment. The results show that the proposed algorithm achieves the highest accuracy on 12 datasets and the best F1 score on 13 datasets, where the differences from the suboptimal values are 0.19%-16.67% and 0.62%-16.67%, respectively. Compared to single-objective particle swarm optimization, the pruned ensemble model of TSMSMOPSO is lighter in ensemble size and exhibits more stable pruning effects.

To solve the problems of pose drift and localization failures by traditional visual-inertial methods in mobile systems of intelligent vehicles and robots under low-light conditions, the novel visual-inertial localization method was developed by incorporating image enhancement techniques into the VINS-Mono framework. In the front end of VINS-Mono algorithm, the input image stream was processed with multi-scale Retinex enhancement. The contrast-limited adaptive histogram equalization and the photometric correction module featuring with adaptive correction factor were applied. The weighted fusion strategy based on the average grayscale value and entropy of the processed images was employed to integrate the results. Using wheeled robot for data collection, the trajectory accuracy comparison experiment was conducted. The results show that compared to the scheme without image enhancement and the scheme with baseline enhancement, the proposed method reduces the root mean square error of trajectory tracking by the average of 22.74% and 8.57%, respectively. The proposed approach significantly improves the localization accuracy of visual-inertial navigation systems under low-light environments.

To solve the problems of most current orbital angular momentum (OAM) antennas with few generated modes and poor return loss, the multi-mode OAM microstrip array antenna using graphene in the terahertz frequency band was designed. The structure, material and dimension of OAM antenna were given, and the dimension was optimized. The effect of graphene on antenna performance was analyzed. The performance of graphene OAM antenna was simulated in CST. The results show that by simply adjusting the feeding phase, the vortex waves of 0-4 modes can be generated with center frequency of 3.700 THz, S11 of -63.00 dB, impedance bandwidth of 0.958 THz and gain up to 9.70 dBi. The excellent performance provides viable approach to the effective generation and control of multi-mode terahertz vortex waves.

To solve the problem that the extended state observer (ESO) with fixed control gain was failed to achieve optimal response speed and disturbance rejection in motor drive systems, the variable-gain ESO-based active disturbance rejection control (ADRC) system for permanent magnet hub motor（PMHM） was established. The effect of ESO control gain on the response speed and disturbance immunity of the drive system was analyzed, and the variation law of the control gain was derived. The hyperbolic tangent nonlinear function was introduced into ESO to realize that the control gain was automatically adjusted according to the motor speed error. The variable-gain ESO was established, and the control system of PMHM was designed. The test platform was constructed, and the tests under complex operating conditions were completed. The results show that compared with the conventional ESO, by the proposed method, the response speed of the motor under dynamic conditions is improved by 16.67% to 25.00%, and the disturbance rejection capability under external load disturbances is improved by 33%, while the recovery time after sudden load change is shortened by 37.50%.

To solve the problem of phase lag of the linear extended state observer (LESO) for back EMF estimation in surface permanent magnet synchronous motor (SPMSM)， the adaptive second order generalized integrator  (SOGI) based LESO was proposed, and the position sensorless control of SPMSM based on SOGI-LESO was implemented. The extended state equation of the 3rd-order LESO was compensated based on the back EMF characteristics with introducing the SOGI into the open loop of the equivalent model to eliminate the steady-state phase error of the back EMF estimation. Considering the issue of closed-loop poles varying with rotational speed caused by the compensation, the adaptive law was designed in the dimensionless system for simplifying the parameter tuning and ensuring the global stability of the system, and the back EMF observer based on SOGI-LESO was obtained. The zero steady-state error characteristics of the proposed method were verified through simulation, and the steady-state, dynamic and robustness tests were conducted on the Speedgoat platform. The results show that the proposed method limits the steady-state position errors to 2.5° over the full speed range and reduces the low-speed back EMF harmonics from 3.27% to 1.24% with maintaining the position estimation accuracy below 3.5° under ±50% parameter perturbations, which effectively achieves high precision and strong robustness across wide speed range.

The stiffness constants of composite double-corrugated steel plate shear wall were investigated by combining theoretical analysis and finite element simulation. Based on the orthotropic plate theory, the theoretical formulas of bending stiffness constant and torsional stiffness constant of the composite double-corrugated steel plate shear wall along the strong and weak axes were derived. The derivation results were compared with the stiffness constants of traditional double-corrugated steel plate shear wall. The finite element model was established using ABAQUS software. The bending and torsional stiffness constants were determined from the finite element simulation results and compared with the theoretical calculations to verify the proposed theoretical formulas. The results indicate that compared to the stiffness constants of traditional double-corrugated steel plate shear walls, the bending stiffness constants of the composite double-corrugated steel plate shear wall along the strong and weak axes are increased. The theoretical formulas for the stiffness constants of the composite double-corrugated steel plate shear wall demonstrate high calculation accuracy.

According to the maintenance requirements and the disease characteristics of rural asphalt road, the rural asphalt road performance assessment system was constructed based on four aspects of road cracks, road defects and deformations, road repair and road facilities. The cluster analysis was conducted based on the related indicators, and the rural roads were divided into daily inspection sections and key inspection sections by the K-means clustering analysis method. The geometric average method was adopted to establish the rural asphalt road performance assessment model, and the parameters of the assessment model were calibrated by the fuzzy analytic hierarchy process and entropy method. The maintenance strategies of rural roads under different evaluation levels were proposed. The rural asphalt road sections in District A of Shanghai were selected to verify the rationality and consistency of the model. The results indicate that the technical condition indices of rural asphalt roads in regular inspection sections and key inspection sections are 94.98 and 87.31, respectively. The established evaluation model can identify the impact of major diseases on road conditions and effectively classify the technical conditions of rural asphalt roads. The model is more suitable for rural asphalt roads in the intelligent inspection mode.

To accurately predict the long-term deflection of reinforced concrete (RC) structures and reduce the experimental research costs, the hybrid model integrating enhanced dung beetle optimization (EDBO) algorithm and back-propagation neural network (BPNN) was proposed, which was named as EDBO-BPNN model. Optimizing the BPNN prediction model by EDBO technology, the predictive capability of the hybrid model was enhanced. The predictive capability of the proposed hybrid model was validated by applying the EDBO-BPNN model and existing models to predict the long-term deflection of RC beams under complex environments. The results indicate that the proposed EDBO-BPNN model achieves high accuracy and excellent  stability in predicting the long-term deflection of RC structures. The coefficients of determination (R2) of EDBO-BPNN model reach respective 0.967 5 and 0.957 7 for the training and testing sets, which are higher than those of the comparison models. The root mean square errors (RMSE) are respective 4.242 5 and 5.442 7 with the mean absolute errors (MAE)  of respective 2.416 1 and 4.008 0, and four error metrics all outperform those of the other compared models. The difference in R2 between the training and testing sets for EDBO-BPNN model is only 0.009 8, which is substantially smaller than that of the comparison models with the difference in R2 of 0.160 7. The box plot distributions from 30 independent runs are the most compact for the proposed model, which demonstrates the strong generalization ability and low risk of over fitting.