To improve the accuracy of path tracking and the vehicle handling stability in the process of automatic driving tracking, a trajectory tracking control method based on model predictive control principle and angle compensation was proposed. The overall control structure was divided into two layers, and the upper layer was the trajectory tracking control layer. According to the obtained road information and the driving state of the vehicle, the front wheel steering control input was determined by the model predictive control algorithm. In the lower layer, the sliding mode control theory was used to design the angle compensator, and the yaw rate deviation of the vehicle was taken as the control target to achieve accurate tracking. The results show that compared with the single-point preview strategy, the trajectory tracking control strategy based on model prediction and angle compensation can better control the vehicle to achieve trajectory tracking, and the peak values of yaw rate and sideslip angle are significantly reduced with better stability.

To improve the seismic performance of prefabricated frame joints, based on bionics design, the replaceable artificial plastic hinge connection structure with the function of structural "fuse" was proposed, and nine replaceable artificial plastic hinge specimens with different parameters were designed. The low cyclic reciprocating load was applied to the replaceable artificial plastic hinge by Abaqus to investigate the seismic characteristics of the structure, such as failure mode, hysteretic performance and stiffness degradation. The results show that the failure position of the replaceable artificial plastic hinge structure is mainly concentrated at the flange buckling replaceable connecting plate, and the failure modes are flange buckling of replaceable connecting plate and shear deformation failure of shear dissipating rod. The hysteresis curve of beam end moment-angle is full with bolt slip phenomenon. The equivalent viscous damping coefficient is 0.30-0.45 with the ductility coefficient of 3.64-14.00, which illuminates that the proposed plastic hinge has good plastic deformation and energy-dissipation capacity.

To improve the accuracy, stability and rapidity of trajectory tracking control for driverless vehicle, backstepping method was respectively integrated with sliding mode variable structure control and fuzzy adaptive control. The position and orientation error differential equation of vehicle with three degrees of freedom was established. The control laws of vehicle speed and yaw rate based on backstepping method were derived, and the stability of the system was verified by Lyapunov stability criterion. The sliding mode variable structure and fuzzy adaptive trajectory tracking control methods integrating backstepping method were established respectively. The steady-state error, overshoot and adjustment time of trajectory tracking were used to verify and compare the performance of accuracy, stability and rapidity. The results show that the stability of sliding mode variable structure trajectory tracking control integrating backstepping is the best, and the overshoot of trajectory tracking is reduced to zero compared with backstepping. The fuzzy adaptive trajectory tracking control integrating backstepping method has the best rapidity, and the adjustment time of trajectory tracking is reduced by 18.2% compared with backstepping.

To predict the dynamic characteristics of the thermoelectric generator (TEG) system, a transient computational fluid dynamics (CFD) model for solving the temperature field distribution of TEG system and an analysis model for investigating the transient response characteristics of TEG modules were established based on COMSOL Multiphysics. A hybrid transient CFD-analysis model was proposed and validated through transient experiments. The results show that due to the thermal inertia effect, the conversion efficiency of TEG system exhibits momentary peak value. Compared to the transient fluctuations of exhaust gas temperature and mass flow rate, there is time lag in the hot and cold side temperatures of the thermoelectric semiconductor. Under the highway fuel economy test (HWFET) cycle conditions defined by the U.S. Environmental Protection Agency, by the transient model, the average output power and the average conversion efficiency of the entire TEG system are solved with 35.63 W and 3.40%, respectively. The average error of the transient model′s output voltage is 6.41%. By the proposed model, the transient response characteristics of the TEG system under transient thermal source excitation can be predicted with high accuracy and short computation time.

The warm-mix flame retardant composite modified asphalt was prepared with warm-mix agent (Sasobit), aluminum hydroxide (ATH), organic montmorillonite (OMMT) and base asphalt, and fourier transform infrared (FTIR) scanning, thermogravimetric (TG) analysis and differential scanning calorimetry were used to investigate the coordination and synergistic effect of warm-mix agent and flame retardant and the flame retardant mechanism. The results show that the modification effect of warm-mix agent and flame retardant on the base asphalt is mainly physical modification. The addition of flame retardant can improve the thermal stability, flame retardant and smoke suppression effects of asphalt, and the flame retardant, smoke suppression effects are improved with the increasing of the amount of flame retardant. The flame retardant can form barrier coating on the asphalt surface with wrapping effect, which can effectively reduce the heat flow rate with excellent flame retardant efficiency and combustion process control ability. The compounded flame retardants of ATH and OMMT can provide the synergistic flame retardant effect, and the flame retardant mechanisms are mainly heat exchange interruption and condensed phase flame retardant. When the compounded flame retardant content is 8%, the asphalt has good flame retardant and smoke suppression properties with good economy.

To investigate the characteristics of fuel cell systems and the influence of air systems on fuel cell characteristics, Simulink simulation software was used to model the fuel cell system with voltage model, air compressor model, cathode model and anode model. The fuel cell air system was controlled respectively through proportional-integral-differential (PID) and fuzzy PID. The results show that when the load current changes, both the feedforward PID and fuzzy PID can make the oxygen excess ratio reach the set constant value of 2.0. However, the fuzzy PID responds faster than the feedforward PID, and the output power of the fuel cell under fuzzy PID control fluctuates less, which makes the fuel cell system more stable.

To solve the problems of the existing depth acquisition methods with missing data and low resolution, the novel depth enhancement method based on soft clustering was proposed and named the soft clustering solver. By the strong edge-preservation of the soft clustering method, the accuracy of depth enhancement could be improved. The affinity matrix derived from the soft clustering was combined with the weighted least square model to establish the confidence-weighted least square model in the solver, and the iteratively based solution method was proposed. To evaluate the proposed method, the experiments on several depth enhancement tasks of depth inpainting, depth super-resolution and depth rectification were conducted. Various evaluation metrics were used, including peak signal to noise ratio(PSNR), structural similarity index measure(SSIM), rooted mean squared error(RMSE) and running time. The results show that for depth inpainting, the average PSNR reaches 42.28 with average SSIM of 98.83%. For depth super-resolution and depth rectification, the average RMSE values are 8.96 and 2.36, respectively. By the proposed method, the image with resolution of 2 048×1 024 pixels can be processed with only 5.03 s.

To investigate the dynamic performance of long-span truss fiber reinforced polymer(FRP) bridge after repairing completion, the 92-meter-span FRP truss bridge model was established based on guide girder pushing. The vehicle-bridge coupling algorithm of long-span composite truss bridge was compiled to analyze the effects of speed, vehicle weight and road surface roughness on the mid-span vertical displacement, vertical acceleration and key member stress of the FRP truss bridge. The results show that the more uneven the road surface is, the greater the impact of the vehicle on the bridge structure is. The vertical displacement is more affected by the road surface level than that by the member stress. In the speed range from 20 km/h to 50 km/h, the impact coefficient is less affected by the speed change. When the vehicle speed exceeds 50 km/h, the influence of speed change on the impact coefficient is gradually increased. The peak curve of vehicle mass to bridge dynamic response shows linear increasing trend. When the vehicle mass is 2.80×104 kg with vehicle speed of 30 km/h, the maximum mid-span vertical displacement is 167 mm, which does not exceed the military emergency bridge displacement limit (L/120), indicating that the bridge design can meet the emergency traffic requirements.

To quantitatively evaluate the thermal stability and combustion characteristics of asphalt, the tests of constant temperature heating, oxygen index and cone calorimetry were used to obtain mass loss, oxygen index, heat release rate (HRR), total heat release rate (THR) and mass loss rate (MLR) of two asphalts of base asphalt and rubber asphalt. The results show that at the identical temperature, the mass loss, oxygen index, peaks of HRR, THR and MLR of rubber asphalt and the value of mass loss rate increased with the increasing of temperature are higher than those of base asphalt, while the thermal stability and combustion characteristics of rubber asphalt are inferior to base asphalt. Comparing three tests, the cone calorimetry test results are more reliable because of its more parameters, which is recommended to evaluate thermal stability and combustion characteristics of asphalt.

Taking the building in hot summer and cold winter zone as research object, the transient system simulation program(TRNSYS）simulation platform was used to build the active phase change energy storage floor module. The simulation model of the active phase change cold storage floor air conditioning system and the conventional fan coil plus fresh air conditioning system was established, and the energy saving and economy were simulated and calculated. The effects of chilled water temperature on energy storage and release characteristics of active phase change energy storage floor, room temperature fluctuation and coefficient of performance (COP) of heat pump were investigated under the condition of night cold storage operation in the cooling season and typical day. The results show that under the condition of phase change cold storage at night，the optimal chilled water temperature is 9 ℃ when the cooling load per square meter is 74.78 W. When the cold storage heat pump adopts the intermittent operation scheme with low electricity price at night, the COP of heat pump operation can be effectively improved with reduced operation cost. Compared with the conventional fan coil plus fresh air air-conditioning system, the energy consumption of the active phase change energy storage floor air-conditioning system in the cooling season is reduced by 30.5% with operating cost reduced by 44.24%, and the comprehensive refrigeration energy efficiency ratio reaches 2.38 in summer.

To solve the problem that in traditional switching strategies for lateral tracking control algorithms in intelligent vehicles, the model matching strategy was always in active state with consuming computational resources and reducing motion smoothness, a control mode switching strategy based on least squares support vector machine (LSSVM) was designed. By the strategy, the control modes were dynamically switched based on error assessment under varying operational conditions. The control effectiveness was evaluated from three aspects of safety, comfort and tracking accuracy. The optimal control amount was executed in real time based on a multi-objective optimization mathematical model. The simulation results show that by the switching strategy, the superior lateral tracking performance in intelligent vehicles can be achieved under varying conditions with path tracking deviation range of ±0.1 m, yaw rate fluctuation range of ±2 （°）/s and jerk value range of ±10 m/s3. Compared to traditional single control algorithms, the proposed strategy exhibits improved tracking accuracy, stability and comfort.

To solve the shortcomings of single-network model with learning preference for data， the crop disease identification method was proposed based on multi-model fusion. The single model performance of four mainstream convolutional neural networks of ResNet50, DenseNet121, Xception and MobileNetV2 was evaluated by the proposed method, and the four single models were respectively conducted by multi-model feature-level and decision-level fusion to obtain the identification output. The feature-level fusion method was used to average, maximize and splice the final output feature layer of each sub-network to achieve efficient complementarity of heterogeneous features. The decision-level fusion method was used to maximize and average the output probability of each sub-network to achieve efficient union of probability distribution decisions. The experimental results on the PDR2018 crop disease datasets show that feature-level fusion outperforms decision-level fusion and single-model methods significantly. Among the fusion methods, splicing compression feature fusion achieves the highest recognition accuracy of 98.44%. The cross-database experiments on PlantDoc data subset and the actual images confirm that the feature fusion method exhibits superior accuracy and generalization performance compared to the single-model method.

A new finite-time adaptive tracking control scheme based on prescribed performance was developed to solve the control problem of non-strict feedback systems with input delays and dynamic uncertainties. The time-delay systems were transformed into delay-free systems by Pade approximation and auxiliary intermediate variable, and the unmodeled dynamics was handled by the dynamic signal generated by the first-order auxiliary system. The prescribed performance adaptive tracking control was implemented by the hyperbolic tangent function, and the stability analysis was presented based on dynamic surface control method. Taking the second-order nonlinear system with unmodeled dynamics and input delay as example, the numerical simulation of the proposed control strategy was conducted in MATLAB environment . The results show that the proposed control scheme can avoid the singularity in the derivation of virtual control, and all signals in the closed-loop system are bounded in finite time. The tracking error can converge to prescribed time-varying region, and the control algorithm is effective.

Based on the prototype monitoring of the separated reinforced soil abutment project of Mingchao Expressway, the numerical model of the bridge abutment was established by FLAC2D software, and the force and deformation of the bridge abutment obtained from numerical simulation were compared and verified with the prototype monitoring results of the bridge abutment. Considering the correlation among  filler type, compaction process and filler mechanical properties, the influence of the filler internal friction angle on the working performance of the reinforced soil abutment was simulated and analyzed. The results show that with the increasing of filler internal friction angle, the horizontal earth pressure and vertical earth pressure of the detached bridge abutment are reduced, and the vertical displacement, the horizontal displacement and the reinforcement strain are also reduced. When the filler internal friction angle is increased to a certain degree, the optimization effect on the bearing performance is decreased. When the reinforced soil filler in the project is selected, the filler internal friction angle should be determined according to the construction conditions on the site, and the performance and economic benefit should be weighed to make reasonable choice.

To solve the problems that single energy system could not give full play to its advantages and could not well satisfy the high density of random load spectrum during the operation of tractor，the composite energy system of electric tractor was designed based on parallel connection of hydrogen fuel cell, power battery and supercapacitor. The hierarchical decoupling control energy management strategy based on Haar wavelet and logical threshold rules was designed to realize the hierarchical decoupling of high frequency signal, sub-high frequency signal and steady state signal, and the power signal was distributed after decoupling. The simulation results show that compared with the power following control strategy and the fuzzy control strategy, the average efficiency of the hydrogen fuel cell based on the layered decoupling control strategy is respectively improved by 2.85% and 1.21%, and the vehicle equivalent hydrogen consumption is respectively reduced by 16.11% and 6.88% in the experimental cycle. The proposed control strategy can improve the vehicle economy on the promise of meeting the power demand of vehicle load, and the fuel cell can output power in efficient and stable working state.

Based on the shallow algal bed reactor, the purification effects of Chlorella vulgaris on carbon, nitrogen and phosphorus elements in secondary effluent from municipal sewage treatment plant were investigated under the conditions with different temperature, light intensity, light-dark ratio, CO2 volume fraction and hydraulic retention time in dynamic test. The depth treatment efficiency of Chlorella vulgaris reactor under long-term stable operation with water quality fluctuations was investigated. The results show that the optimal environmental conditions for depth treatment of sewage by Chlorella vulgaris are with temperature of 26 ℃, light intensity of 6 000 lx, light-dark ratio of 14 h∶10 h and CO2 volume fraction of 1%. Under the optimal operation conditions, when the reactor is operated continuously and stably with the hydraulic retention time of 1.0-1.5 d, the removal rates of chemical oxygen demand (COD), total phosphorus (TP) and ammonia nitrogen (NH3-N) in secondary effluent are 52.0%-86.8%, 75.6%-87.4%, 59.1%-75.9% and 64.0%-68.0%, respectively.

To investigate the chaotic characteristics of mixed traffic flow and discern the factors influencing the degree of chaos in mixed traffic platoons, the Cao method and the improved Cao method were employed based on the traditional traffic flow theory to determine the delay time and embedding dimension of the mixed traffic flow. The phase space of mixed traffic flow sequences was reconstructed to determine the chaotic characteristics by calculating the maximum Lyapunov exponent. The influential parameters of the proportion of intelligent connected vehicle (ICV) using cooperative adaptive cruise control (CACC) and the delay time in mixed traffic flow were analyzed. The results show that when the maximum Lyapunov exponent of the headway sequence during the car-following process is less than zero, chaos exists in the mixed traffic flow. Increasing the proportion of CACC vehicles can mitigate chaos in certain time intervals. The car-following system tends to be stable when the proportion of CACC vehicles reaches 0.6. The delay time of CACC vehicles significantly affects chaos, and maintaining low communication delays is essential for CACC vehicles to effectively suppress chaos.

Based on Huangshandong Tunnel project, field data monitoring and numerical simulation by finite element analysis software of Midas GTS NX were used to analyze the laws of vault settlement, peripheral convergence and effective stress change of surrounding rock under different expansion widths. The length and ring distance of anchor bolts were changed to discuss the variation laws of effective surrounding rock stress, lining stress and bolt stress under different bolt lengths, and the optimal bolt length and bolt spacing under different expansion widths were determined. The results show that the decreasing sequence of effective surrounding rock stress is the compressive stress at the arch foot, the compressive stress at the arch waist, the compressive stress at the arch top and the tensile stress at the invert. With the increasing of excavation width, the compressive stress at the top of arch is decreased, while the compressive stress at the waist is increased with unconspicuous  increasing trend. Under the influence of the single-side expansion mode, the stress of expansion side is greater than that of the non-expansion side. Due to the influence of stress concentration, the compressive stress at the arch foot is increased with the increasing of excavation width. For the expansion width of 1.5-2.5 m, the length and spacing of the anchor rod are respective 2.5 m and 1.7 m, and for the expansion width of 3.5-4.5 m, the length and spacing of the anchor rod are respective 3.0 m and 1.2 m.

To solve the problems of picking mushrooms manually in the factory mushroom growing environment with consuming time and labor， a mushroom picking robot was designed and developed. The mechanical structure of the picking robot system was designed by the modular design method, and the positive and negative kinematics solutions of the robot were deduced based on the D-H coordinate method to analyze the dynamic performance of the picking arm. The multi-objective optimization model of picking arm size structure was established for maximizing the picking efficiency, and the optimal solution was obtained with rapid iteration by genetic algorithm. The Adams virtual prototype model was established, and the picking dynamics simulation tests of the robot model before and after optimization were conducted. The simulation results show that under the same motor output torque, the maximum angular velocities of the big arm joint and the small arm joint are increased by 22.9% and 18.6%， respectively, while the single picking time is shortened from 1.60 s to 1.36 s with the picking speed increased by 15%. The developed physical prototype is suitable for the large area automatic picking operation in the sall multi-layer mushroom bed under the factory mushroom growing mode, and the single picking time is 2.0 s.

For the security concerns of communication data and key in automotive ethernet, an automotive ethernet security communication method based on dynamic keys was innovatively designed. The attack tree model for automotive ethernet security threats was established to analyze the faced security threats. The comprehensive architecture for automotive ethernet security communication methods was constructed to address the threats, and the dynamic key algorithm model was designed to generate implicit dynamic keys. The secure communication method based on dynamic keys was designed to encrypt original data through dynamic keys, and the digests were generated and compared at the receiving end. Communication and network attack processes between domains were simulated using Linux-based systems, I.MX6ULL microcontrollers and Windows devices, and the difference degrees of keys, ciphertexts and plaintexts were compared. The communication process was attacked to analyze the data interactions by Wireshark software. The results show that the proposed method can effectively ensure the key security and the confidentiality, freshness, authenticity and integrity of automotive ethernet data, and it can effectively counter tampering attacks and replaying attacks.

To analyze the impact of the power system parameter matching design and energy management scheme on the vehicle performance, taking a fuel cell vehicle as research object, the parameters of main power system components were matched. The power-following energy management strategy and the fuzzy control-based energy management strategy were proposed to improve the economic performance of the vehicle, and the sliding average filtering algorithm was used to optimize the output of the fuzzy control strategy. To model the vehicle and energy management system, the joint simulation platform was established based on AVL-Cruise and Simulink to verify the power performance of the vehicle, and the economic performance of the vehicle under three control strategies was comparatively analyzed. The results show that the fuel cell output power curve of the optimized energy management strategy is smoother and is always within the high-efficiency output interval. Based on the principle of fuel cell priority protection, the optimized energy management strategy has the best performance.

By the original estimation method, the state of charge (SOC) of lithium battery pack is measured after battery discharge. When the battery internal resistance is large, it is difficult to obtain clear open circuit voltage, leading to errors in SOC estimation of lithium battery pack. To solve the problem, the SOC estimation method of lithium battery pack based on segmented polymerization and Kalman filter was designed. Based on the construction of equivalent circuit model, the parameters of lithium battery were identified, and the SOC estimation indexes such as open circuit voltage were defined. The feedback path of lithium battery was switched by piecewide polymerization, and the SOC value of lithium battery was estimated by linearly recursive Kalman filter. The experimental results show that under the pulse discharge condition of lithium battery, the estimation result of the proposed method is basically consistent with the actual SOC value, and the estimation error can be controlled within 0.4% when SOC is 0.6.

The annealing was used to remove the adsorbed gases and impurities on graphene film, and the state of graphene surface was changed. The surface of graphene film before and after annealing was characterized in situ by atomic force microscope and Kelvin probe force microscope. The surface morphology and surface potential map of graphene film before and after annealing were obtained respectively. The work function was calculated according to the surface potential map, and the change mechanism of work function induced by annealing was analyzed. The results show that the annealing treatment causes the escape of water molecular between the graphene film and substrate, which reduces the gap between graphene film and SiO2 substrate and the P-type doping level of graphene film， resulting in the increasing of the Fermi level and the decreasing of work function.

The high-speed digital camera was used to visualize the evolution of electrostatic atomization, where the capillary was considered as spraying nozzle. The morphology of electrostatic atomization pattern and the transformation of the instability breakup pattern at high flow rates were investigated, and the effects of physical parameters, applied voltage and liquid flow rate on the instability breakup of electrified jets in electrostatic atomization were explored. The evolution from cone-jet to simple-jet was studied using hemispherical capillary, and simple-jet atomization of deionized water and ethanol was conducted using common capillary. The results indicate that cone-jet mode can be gradually transfered into simple-jet mode with the increasing of flow rate at specific voltage. For the initial state of jetting, with the increasing of external potential, the instability breakup modes of simple-jet for deionized water at high flow rates are mainly varicose instability breakup and kink instability breakup. Anhydrous ethanol with low surface tension is easy to be overcome by electric force and easier to break, and the whipping assisted bifurcation and ramified modes also appear.

To solve the problem that the cyclist was vulnerable in the traffic accident, a trajectory prediction method was proposed based on the intention identification of cyclist. The diverse characteristics of cyclists were extracted from the perspective of cyclists and vehicles, such as the direction of movement, the probability of returning and the relative position of the vehicle, and the dynamic Bayesian model network was established under the intersection scenario. The influencing factor of cyclist intention was analyzed to obtain the probability of cyclist intention. According to the results of intention identification, the relevant traffic scenarios were built, and the cycling motion equation was given. Through the cycling trajectory prediction algorithm based on particle filtering methods, combined with the motion equation and observation equation, the future trajectory of the cyclist was simulated with particles simulation set. The data collection hardware platform was set up based on laser radar and monocular cameras to obtain the information related to the cyclist movement and posture. Based on the collected 7 358 frame data, the length of the cyclist entering the intersection area was calculated to evaluate the algorithm. The results show that the proposed algorithm can basically identify the intention of the cyclist 0.24-0.54 s before the riding person crossing the street and can predict the rider′s track set of future 5.0 s. The algorithm is of great significance for improving road security.

Based on the characteristics of slotless brushless DC motor with small size, high power density and cogging torque, the slotless brushless micro DC motor for catheter pump was designed using flexible printed circuit board (FPCB) technology, and the loss analysis and optimal design of slotless brushless micro DC motor for catheter pump were conducted based on 3D multi-physical field coupling model analysis method. The results show that the copper loss of the slotless brushless micro DC motor for catheter pump accounts for about 88% of the whole motor loss, which is the main source of motor heating. In the temperature field simulation, the temperature change can change the motor loss and affect the motor output torque. In the flow field simulation model, the motor temperature can increase the human blood temperature and affect the blood compatibility. The flowing blood takes away part of the heat and reduces the motor temperature. The output torque of the pump motor with optimized winding is improved, and the loss is reduced.

To solve the problem that conventional linear Kalman filtering was increasingly unable to meet the demand of multi-motorized target tracking accuracy, a cooperative tracking method based on adaptive multi-model particle filtering was proposed. The host vehicle and the cooperative vehicle respectively executed the adaptive interactive multi model particle filter (AIMM-PF) algorithm to obtain the motion states of the target vehicles in the environment. By the cooperative vehicle, the tracked target state was sent to the host vehicle through vehicle-to-vehicle communication. The data association and data fusion techniques based on the Hungarian algorithm and the fast covariance crossover algorithm were utilized to achieve cooperative tracking of multiple maneuvering targets. The V2V communication, radar and localization simulation system were built to sense and track seven target vehicles within 200 meters range with two intelligent vehicles as the host vehicle and the cooperative vehicle, and the simulation experiments were completed. The results show that compared with the traditional single-vehicle tracking, by the cooperative tracking, the sensing range is expanded, and the tracking error is reduced by 31.1% without affecting the tracking efficiency.

To quantitatively analyze the road congestion around primary and secondary schools and improve the congestion around schools, the traffic congestion data in school season and non-school season were compared. The congestion index of the road around school was established, and the congestion evaluation function was constructed. The influence degree of each road by school travel was analyzed, and the key congested roads were identified. An algorithm for obtaining traffic big data based on the map open platform was proposed. The roads around 79 primary and secondary schools in the main urban area of Nanjing were analyzed, and the congestion value of each road affected by school travel was given. The results show that the overall congestion degree in school season is higher than that in non-school season, and the average congestion index is 1.973 in school season, which is 18.57% higher than that of 1.664 in non-school season.

To investigate the effect of inlet gas relative humidity on the performance of proton exchange membrane fuel cell stack under variable load conditions, the dynamic cycling condition was designed with two step conditions of current densities ranging from 0.10 to 0.35 A/cm2 and from 0.10 to 1.00 A/cm2, and steady state condition of current densities of 0.10, 0.35 and 1.00 A/cm2. Five sets of tests with humidity levels of 20%, 30%, 50%, 70% and 80% were performed. The results show that the sensitivity of the stack to humidity is different at different current densities under the constant current condition, and humidity has great impact on the stack performance at high current density. The stack voltage appears to be downstroke and upstroke in the variable load process, and the amount of voltage downstroke in the loading process is related to the variable load amplitude and humidity, while the voltage consistency of single cell of the stack is deteriorated and related to the inlet gas relative humidity.

According to the independent and controllable advantage of the torque of each wheel for four-wheel independent driving electric off-road vehicle, to improve the traction and off-road ability of off-road vehicle, a coordinated control strategy of driving torque was proposed. The torque of the driving motor was pre-distributed according to the load of each axis. The optimal estimation of road slip rate was realized by the extreme value searching algorithm based on wheel slip rate and road adhesion coefficient. Proportional-integral-differential（PID）control-sliding mode control（SMC）and state-based driving force redistribution method were used to achieve the coordinated distribution of wheel drive torque for restraining the wheelslip under bad road conditions. Through the co-simulation of CarSim/MATLAB and hardware-in-the-loop（HIL）test, the experimental verification of climbing steep slope, single adhesion road surface, split road and continuous unrolling road was carried out. The results show that by the proposed control strategy, the torque of the driving wheel can be distributed according to the actual working conditions, and the slip rate of wheels is reduced to realize the optimization of the driving force of whole vehicle.