Electric vehicle intelligent starting and acceleration control method and system based on AI

Abstract

The invention relates to the technical field of electric vehicle control, in particular to an AI-based electric vehicle intelligent starting and acceleration control method and system. The method comprises the following steps: acquiring a driving log of a vehicle owner, performing intelligent vehicle optimization after starting, and constructing an intelligent starting mode selection strategy; physiological state data of a driver are obtained, state change evolution analysis is carried out, and a personalized driving portrait is constructed; holographic state sensing and vehicle virtual simulation are carried out, and a vehicle synchronous digital simulation model is constructed; radar feedback sensing information and a real-time monitoring image in front of a vehicle are collected, traffic situation prediction and vehicle scene prediction and perception are carried out, and a vehicle scene prediction and perception model is constructed. According to the method, the driving starting mode of the vehicle is intelligently selected, vehicle owner requirements and environment available power constraints are matched, and vehicle acceleration efficiency maximization and safety optimization are achieved.

Description

Technical Field

[0001] The present invention relates to the field of electric vehicle control technology, and in particular to an AI-based electric vehicle intelligent starting and acceleration control method and system. Background Art

[0002] Continuous breakthroughs in artificial intelligence (AI), particularly the widespread application of deep learning, reinforcement learning, and big data analytics, have provided novel solutions for intelligent starting and acceleration control of electric vehicles. By integrating multi-source vehicle sensor data, driver behavior, and environmental dynamics, AI-based intelligent control methods enable precise modeling and real-time optimization of the vehicle's powertrain, effectively improving power response speed and energy efficiency. Furthermore, these intelligent control methods possess strong adaptive capabilities, dynamically adjusting starting and acceleration strategies based on the driver's individual needs and changing road conditions, significantly enhancing driving comfort and safety. However, current EV starting and acceleration control technologies still face numerous challenges. Firstly, the complex and ever-changing vehicle operating environment makes it difficult for traditional control algorithms to capture and process this rich dynamic information in real time, resulting in insufficient control strategy adaptability. Secondly, the significant differences in individual drivers make existing methods difficult to achieve truly personalized powertrain requirements, limiting further improvements in system performance. Furthermore, there remains a lack of efficient intelligent solutions for the joint analysis and real-time coordination of complex multi-dimensional parameters, such as battery pack status, motor performance, and vehicle dynamics, hindering the overall optimization of the starting and acceleration process. Therefore, a more intelligent EV intelligent control approach is needed. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes an AI-based electric vehicle intelligent starting and acceleration control method and system to solve at least one of the above technical problems.

[0004] To achieve the above objectives, the present invention provides an AI-based intelligent starting and acceleration control method for electric vehicles, comprising the following steps: Step S1: Obtain the owner's driving log, perform post-start intelligent vehicle optimization, and build an intelligent start mode selection strategy; Step S2: Acquire the driver's physiological state data, analyze the state change evolution, and build a personalized driving profile; Step S3: Collecting battery pack monitoring parameters, performing holographic state perception and vehicle virtual simulation, and building a vehicle synchronous digital simulation model; Step S4: Power demand prediction is performed based on the vehicle synchronous digital simulation model and personalized driving profile, and battery pack power is redistributed based on the intelligent start mode selection strategy.

[0005] In this specification, an AI-based electric vehicle intelligent starting and acceleration control system is provided, which is used to execute the above-mentioned AI-based electric vehicle intelligent starting and acceleration control method, including: The intelligent start module is used to obtain the owner's driving log, perform intelligent vehicle optimization after startup, and build an intelligent start mode selection strategy; The driving profile module is used to obtain the driver's physiological status data, analyze the evolution of the status change, and build a personalized driving profile; Synchronous simulation module, used to collect battery pack monitoring parameters, perform holographic state perception and vehicle virtual simulation, and build a synchronous digital simulation model of the vehicle; The scene prediction and perception module is used to collect radar feedback sensor information and real-time monitoring images in front of the vehicle, perform traffic situation prediction and vehicle scene prediction and perception, and build a vehicle scene prediction and perception model; The adaptive power distribution module is used to predict power demand based on the vehicle's synchronized digital simulation model and personalized driving profile, and redistribute battery pack power based on the intelligent start mode selection strategy.

[0006] The beneficial effects of the present invention include: By analyzing historical driving behavior data, the system can identify the user's typical post-start driving patterns (such as rapid acceleration, slow driving, and preheating), and then develop a personalized start-up response curve, achieving a personalized intelligent start-up experience. Historical data analysis helps the system predict the user's start-up rhythm and analyze the user's vehicle use scenarios, such as commuting and traveling. It also identifies the user's urgency and intelligently selects and preloads the vehicle's driving mode to better match the owner's travel schedule. This optimizes the vehicle's pre-start response time and energy pre-allocation, reduces start-up abruptness, and improves ride comfort. A real-time driver profile is constructed using physiological signals such as heart rate, pupil, and electromyography to determine the driver's current mood and stress state, providing context-based decision-making for start-up and acceleration control. If the system detects that the driver is not in optimal condition, it can delay start-up and limit initial acceleration response, helping to prevent misoperation and accidents. This profile model will provide a data foundation for subsequent driving style learning and power response adjustment, promoting the evolution of acceleration control towards "context-aware driving." By collecting multiple parameters such as SOC, voltage balance, and temperature gradients, the system assesses battery health and power delivery capabilities in real time, improving the safety and accuracy of the drive system. Combined with digital twin simulation, the system can simulate the impact of different starting and acceleration strategies on battery load, thereby formulating more optimal energy distribution and output plans. Holographic perception helps detect battery anomalies in advance, reducing wear and tear and sudden failures, maximizing battery life and minimizing repair costs. The system proactively adjusts acceleration strategies to match current road conditions by sensing and predicting factors such as traffic flow, preceding vehicle behavior, and obstacles. Incorporating forward vision and radar information into the acceleration control logic effectively avoids overacceleration or delayed response due to unclear road conditions. By accurately identifying scenario types (e.g., congestion, highway, and open road), the system selects the appropriate acceleration response to maximize efficiency and optimize safety. The system dynamically adjusts motor output power based on comprehensive calculations based on factors such as battery status, vehicle load, and environmental resistance, achieving real-time optimal energy utilization. Through model-driven real-time calculations, the system can rapidly adapt to complex traffic conditions such as uphill and downhill slopes, congestion, and sharp road changes, improving maneuverability and energy efficiency. Combining virtual simulation models with real-world scenario predictions, a data-physics-behavior control framework is established to achieve optimal matching of global power and acceleration strategies. Based on a profiling model, the system predicts the driver's upcoming maneuver intentions (such as sudden acceleration or a steady start) and proactively adjusts battery power distribution and output mode to ensure consistent responsiveness and user experience. By combining profiling with the actual environment, the system dynamically adjusts the acceleration curve slope and output speed to achieve "follow-the-hands" control. Personalized acceleration strategies help reduce unnecessary high power output, extend battery life, and reduce overall energy consumption, ultimately achieving green and energy-saving control goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1This is a schematic flow chart of the steps of an AI-based electric vehicle intelligent starting and acceleration control method of the present invention; DETAILED DESCRIPTION It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0008] This application provides an AI-based electric vehicle intelligent starting and acceleration control method and system. The execution entities of the AI-based electric vehicle intelligent starting and acceleration control method and system include, but are not limited to, the following: mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc. equipped with the system, which can be regarded as general computing nodes of this application. The data processing platform includes, but is not limited to, at least one of an audio and image management system, an information management system, and a cloud data management system.

[0009] See also Figure 1 The present invention provides an AI-based intelligent starting and acceleration control method for electric vehicles, comprising the following steps: Step S1: Obtain the owner's driving log, perform post-start intelligent vehicle optimization, and build an intelligent start mode selection strategy; Step S2: Acquire the driver's physiological state data, analyze the state change evolution, and build a personalized driving profile; Step S3: Collecting battery pack monitoring parameters, performing holographic state perception and vehicle virtual simulation, and building a vehicle synchronous digital simulation model; Step S4: Power demand prediction is performed based on the vehicle synchronous digital simulation model and personalized driving profile, and battery pack power is redistributed based on the intelligent start mode selection strategy.

[0010] In the embodiment of the present invention, see Figure 1 , is a schematic flow chart of the steps of an AI-based electric vehicle intelligent starting and acceleration control method of the present invention. In this example, the steps of the AI-based electric vehicle intelligent starting and acceleration control method include: Step S1: Obtain the owner's driving log, perform post-start intelligent vehicle optimization, and build an intelligent start mode selection strategy; In this embodiment, a complete driving monitoring log is continuously collected from each vehicle start to the beginning of steady-state driving via the onboard T-BOX, central gateway controller, or advanced driver assistance system (ADAS) platform. Collected data includes, but is not limited to, vehicle start time, initial accelerator / accelerator pedal opening, pedal depression duration, acceleration command response time, initial acceleration curve, vehicle speed trajectory, torque response data, battery output power, motor speed, tire slip, vehicle posture parameters (such as pitch angle), whether a start failure or response delay occurred, and the driver's waiting time before starting the vehicle. These logs are segmented by start cycle, with an average of 80 to 150 start behavior records collected per driver per month, forming a high-density behavioral sample library. Each start behavior is then analyzed item by item. The first step is to classify start responsiveness. Based on the acceleration curve and response time, behaviors are categorized into "quick response," "smooth start," and "hesitant delay" types. The second step is to classify acceleration trajectories after start using time series clustering (DTW) technology to identify similar behavioral patterns. Taking actual experiments as an example, six main startup behavior patterns were identified from 30,000 startup behaviors by 200 users. The "short response-medium acceleration" pattern accounted for 27.6%, and the "slow start-rapid acceleration" pattern accounted for 21.3%. The third step was to introduce a behavioral evaluation model to comprehensively score each startup behavior based on energy efficiency, comfort, and safety margins. Evaluation metrics included acceleration jerk (variation), peak energy consumption, battery temperature rise, and peak tire slip rate.

[0011] After completing the behavioral analysis, the driver's starting behavior patterns are clustered and modeled using unsupervised learning algorithms (such as K-means++ and GMM Gaussian mixture models). Drivers are then categorized into different behavioral preference types, such as "conservative starter," "responsive starter," "energy-saving starter," and "aggressive starter." Furthermore, the consistency of each starting behavior type under different environmental conditions (such as sunny and rainy days, hill starts, and congested roads) is analyzed to construct a "starting behavior stability feature vector" for the driver. Based on these analysis results, an intelligent start mode selection strategy engine is developed. This strategy automatically retrieves the driver's behavioral preference model and current environmental perception data at each vehicle start. Through multi-factor rule matching and weighted fusion decision-making, the most appropriate start control mode is selected. Available start modes generally include energy-saving start mode, responsiveness priority mode, comfortable and stable mode, anti-skid control mode, and self-learning optimization mode. For example, when it is determined that the car owner has a "responsive" driving style and the current environment is a high-temperature, high-adhesion road with smooth traffic, the "response priority mode" will be enabled first; and in a rainy environment on a slope, even if the car owner prefers quick start, the "anti-skid control mode" will be activated first to ensure safety.

[0012] Step S2: Acquire the driver's physiological state data, analyze the state change evolution, and build a personalized driving profile; In this embodiment, multimodal sensing is used to acquire driver physiological state data. In practical applications, vehicles can continuously collect driver physiological characteristics through sensors integrated into seats, steering wheels, dashboards, and even wearable devices (such as PPG photoelectric sensors, EDA conductivity detectors, ECG modules, eye trackers, EEG headbands, and temperature / humidity detection modules). Key data dimensions include heart rate (BPM), heart rate variability (HRV), skin conductivity (GSR), palm temperature, eye movement frequency and eye closure duration, facial expression tension, myoelectric activity, and palm perspiration levels. This data is typically sampled in cycles of 500ms to 1s and initially processed and cached in the vehicle controller to form time series data on the driver's physiological state during driving. For example, during a long drive, the driver's heart rate, skin resistance, and pupil dilation state are recorded once per second, forming a data vector set of over 3,000 continuous state nodes. This then enters the state change evolution analysis phase. During this process, dynamic time warping (DTW), principal component analysis (PCA), and variability detection algorithms are used to identify patterns and extract trends from continuous physiological state vectors. This allows the identification of physiological state evolution patterns under specific driving tasks or scenarios (such as high speeds, nighttime driving, urban congestion, and frequent start-stop driving). For example, cluster analysis of skin conductance and heart rate fluctuations in 50 drivers during rush hour revealed a clear "emotional tension-high responsiveness" pattern in 32 drivers, while the other 18 exhibited a "emotional stability-low involvement" trend. This analysis is further correlated with driving behavior data (such as accelerator pedal force, frequency of sudden acceleration events, and vehicle yaw rate changes) to clarify the degree of coupling between physiological state changes and actual behavior. Based on these behavioral-physiological state relationships, a personalized driving profile modeling engine is constructed. This model utilizes a graph neural network (GNN) or Transformer architecture, treating each physiological state node as a node entity in the graph, with different behavioral triggers or state transitions forming weighted edge relationships within the graph. The model training process is based on supervised label learning and partial self-supervised learning mechanisms. By continuously mapping feature vectors to individual driver data across multiple scenarios, it ultimately constructs a personalized, up-to-date driver status profile. This profile includes not only behavioral style classifications (e.g., conservative, nervous, aggressive) but also response patterns under different conditions (e.g., increased acceleration response with increased heart rate, increased steering frequency with increased GSR), providing a crucial reference for subsequent control strategy selection. The profile model is not a static template but rather has adaptive updating capabilities. After each completed drive, the deviation between the profile's predicted behavior and the actual behavior is compared. If there is persistent deviation, a model update mechanism is automatically triggered, retraining some neural network weights to maintain consistency between the profile and the driver's state.Through this mechanism, a dynamic modeling effect that "varies from person to person and changes with time" can be achieved, ensuring that even if the driver's condition is affected by factors such as lifestyle, mental state or age changes, their core driving characteristics and behavioral logic can still be accurately captured.

[0013] Step S3: Collecting battery pack monitoring parameters, performing holographic state perception and vehicle virtual simulation, and building a vehicle synchronous digital simulation model; In this embodiment, with the vehicle's battery management system (BMS) at its core, a high-precision sensor network distributed within and around the battery pack continuously acquires multi-dimensional monitoring parameters during battery pack operation. These parameters primarily include single cell voltage (typically sampled with an accuracy of ±5mV), battery module temperature (sampled using thermistors or thermocouples with an accuracy of ±0.5°C), battery SOC (state of charge), SOH (state of health), battery internal resistance, charge and discharge currents, battery pack heat flow variations, insulation status, and external cooling conditions. Data is refreshed at a frequency of 1Hz to 10Hz and synchronized and cached with a central controller via the vehicle's CAN bus, Ethernet, or over-the-air (OTA) connection. For example, in a mass-produced electric vehicle, approximately 200,000 battery operating status data points can be collected per hour, covering over 50 physical quantities, ensuring sufficient timeliness and resolution for dynamic modeling. This then enters the holographic state perception phase. This process first uses multi-source data fusion algorithms (such as Kalman filtering and multi-dimensional time series fitting) to reduce noise, normalize, and remove outliers from the sensor data to ensure the reliability of the input model. Based on this, local clustering and time series aggregation methods are used to identify temperature gradients, voltage differences, and impedance evolution trends across different regions and modules within the battery pack. For example, during high-power acceleration, the temperature of the module in the lower right corner of the battery pack rises faster than in other areas, and the corresponding module's internal resistance increases at 1.8 times the average, indicating a preliminary risk of local thermal imbalance. This analysis is organized in a graph structure, with each cell or module as a node and the coupled thermal and electrical behaviors between them as edges. By constructing a "battery holographic state diagram," comprehensive perception, dynamic prediction, and safety warnings of the battery's operating state are achieved. A SOX dynamic evolution model (including SOC, SOH, and SOP) is also introduced to predict the battery's remaining capacity under different operating conditions, thereby dynamically extending the state from "current observation" to "future prediction." Once holographic perception is complete, the vehicle virtual simulation phase will be initiated based on this multi-dimensional perception data and evolutionary graph. During this phase, the battery thermal management model, motor control model, chassis dynamics model, and external environment model are integrated to build a complete vehicle-level synchronized simulation structure within a digital twin platform (e.g., based on MATLAB / Simulink, CarSim, or a self-developed physical simulation platform). The focus is on the battery pack's response under different operating conditions (e.g., 0–60 km / h acceleration, continuous uphill climbs, frequent starts and stops, energy regeneration, etc.), including output capacity, temperature rise rate, pressure differential limit, and range degradation. For example, three simulated starting accelerations were performed under the conditions of an ambient temperature of 35°C, a battery SOC of 65%, and a fully loaded vehicle. The simulation results showed that at the fourth second mark of the third start, the battery's thermal response significantly increased, with a cell temperature difference of 6.2°C. Simultaneously, the maximum output power decreased by approximately 12%, indicating performance degradation under continuous load.Such simulations will be repeated to build a multi-scenario, multi-variable battery response database for the entire vehicle.

[0014] Step S4: predicting power demand based on the vehicle's synchronized digital simulation model and personalized driving profile, and redistributing battery power based on an intelligent start mode selection strategy; In this embodiment, a synchronized digital simulation model of the vehicle is used as input to quantitatively analyze the vehicle's current operating state and potential, thereby calculating a "Vehicle Comprehensive Performance Index" (VCPI) representing the vehicle's current performance limit. This index is primarily based on the battery pack's real-time output capacity, motor efficiency, electric drive response delay, chassis load status, and tire adhesion. Its calculation core relies on a multi-dimensional dynamic weight fusion model, which normalizes performance parameters from different physical domains and assigns weights. For example, in one simulation run, a certain electric vehicle model was subjected to 10 0–60 km / h acceleration simulations. The battery SOC was maintained at 60%, the average maximum motor output torque was 180 N·m, and the tire sideslip angle remained stable within 2.5°. The measured VCPI value in this state was 0.84 (out of a maximum score of 1.0), indicating that the vehicle still has high acceleration potential. However, the battery temperature was approaching the safety threshold (42.3°C), requiring controlled power release. Environmental constraint parameters acquired from the vehicle's scene prediction perception model are incorporated, particularly key factors such as road slope, adhesion coefficient, drag parameters, traffic density, and the distribution of dynamic objects ahead. The goal of this phase is to translate environmental constraints into a power release boundary—that is, to determine the acceleration and power ceiling that the vehicle can safely and reasonably release under the current traffic and road conditions. By simulating vehicle response and environmental interaction under varying accelerations and integrating scenario simulation tools (such as SUMO or a proprietary traffic flow prediction engine), a "power release envelope" is constructed. For example, in a real-world scenario, the current road slope is detected to be 3.5%, the adhesion coefficient is 0.65, and there are two pedestrians and a slow-moving vehicle within 15 meters ahead. There is a 75% probability of active braking within the next two seconds. Based on this, the maximum releasable acceleration is limited to 1.2 m / s², and a limit window (0-140 N·m) is imposed on the motor torque.

[0015] Adjustments are made based on a multi-objective optimization algorithm to simultaneously meet the three control objectives of power responsiveness, energy efficiency, and safety margins. Common methods include reinforcement learning-based policy optimization (such as DDPG and PPO), mixed integer programming (MILP), and nonlinear constrained optimization (such as SQP). The vehicle comprehensive performance index (VCPI) is used as the core parameter for current power capability, and the power release parameters from the scenario prediction model serve as boundary constraints. An objective function is constructed for optimization. The optimization process, with an update cycle of seconds, adjusts the power distribution curve in real time, ensuring smooth and efficient vehicle launches under actual operating conditions without triggering energy consumption peaks or exceeding safety thresholds. For example, in one simulation, the optimal power release path for a specific acceleration condition was evaluated as: 90 N·m in the initial stage, 130 N·m in the middle stage, and convergence to 70 N·m in the final stage. Overall power consumption was controlled at 1.8 kWh / 100 km, acceleration time was 6.3 seconds, the Jerk index was kept within 1.4 m / s³, and a user perception score exceeding 80%. A power distribution mapping relationship is constructed using a state-space model. Specifically, specific driving requirements, vehicle status, and environmental characteristics are mapped to a corresponding torque-current distribution matrix, forming a state-driven, responsive power control logic. This logic supports dynamic adjustment and feedback correction for each acceleration request, enabling rapid response to unexpected operating conditions and complex traffic situations. For example, if a sudden lane change by a preceding vehicle is detected, reducing available acceleration space, the system can immediately reduce acceleration torque output and activate energy recovery logic to achieve a stable control transition.

[0016] In this embodiment, time series modeling is performed on historical driving data, using a recurrent neural network (RNN) or long short-term memory (LSTM) network to predict the driver's behavior in specific scenarios. For example, it can identify a driver who frequently uses a "short burst of high-intensity acceleration followed by rapid deceleration" pattern during morning rush hour in urban areas, with an average starting torque requirement of approximately 120 N·m and an average acceleration of 1.6 m / s² within three seconds of starting. Furthermore, this behavior can be linked to physiological state profiles. For example, when a driver is in a state of high stress, their power demand pattern tends to be more aggressive. Ultimately, a "target power demand prediction" is output, including the predicted torque value, acceleration value, duration, and acceptable delay range. Based on the predicted target power demand, the current power distribution strategy is dynamically adjusted, focusing on battery pack power redistribution. This requires not only meeting the driver's power request but also considering the battery pack's current output capacity, temperature control status, health status, and safety margins. During power redistribution, the battery pack's internal status information (such as each module's state of charge (SOC), state of hydration (SOH), voltage consistency, and thermal profile) is utilized to adjust the power output weight of each module using a multi-constraint optimization algorithm. For example, when a predicted acceleration request reaches 150 N·m, lasts 2.8 seconds, and has an estimated peak power demand of 52 kW, the system evaluates the load capacity of multiple battery modules. Module A's current temperature has reached 45°C, approaching its thermal management limit, while module C is under a lower load at 32°C. The system then reallocates the current path, increasing module C's output ratio and reducing module A's power output, ensuring that the battery maintains thermal balance and health while meeting demand. This process is accomplished collaboratively by a dynamic current controller and a model predictive control (MPC) algorithm, with real-time adjustments made dozens of times per second. This generates an "intelligently matched acceleration control engine" with customized driving characteristics. This engine is not a single control model, but rather an intelligent control framework with multiple input parameters, dynamic response capabilities, and individual adaptability. It integrates driver profile parameters, environmental perception parameters, vehicle status parameters, and battery pack limit parameters to select the optimal response before each acceleration request is triggered. The control engine uses reinforcement learning (such as PPO or DDPG) combined with rule-based guidance to fine-tune the strategy. For example, different drivers will trigger different torque release paths and energy consumption management strategies for the same acceleration request. For Type A drivers (conservative), the energy-optimizing path is prioritized; for Type B drivers (aggressive), response speed and acceleration are prioritized; and for Type C drivers (conservative), the power release limit is further compressed and a more secure buffer is introduced.

[0017] In this embodiment, the detailed implementation steps of step S1 include: Obtain the owner's driving log; Define a time series behavior cycle and extract periodic vehicle startup behavior from the owner's driving log to obtain the vehicle startup behavior data stream within the cycle; Performing driving behavior analysis after each start of the vehicle start-up behavior data stream to generate driving behavior characteristics; Extracting a vehicle startup timestamp from the vehicle startup behavior data stream; Performing periodic time series fitting on the timestamps to construct a periodic timestamp marker graph; Intelligent vehicle startup optimization is performed based on driving behavior characteristics and cycle timestamp marking graph, and an intelligent startup mode selection strategy is constructed.

[0018] In this embodiment, intelligent vehicle-mounted devices (such as the OBD data acquisition module, CAN bus access module, BMS, and T-Box remote communication module) collect and aggregate the owner's historical driving data logs to establish a foundational data pool for subsequent intelligent start-up mode strategies. The collected data primarily covers the following five categories: 1) vehicle start time and motor activation time; 2) vehicle speed changes, acceleration curves, and braking force within the first 10 minutes after start-up; 3) battery state of charge (SOC) and voltage fluctuations during start-up; 4) environmental data such as temperature, humidity, and traffic density; and 5) driving behavior, such as accelerator pedal response and throttle response curves. The recommended data sampling frequency is 1Hz to 10Hz to fully capture vehicle state changes. The recommended collection period covers at least 30 days of continuous driving behavior to reflect the owner's consistent usage habits. After collection, the data undergoes cleaning and normalization, including outlier removal (such as GPS drift and current fluctuations), timestamp standardization, and data null value interpolation. The final output is a unified formatted driving monitoring log sequence. The "vehicle start behavior cycle" from typical driving behavior is extracted and a periodic start behavior data stream is constructed. First, by performing time series frequency statistics on historical data, high-frequency daily start periods (such as 7:30 AM and 5:40 PM) are identified. A start-stop clustering algorithm based on a sliding time window (such as DBSCAN combined with time density clustering) is then used to identify periodic behavior. Each "vehicle ignition start to stable acceleration phase" is defined as a "start behavior unit," which is used to divide the starting time of each cycle. All relevant data within this cycle (such as current, voltage, ambient temperature, acceleration change, etc.) is extracted and combined to form a periodic start behavior data stream.

[0019] To improve the accuracy of cycle identification, a minimum number of periodic start behaviors (e.g., no less than three per week) can be set as a filtering criterion. Each cycle data stream consists of multiple fields, including the start timestamp, vehicle speed change rate, battery voltage response, and acceleration curve, providing a complete input structure for subsequent behavioral pattern recognition. Driving performance after the start behavior is quantitatively analyzed to extract the driver's driving behavior characteristics. This method utilizes a segmented time window analysis technique, using the first 1 minute, 3 minutes, and 5 minutes after the start behavior as different time windows. Key driving behavior indicators, such as average acceleration, maximum speed, throttle response curve slope, frequency of sudden acceleration, and energy consumption growth rate, are extracted to form a multidimensional behavioral feature vector. To model the stability and individual characteristics of driving behavior, rate of change and fluctuation amplitude are introduced as supplementary features, such as acceleration standard deviation, voltage fluctuation amplitude, and number of braking mutation points. These behavioral features are arranged chronologically to form a "driving behavior evolution map." A dynamic time warping (DTW) algorithm is also introduced to compare the similarity of behavioral profiles across different cycles, thereby identifying consistent driver behavior and typical driving patterns. These behavioral characteristics serve as input for the intelligent start optimization algorithm, predicting the vehicle's required energy output curve and acceleration expectation model after start, thereby improving the accuracy and comfort of the starting response. In the periodic data stream, the start timestamp marks the beginning of each "ignition + motor activation" event and is directly related to the temporal distribution of driving behavior and daily habits. In this step, the motor start signal points recorded by the vehicle control unit (VCU) are combined with GPS data and the current activation signal output by the BMS to extract the start timestamps (precisely recorded in year, month, day, hour, minute, and second) from all historical data. These timestamps are then organized to form a continuous "start time series."

[0020] For further processing, timestamps can be aggregated at weekly or monthly granularity, and a start density map can be plotted over a 24-hour period to identify high-frequency start periods. Furthermore, daily and weekly start frequency and standard deviation statistics are used to determine whether the owner has a highly regular travel habit. This timestamp information will serve as a crucial component of the time variable in the subsequent construction of a periodic prediction model. After obtaining the start timestamp sequence, a "periodic time distribution model" of vehicle starts must be constructed. This step employs periodic function fitting techniques to analyze the start time series, such as extracting the primary cycle frequency using a Fourier transform (FFT) or using weighted least squares fitting to approximate high-frequency start periods. The results are presented as a periodic timestamp marker map, which divides a 24-hour day into smaller time periods (e.g., 15-minute segments) and plots the probability density of start events within each segment. For example, if a high start probability occurs continuously between 07:15 and 07:30, it will be labeled as a "high-confidence periodic start point" and analyzed in conjunction with the corresponding driving behavior characteristics. This timestamp signature map not only reflects the driver's typical start-up rhythm but also provides a temporal basis for the next step in intelligent mode selection. Specifically, when comparing weekdays and weekends, this signature map can be used to determine the degree of rhythm variation, enabling more granular scenario recognition. A driving mode classification model (using unsupervised clustering algorithms such as K-Means and HDBSCAN) is constructed to classify different combinations of driving behavior characteristics into patterns such as "steady cruising," "frequent rapid acceleration," and "urban commuting." The cycle timestamp signature map is used to match the current time to predict the user's upcoming start cycle type, and this is then combined with the behavioral characteristic model for inference. Ultimately, intelligent start-up mode recommendations are implemented based on both time and behavior. For example, if the identification indicates "commuting mode with a high frequency cycle at 7:30 AM," the starting current output curve, optimal energy release path, and driving torque response model are intelligently configured at the moment of ignition. The BMS current preheating mechanism is also pre-activated to reduce acceleration response delay and enhance the overall driving experience. This strategy's implementation proposal combines software and hardware deployment: The strategy is implemented by the vehicle's intelligent control unit, executed through closed-loop regulation between the VCU and the motor controller, and continuously optimized through cloud-based cyclical behavior models. This ultimately creates a personalized, scenario-adaptive intelligent start-up mode selection mechanism, driving electric vehicles towards greater intelligence, energy efficiency, and efficiency.

[0021] In this embodiment, the specific steps of performing intelligent vehicle startup optimization based on driving behavior characteristics and periodic time stamp mark graph and constructing an intelligent startup mode selection strategy are as follows: Identify the owner's real-time vehicle entry time; Perform optimal similarity matching calculation on the owner's real-time vehicle entry time based on the periodic timestamp marker graph to obtain the most similar startup timestamp in history; Analyze the owner's vehicle usage scenario based on the most similar historical startup timestamp to obtain the current owner's vehicle usage scenario; Calculate the travel time delay of the most similar start timestamp in history to obtain the delay parameter of this trip; Intelligently select the vehicle driving mode based on the driving behavior characteristics and the delay parameters of this trip to obtain a preloaded driving mode; Perform intelligent vehicle startup optimization based on preloaded driving modes and build an intelligent startup mode selection strategy.

[0022] In this embodiment, the starting point of the user's actual use of the vehicle is captured to provide a time reference for subsequent decision-making. The identification of vehicle entry time mainly relies on the interaction behavior signals between the car owner and the vehicle, including car key sensing, door unlocking records, smartphone APP connection, NFC or Bluetooth near-field communication, etc. For example, when the car owner carries the smart key and approaches the vehicle at a certain distance (usually within 1.5 meters) and triggers the unlocking behavior at the same time, this time point will be recorded as the starting time of entering the vehicle. In order to improve the accuracy and anti-interference of recognition, the occurrence of entry events is usually determined by multi-signal fusion. For example, the owner's mobile phone Bluetooth is successfully paired with the vehicle, the door unlocking signal is reported, and the seat sensor in the car is activated. After multiple signals are cross-verified, the confirmation logic of the "vehicle entry event" will be triggered. The timestamp of this event is a high-precision record at the millisecond level, usually synchronized with the UTC standard time. During the experiment, the recognition mechanism was deployed on 100 electric vehicles, and the behavioral data of the owners entering the vehicles were collected. On average, each person recorded about 2 to 3 valid entry events per day, and the recognition accuracy rate reached over 98%. For users with regular commuting, the entry time is highly consistent with their periodic behavior, providing a solid foundation for subsequent time matching and behavior prediction.

[0023] After obtaining the real-time vehicle entry time, this time point is matched against the owner's periodic timestamp signature graph for optimal similarity. The periodic timestamp signature graph is a daily temporal behavior graph constructed from vehicle start data from the past 30 to 90 days. It contains the probability distribution of vehicle starts by the owner at different time periods each day. The current entry time is matched against all marked times in the graph. A sliding time window search strategy, combined with a Gaussian weighting mechanism or a simple temporal distance sorting algorithm, is used to find one or more historical start time records closest to the current time point. For example, if the current time is 07:48, the periodic graph is used to find the most frequent start times between 07:45 and 08:00 for the past 28 days. The record closest to the current time is selected as the "historical most similar start timestamp." The matching process not only considers temporal distance but also incorporates multiple constraints, such as the completeness of the driving behavior at that time point, normal battery usage, and the complete travel path after the start, to avoid selecting anomalous data as matching samples. In the experiment, a matching calculation can be completed within 2 seconds on average, and the matching accuracy is maintained above 92%, which is especially stable for users with regular schedules.

[0024] After obtaining the most similar historical startup timestamp, the vehicle usage scenario analysis phase begins. The goal is to infer the driver's current vehicle usage intention and scenario type, helping to determine whether preheating is necessary, selecting power or energy-saving mode, and other actions. This analysis relies on a behavioral label classification model trained on driving behavior data from the 30 minutes following the most similar historical startup time. This model includes driving duration, average speed, maximum speed, battery drain, and whether the vehicle traveled on expressways or complex roads. These behavioral characteristics are then combined with typical vehicle usage scenarios (such as morning rush hour commuting, weekend shopping trips, child pickup, and short nighttime trips) to form a training model. Inference is then performed using a similarity algorithm or a supervised classification model (such as a decision tree or support vector machine). Assuming that at 7:50 a.m., most trips historically lasted around 20 minutes, routed to the office, and maintained a stable average speed of less than 30 km / h, the current trip is likely a "daily commuting" scenario. A large sample of user data reveals that vehicle usage at specific times exhibits high stability and repeatability, particularly in weekday mornings and evenings, where user behavior scenarios show a high consistency rate of 87%. Weekend scenarios are more diverse but still exhibit a certain degree of regularity. The time difference between the matched most similar historical start timestamp and the current vehicle entry time is calculated to determine whether the current trip is delayed or advanced. This delay parameter is more than a simple difference in absolute time; it is a crucial indicator of changes in the user's travel rhythm. If the current entry time is 15 minutes later than the historically similar time point, it is inferred that the user's travel plans may have been delayed due to temporary factors, necessitating a reassessment of their driving response strategy. Experimental data shows that for commuters, the probability of travel time delays exceeding 10 minutes is 22%, with delays typically fluctuating between 5 and 20 minutes. To quantify the impact of these delays on vehicle start-up response, delays are categorized into levels: for example, under 5 minutes is considered normal, 5 to 15 minutes is considered moderate, and over 15 minutes is considered severe. Each level corresponds to a different response strategy. Especially in the case of delayed travel, it is estimated that the driver may want to reach the destination as soon as possible, so a more positive acceleration response is given in the subsequent startup optimization.

[0025] Combining the driving behavior characteristics, vehicle usage scenario category, and trip delay parameters analyzed previously, the vehicle enters the intelligent driving mode selection phase. This is controlled by a multi-input, multi-factor AI decision-making model that integrates periodic behavior maps, scenario labels, driving style, and delay conditions to determine whether to select energy-saving, standard, dynamic, or adaptive driving modes. For example, if the driver prefers faster acceleration in the same scenario during a weekday commute and the current trip delay exceeds 10 minutes, the driver will be more likely to select the dynamic response mode, which provides faster starting acceleration and acceleration response frequency. The model primarily utilizes an ensemble learning architecture, such as XGBoost or a reinforcement learning strategy based on LSTM time series modeling. The model training data is derived from thousands of past driving behavior records, and the output decision value is the mode label and corresponding control parameter configuration. In actual deployment, driving mode decisions are completed within 2-3 seconds after vehicle entry, rapidly responding to the driver's personalized needs. The model's self-learning capabilities also allow for real-time optimization based on driver feedback, continuously adjusting its understanding of driver habits over time.

[0026] In this embodiment, the detailed implementation steps of step S2 include: Based on the vehicle-mounted multimodal sensor array, the driver's heart rate variability, eye movement trajectory, eye blink frequency and closure degree are collected in real time to construct the driver's physiological state data; Performing multi-time point state change evolution analysis on the multidimensional feature vector to obtain multi-time point state evolution characteristics; Mining potential driving intentions and evaluating driving scene urgency based on multi-time point state evolution characteristics to generate potential driving intentions and driving scene urgency; Based on potential driving intentions and the urgency of driving scenarios, personalized real-time portrait modeling is performed to build a personalized driving portrait.

[0027] In this embodiment, a multimodal onboard sensor array collects multiple physiological indicators, including the driver's heart rate variability (HRV), eye movement trajectory, blink frequency, and eye closure, in real time to construct a multidimensional feature vector of the driver's physiological state. The sensor array typically includes a photoplethysmography (PPG) sensor embedded in the steering wheel, a mid-infrared / near-infrared camera in the cockpit for eye tracking and eye closure detection, and a micro-vibration sensing module in the seat. These sensors provide 30 data samples per second, ensuring high-frequency capture of physiological state changes. In an experimental setup, the system was deployed in 50 electric vehicles, and 200 users were tracked during daily driving for three months. Heart rate variability was analyzed by analyzing the standard deviation of the RR interval (SDNN) to extract autonomic nervous system states; eye movement trajectory was modeled by the movement path of the pupil center within the visual field; blink frequency was based on the number of blinks per minute, and the duration of each blink closure was combined to determine fatigue level. These data are uniformly time-stamped and standardized, and a multi-dimensional physiological state vector (usually including 1015 feature dimensions) is formed according to a fixed time window (e.g., every 5 seconds) to form the basic expression of the driver's physiological state.

[0028] The multi-dimensional feature vectors are analyzed for their state evolution at multiple time points to uncover the dynamic short-term evolution of the driver's physiological state. Specifically, the feature vector sequences at different moments are fed into a time series modeling framework, such as a long short-term memory (LSTM) or a gated recurrent unit (GRU), using a sliding time window (e.g., the past 60 seconds with a step size of 5 seconds). This captures the trend changes and burst patterns of the driver's physiological state within a unit of time. For example, a sustained downward trend in HRV, accompanied by a decrease in blink frequency and an increase in eye closure, indicates that the driver is gradually entering a state of fatigue or decreased attention. In the experiment, an LSTM-based temporal evolution feature extraction model was trained to label and identify five typical state evolution trends (focused → distracted, alert → fatigued, relaxed → anxious, alert → tense, and stable → chaotic), achieving a classification accuracy of over 88%. This model continuously outputs a label for the driver's state evolution and a state change speed indicator at the current point in time, providing in-depth behavioral evidence for subsequent driving intention and scenario assessment. The time-point state evolution features are input into the potential driving intention recognition and driving scenario urgency assessment modules to further infer the driver's current driving intention (e.g., rushing, relaxed driving, inattention), and determine whether the driving task is high-risk or high-stress. The intention recognition component utilizes a multi-layer neural network model for classification. Its training samples are based on real-world driving behavior and physiological state data, and a supervised sample set is established using driver subjective questionnaires and behavioral labels. Five typical driving intention types are identified: "rushing," "relaxing," "passive fatigue," "alert and defensive," and "highly focused." Each intention type corresponds to a specific set of physiological state evolution characteristics. For example, "rushing" is often characterized by decreased HRV but increased eye movement frequency, decreased blinking, and high concentration. "Relaxing" exhibits higher HRV and a relatively flat eye movement path. Furthermore, a decision tree combined with a behavioral statistical model assesses the urgency of the driving scenario, including the frequency of sudden acceleration, the number of throttle changes, route congestion, and the degree of synchronization with the driver's physiological responses. This comprehensive score outputs the scenario urgency level. The evaluation results are divided into three categories: "low emergency", "medium emergency" and "high emergency". Experimental data show that the model's scene evaluation accuracy remains above 90% under actual driving conditions, and can issue early warnings for sudden dangerous conditions 1 to 2 minutes in advance.

[0029] After assessing both potential driving intention and the urgency of the driving scenario, a personalized real-time driver profile is constructed, creating a dynamic driver state profile. This profile not only encompasses the driver's current physiological state and behavioral trends but also tracks changes in similarity with historical behavioral profiles, forming a dynamic adaptive model of the driver. This profile modeling utilizes a vector graph approach. Each driver possesses a "psychological-behavioral profile" that evolves over time. An autoencoder compresses high-dimensional features, and a graph neural network (GNN) is used to model the relationship structure of the profile. Before each drive is initiated, the system dynamically recommends the most appropriate driving mode and control parameters based on the profile state, combined with the intention and scenario level. For example, if the driver is detected to be in a "high-urgency state" and their historical profile indicates a tendency to accelerate rapidly in similar situations, the "power boost + response optimization" control mode will be applied to the launch strategy. The profile model also supports real-time iterative updates, proactively fine-tuning the next launch and control strategy based on feedback from the differences between the most recent driving behavior and the profile. Experimental results show that after adopting the portrait-driven personalized control strategy, users' driving satisfaction scores increased by 14%, and the accuracy of predicting behavioral deviations increased to more than 93%, greatly improving the adaptive capability of intelligent control.

[0030] In this embodiment, the detailed implementation steps of step S3 include: Collect battery pack monitoring parameters; Perform temperature distribution analysis on the multi-dimensional monitoring parameters, calculate voltage differences among multiple battery cells, identify internal resistance changes, and perform holographic state perception to construct a holographic state perception map of the battery pack; Collect the motor rotor position, stator temperature, and magnetic field strength, and perform motor performance analysis to obtain a real-time profile of the motor performance; Obtain vehicle posture, tire pressure, suspension load, and fit chassis status information; Conduct vehicle dynamics evolution analysis on chassis status information and construct a vehicle dynamics characteristic map; Conduct vehicle virtual simulation based on the vehicle dynamics characteristic map, motor performance real-time portrait, and battery pack holographic state perception map to build a vehicle synchronous digital simulation model; Collect radar feedback sensor information and real-time monitoring images in front of the vehicle to perform traffic situation prediction and vehicle scene prediction perception, and build a vehicle scene prediction perception model.

[0031] In this embodiment, the onboard battery management system (BMS) acquires multi-dimensional monitoring parameters of the battery pack in real time. These parameters include voltage, current, temperature, internal resistance, SOC (state of charge), and SOH (state of health) for each individual cell, with the acquisition frequency typically around 15Hz. For example, a 75kWh ternary lithium battery pack typically consists of 500 cells distributed across several modules. The operating status of each cell is recorded individually, and a spatial topology is established between the cells and modules to facilitate subsequent thermal distribution and voltage consistency analysis. The collected data is transmitted to a central computing module via the CAN bus or Ethernet. Data integrity and sampling frequency are verified in real time to ensure that each frame of data supports high-resolution time series analysis. This process forms the foundation for battery status awareness and effectively identifies potential battery anomalies, aging trends, and thermal management risk points.

[0032] After data collection is complete, a comprehensive analysis of these multi-dimensional monitoring parameters is conducted, including battery temperature distribution modeling, cell voltage variation measurement, and internal resistance trend identification. The temperature analysis uses thermal imaging to fit the spatial arrangement of the module and cells. Using a thermal balance model, the temperature diffusion path under high load or rapid charging is simulated to identify hotspots or uneven cooling. Cell voltage variation analysis calculates voltage consistency between cells using the maximum and minimum values. Time series voltage response curves are used to identify cells with hysteresis or abnormal capacity decline. Internal resistance analysis uses changes in voltage response during charge and discharge to infer the dynamic trend of AC internal resistance, and combines historical data to identify aging signatures. All this information is integrated into a multi-layered "battery pack holographic state perception map." This map uses modules as the basic unit, with each node carrying composite attributes such as temperature, voltage, and internal resistance. Edges represent the thermal / electrical coupling between cells. This map provides structural support for subsequent battery management optimization, early warning mechanisms, and energy scheduling strategies. In 300 hours of actual vehicle operation data, the perception map can accurately capture more than 80% of abnormal battery cell behaviors and temperature control deviation points, greatly improving perception accuracy.

[0033] Key motor operating parameters, including rotor position, stator temperature, and magnetic field strength, are collected as core data for motor performance analysis. Rotor position is provided in real time by a high-precision encoder or Hall effect sensor with a resolution of up to 0.1°, used to assess the dynamic response and angular velocity changes of the motor's rotation. Stator temperature is measured using thermocouples embedded in the coil or infrared temperature sensors, with a measurement range of -40°C to 180°C, providing a real-time reflection of the motor's thermal load. Magnetic field strength is sensed by flux sensors or magnetoresistive devices to monitor the uniformity and transient changes in the magnetic field distribution. These parameters are used to model the motor's efficiency, thermal stability, and magnetic energy conversion efficiency. Combined with multiple vehicle operating load characteristics, a "real-time motor performance profile" is generated. This profile includes not only efficiency curves and thermal efficiency maps, but also indicators such as electromagnetic torque fluctuation rate, response lag time, and high-temperature risk areas. This profile is refreshed every 5 seconds through a dynamic update mechanism. In an experiment involving 1,500 kilometers of measured data, real-time profiling accurately predicted six signs of motor high temperature and two abnormal magnetic flux deviations, significantly enhancing the motor's fault prediction capabilities and adaptive adjustment basis.

[0034] Further information is obtained about the vehicle chassis' real-time status, including vehicle attitude (such as pitch, roll, and yaw angular velocity), tire pressure, and dynamic load distribution of the suspension. Vehicle attitude is typically acquired via a six-axis inertial measurement unit (IMU) with a sampling frequency exceeding 100Hz per second. Tire pressure is continuously monitored and reported once a minute by a TPMS. Suspension loads, on the other hand, rely on force sensors or electronically controlled damping feedback parameters at the vehicle's four corners, providing real-time information on suspension compression and force transmission. By integrating these sensor data and leveraging vehicle kinematic models and multi-body simulation, the real-time operating state of the chassis is fitted. This state not only reflects the vehicle's stability under various road conditions but also predicts its responsiveness during extreme maneuvers. For example, during cornering, real-time suspension load and attitude changes can be used to determine whether the roll limit is about to be exceeded, and this risk signal is fed back to the driver assistance module. In experimental testing, this fitted model accurately reconstructed the chassis state under highly dynamic maneuvers and successfully identified 16 attitude imbalance events during high-speed obstacle avoidance, providing predictive support for vehicle dynamic control.

[0035] By integrating chassis state information with motor response data, the system analyzes the dynamic evolution of the entire vehicle and constructs a "full-vehicle dynamics characteristic map." This map, with time as the primary axis, constructs the dynamic behavior evolution path using multidimensional feature vectors, encompassing key dimensions such as acceleration changes, yaw rate response, motor torque distribution, suspension force state, and body posture change trajectory. The map is constructed using methods such as time series data clustering, principal component dimensionality reduction, and correlation graph analysis to efficiently extract the dynamic response characteristics of the vehicle under different driving modes. For example, the characteristic map clearly shows that in "city congestion mode," the vehicle dynamics exhibit a state evolution pattern of small fluctuations and low speeds and high frequency, while in "high-speed cruising mode," it exhibits a highly stable and low-energy output path. This map can be used for both retrospective analysis of driving behavior and as an important input for AI models to determine the vehicle's dynamic state. Validated through 200 hours of real-world vehicle sampling and comparison with the map, the accuracy of dynamic state recognition reached 94%, significantly improving modeling capabilities in complex driving scenarios. Using the constructed holographic state perception map of the battery pack, real-time motor performance profiles, and vehicle dynamics maps as core inputs, the virtual simulation phase begins, constructing a synchronized digital simulation model of the vehicle. This model is not a traditional static simulation, but rather a "digital twin" that runs in parallel with the vehicle's real-time data stream. It can proactively assess the potential energy consumption, thermal load, motor efficiency fluctuations, and vehicle dynamic response of each control action before AI control decisions are made. The simulation platform is typically deployed on a high-performance edge computing unit on the vehicle, using a real-time computing engine (such as Simulink Real-Time or an AUTOSAR platform-compatible model). Before each driving maneuver is initiated, the simulation model is run for 300-500ms, performing a rapid calculation and outputting performance comparison data for multiple control strategies. Experimental results show that this virtual simulation mechanism can detect 8% of potential motor overheating risks in advance, effectively reduce peak energy consumption by 5%, and improve the real-time feedback accuracy of the control strategy. The synchronous digital simulation model is the key bridge for AI control to move from "perception-driven" to "predictive-driven", marking the entry of vehicle startup and acceleration control into the intelligent stage of visualization, quantification and parallelization.

[0036] In this embodiment, the acquisition of radar feedback sensor information and real-time monitoring images in front of the vehicle, traffic situation prediction and vehicle scene prediction perception, and construction of a vehicle scene prediction perception model include the following steps: Perform all-round environmental scanning based on the vehicle-mounted laser radar to obtain radar feedback sensor information; Calculate road slope, road friction coefficient, and wind speed and direction based on radar feedback sensor information to obtain environmental parameter characteristics; Acquire real-time monitoring images in front of the vehicle; perform traffic flow state recognition and signal light change cycle calculation on the real-time monitoring images in front of the vehicle to obtain traffic state information characteristics; Based on the real-time monitoring image in front of the vehicle, deep visual recognition is performed to analyze the dynamic distribution of pedestrians and vehicle displacement changes to obtain the dynamic target distribution characteristics of the scene; Traffic situation prediction is performed based on traffic status information characteristics and scene dynamic target distribution characteristics, and a traffic situation prediction map is constructed; Carry out vehicle scene prediction perception based on traffic situation prediction map and environmental parameter characteristics, and build a vehicle scene prediction perception model.

[0037] In this embodiment, a vehicle-mounted LiDAR performs a full-scale environmental scan, acquiring radar feedback sensor information as the primary data source for environmental modeling. LiDAR provides 360-degree, high-resolution point cloud data, capturing the spatial structure and obstacle distribution around the vehicle. In actual deployments, 32- or 64-line rotating LiDARs are used, mounted on the roof or sides of the vehicle. The point cloud refresh rate is 10 Hz to 20 Hz, with a horizontal angular resolution of 0.2 to 0.4 degrees, a vertical angle of over 30 degrees, and a detection range exceeding 120 meters. Existing smart new energy vehicles have a variety of radar types that generally meet the hardware parameter requirements of this technology. In dynamic environments, this point cloud data is continuously recorded in a time series, forming a spatiotemporally consistent structure. The point cloud data is preprocessed to filter out static background and high-frequency noise points, while also using voxel grid downsampling to reduce the data dimensionality. The preprocessed point cloud information is fed into the perception engine as core input, providing basic data support for subsequent environmental parameter estimation. After obtaining radar feedback, physical environmental parameters are estimated from the point cloud data, including road slope, road friction coefficient, and wind speed and direction. The radar's 3D point cloud is projected onto the ground area in front of the vehicle. By fitting the angle between the normal vector of the ground point set and the vehicle coordinate system, the longitudinal and transverse slopes of the road ahead are accurately calculated. In the experiment, the least-squares plane fitting method was used to estimate the slope, with an accuracy of ±1.2 degrees. Friction coefficient estimation is based on the reflectivity of the road surface point cloud and historical weather models for classification. For example, rainy, slippery road surfaces exhibit reduced reflectivity and more uniform point cloud density, identifying these areas as low-friction areas. Wind speed and direction estimation is derived from a model combining the radar point cloud change rate and the vehicle's own acceleration correction residuals. Especially at high speeds, the windward point cloud offset and velocity difference can be used as quantitative indicators of wind disturbances. These calculation results are combined to form a set of environmental parameter features, which are updated once per second and used to dynamically adjust the driving control strategy. Real-time image data from the vehicle's front is collected via an onboard camera to identify traffic flow states and analyze signal cycle times, thereby capturing traffic status information. Images are acquired using a high-definition RGB camera with a resolution of 1920×1080 and a frame rate of at least 30 fps, mounted in the center of the front windshield or on the front bumper. After processing these images with a convolutional neural network (CNN), traffic flow parameters such as vehicle density, average speed, and vehicle spacing within the lane ahead can be identified, and traffic flow conditions such as free flow, steady flow, or congestion can be determined. For traffic light recognition, the YOLOv5 or EfficientDet architecture is used to detect the position of traffic lights. The signal cycle is calculated from the inter-frame brightness curve to determine the current light phase duration and remaining time. By modeling the sequence of signal light state changes over the past 30 to 60 seconds, signal change trends can be predicted in advance, enabling adaptive acceleration prediction in dynamic signal environments.Experimental data show that the accuracy of traffic light recognition reaches 97%, and the average signal cycle prediction error is within ±1.3 seconds, providing key rhythm information support for intelligent starting strategies.

[0038] After acquiring traffic state characteristics, the system then performs deep visual recognition based on camera images to analyze dynamic elements in the scene ahead, particularly the spatial distribution and motion trends of pedestrians and vehicles. This system relies on the combined application of a depth estimation network and an object tracking algorithm. For example, models such as Monodepth2 or DPT are used to construct a depth map of the image ahead, followed by SORT or DeepSORT for temporal object tracking. It not only identifies object types (pedestrians, cyclists, motorcycles, cars, etc.), but also calculates their motion trajectories and potential spatial locations over the next 13 seconds, constructing a dynamic object distribution feature map. In specific scenarios such as crosswalks, complex intersections, or school areas, the system automatically enhances pedestrian detection sensitivity, using keypoint recognition and skeleton tracking to predict pedestrian behavior, such as gait speed, direction, and intention. In experimental road testing, the system stably tracked 812 dynamic objects with recognition latency under 150 milliseconds and achieved over 86% accuracy in object motion prediction, providing the foundation for dynamic obstacle avoidance and energy-saving activation strategies.

[0039] Traffic state information features are integrated with the dynamic target distribution characteristics of the scene to perform traffic situation forecasting and analysis, and a traffic situation prediction map is constructed. This map presents the dynamic evolution paths, potential conflict points, and state change trends of all key targets in the entire traffic scene ahead within a short period of time (15 seconds). Using a graph neural network (GNN) structure, each target (vehicle, pedestrian, traffic light) is treated as a graph node, and the relative motion relationship between different targets, predicted trajectory overlap, and spatial proximity are used as graph edge weights. After training, the model outputs a scene situation tensor. This map effectively reflects potential risk hotspots, signal-induced acceleration / deceleration nodes, and path accessibility levels. In a complex intersection scenario, the traffic situation prediction map identified 17 potential vehicle conflict trends and provided intervention prompts 0.81.5 seconds in advance, providing a safety redundancy window for the vehicle control module.

[0040] Traffic situation prediction maps and environmental parameter characteristics are fed into the AI ​​perception and decision-making module to construct a "vehicle scenario prediction perception model," the final high-level fusion model before intelligently initiating vehicle control strategies. This model, based on a Transformer architecture and incorporating a temporal attention mechanism, continuously predicts short-term future scenario states. It also incorporates the vehicle's current state (such as remaining battery charge, motor responsiveness, and dynamics) for personalized adjustments. In the model output, each scenario configuration (for example, green traffic light, flat road, low-density traffic ahead, and no pedestrians) is mapped to an optimal starting response strategy, including parameters such as motor output delay, maximum starting power, and battery discharge curve. Experimental data shows that before and after control strategy optimization, the average vehicle starting time was shortened by 0.9 seconds, power consumption was reduced by 6.3%, and starting comfort scores improved by 15%. Model inference latency was kept below 300 milliseconds, fully meeting the requirements of real-time deployment.

[0041] In this embodiment, step S4 includes the following steps: Calculate the comprehensive vehicle performance index for the vehicle synchronous digital simulation model to generate a dynamic performance baseline; Perform environmental constraint assessment on the vehicle scene prediction perception model, calculate the maximum available acceleration of the scene, and generate the maximum available acceleration parameters of the scene; Perform multi-objective optimization based on the maximum available acceleration parameters and dynamic performance baseline of the scenario to generate the optimal starting torque curve; Perform environmental adaptive power distribution based on the optimal starting torque curve and build a power distribution strategy; The power demand is predicted based on the personalized driving profile, and the power distribution strategy is used to redistribute the battery pack power to build an intelligent matching acceleration control engine.

[0042] In this embodiment, the vehicle's comprehensive performance index is calculated for a synchronized digital simulation model, relying on a pre-built multi-source data-driven model encompassing the battery, motor, chassis, and dynamics profiles. This model integrates factors such as maximum power output, motor response characteristics, vehicle mass, and aerodynamic parameters to simulate the vehicle's dynamic response under multiple typical driving conditions. By performing structured analysis on the simulation results for different scenarios, such as quantifying parameters such as acceleration curves, vehicle take-off time, torque output curves, yaw rate change, and energy consumption, this heterogeneous data is normalized into a dimensionless comparison benchmark. The weighting of each performance indicator is then determined using the analytic hierarchy process (AHP) to ultimately generate a comprehensive performance index for the vehicle under different operating conditions. This comprehensive index not only reflects the vehicle's instantaneous performance capabilities but also, through time series analysis, depicts the performance evolution curve over the entire acceleration process. By applying a curve envelope to the performance index under all operating conditions, a dynamic performance baseline covering different driving scenarios is generated. This baseline represents the theoretical maximum performance boundary achievable by the vehicle under different conditions, given current hardware capabilities and control strategies. It serves as a key reference baseline for subsequent multi-objective optimization of the control strategy. The vehicle's scenario prediction perception model is evaluated for environmental constraints, and based on this, the vehicle's maximum available acceleration capability in a specific scenario is calculated. This step utilizes the previously constructed multimodal perception system as input for environmental conditions, including traffic signal information identified by cameras, terrain and obstacle distribution acquired by radar, vehicle posture information provided by the IMU, and wind speed, direction, and slope estimated by sensors. The environmental perception module comprehensively models information such as road surface conditions, adhesion coefficient, and ramp angle. This information is combined with dynamic target features such as real-time traffic density, preceding vehicle behavior, and pedestrian activity areas to constrain the calculation of the vehicle's safe acceleration in the current environment. For example, on slippery, curving urban roads, an empirical regression model combined with real-time road surface reflection information determines that the current road adhesion coefficient is insufficient to support standard acceleration, resulting in a dynamic downward adjustment of the maximum available acceleration. During this process, the vehicle's inherent power capacity remains in the model as a hardware upper limit, but an "environmentally acceptable acceleration limit" is generated based on the environmental modeling results. This value is continuously updated over time and serves as a dynamic lower bound for the performance baseline. By integrating traffic flow behavior, the system can also predict interaction risks during acceleration, such as potential braking by the preceding vehicle or the possibility of an oncoming vehicle entering the lane. This allows for further safety redundancy reduction in acceleration, ensuring stable, controllable, and unobtrusive vehicle behavior, providing precise dynamic boundaries for intelligent control strategies.

[0043] Based on the aforementioned dynamic performance baseline and the scenario's maximum available acceleration parameters, the core control optimization phase begins: a multi-objective optimization search to generate the optimal starting torque curve. This process aims to improve dynamics, ensure safety, and balance comfort and energy efficiency. A global search is conducted using the motor output torque variation sequence as the optimization variable. First, an optimization time window is defined (for example, the 0-5 second starting phase). This phase is discretized into multiple time slices, each corresponding to a motor output torque value, forming a time series torque control curve. To achieve this optimization search, a multi-objective genetic algorithm based on an evolutionary mechanism is used. Training with a large number of simulation samples yields a torque output solution with high adaptability. Each candidate curve is then simulated in a simulation platform during vehicle acceleration, and performance scores are calculated across five dimensions: proximity to maximum acceleration, acceleration smoothness, overall energy consumption, tire slip probability, and vehicle stability. Curves that fail to meet the scenario's maximum available acceleration limits are eliminated. After extensive iterations and screening, an optimal starting torque curve is ultimately generated that balances all objective constraints. This curve may provide high response torque at the start of a launch to enable a rapid breakaway, then gradually increase to a stable plateau to ensure continuous traction. It also automatically converges as high load approaches to mitigate the risk of slip. This curve serves not only as a reference input for the motor controller but also, through parameterization, adjusts in real time to changing environmental constraints, forming a dynamic curve template with adaptive characteristics. This coordinates the various vehicle power controllers to implement an environmentally adaptive power distribution strategy. Specifically, it analyzes multiple factors in real time, including the drive motor efficiency range, the current battery temperature (e.g., suppressing high torque requests below 0°C), the transmission efficiency between the motor and the wheel, and the activation status of the energy recovery mode. In single- or dual-motor configurations, the front-to-rear drive ratio is adjusted in real time based on adhesion conditions and tire load to avoid power slip and energy waste. Furthermore, in complex environments such as slopes or icy roads, the starting torque limit is adaptively lowered to improve acceleration response time and the anti-slip control level is automatically increased to enhance stability. This strategy not only ensures the intelligent starting capability of electric vehicles in different climates, terrains and traffic environments, but also has the ability to adaptively adjust to long-term operating conditions such as battery aging and driving performance degradation, realizing closed-loop tuning of intelligent starting and acceleration control under all working conditions.

[0044] In this embodiment, the power demand prediction based on the personalized driving profile and the battery pack power redistribution based on the power distribution strategy to build an intelligent matching acceleration control engine include the following steps: Perform power demand prediction based on personalized driving profiles to generate target power demand prediction values ​​for driver licenses; Dynamic acceleration trajectory planning is performed based on the predicted power demand value of the driver's license target to generate an acceleration trajectory that matches the power demand; Based on the acceleration trajectory, the power distribution strategy is used to redistribute the battery power and build an intelligent matching acceleration control engine; Perform instant acceleration control based on the intelligent matching acceleration control engine and calculate the acceleration interval; Braking energy recovery and intelligent battery balancing management are performed during acceleration intervals, and vehicle intelligent starting and acceleration control operations are executed based on the intelligent starting mode selection strategy.

[0045] In this embodiment, a personalized driving profile is modeled to accurately predict the driver's power requirements in different scenarios. This driving profile is based on the collection and labeling of long-term driving behavior data, including accelerator pedal stroke amplitude, duration, throttle response frequency, starting timing, acceleration / deceleration style (aggressive / smooth), driving time distribution, and road type preference. During the training phase, supervised learning models (such as XGBoost and Random Forest) are used to extract correlations between different driver characteristic dimensions and power requirements. For example, after collecting approximately 800 km of driving behavior from a driver over three months, statistics revealed that 90% of the driver's starting acceleration on urban roads remained within the 0.8-1.3 m / s² range, while on elevated roads, the driver preferred 2.2 m / s². Combining environmental parameters, real-time driving context, and personal style factors (such as a preference for energy efficiency and high responsiveness), the system outputs a predicted target power requirement for the driver in the current scenario. This predicted value is dynamically adjusted over time to form a "personalized power expectation curve" that directly guides subsequent power trajectory planning.

[0046] Next, based on the predicted target power demand, the dynamic acceleration trajectory planning phase begins. This process aims to convert the power demand into a dynamic time-velocity-acceleration trajectory, which is used to guide motor output control and torque response distribution. First, the vehicle dynamics simulation module is invoked to determine the feasible acceleration range under current physical conditions based on parameters such as vehicle weight, slope, wind resistance, and tire adhesion. Next, based on the driver's desired power level, deep sequence modeling (such as Bi-LSTM) is used to generate a continuous and smooth acceleration trajectory within a 3-5 second time window. The planning process also incorporates a "driving comfort" constraint to control the acceleration jerk within an acceptable range, ensuring that the power output does not cause discomfort to the occupants. The forward traffic environment is also considered. For example, if the preceding vehicle is predicted to brake or swerve within 2 seconds, the planned trajectory will automatically suppress power output and lengthen the acceleration curve. The resulting acceleration trajectory not only meets the driver's expectations but also exhibits environmental adaptability and optimizes energy efficiency, providing the most direct basis for engine control.

[0047] The generated acceleration trajectory is the core of the vehicle's power distribution strategy, specifically the real-time allocation and coordinated management of battery pack power. The key to this stage is translating the target acceleration into a practical battery power output strategy to avoid over-discharge, thermal runaway, or cell voltage drift. The acceleration trajectory is subdivided into several power output points. Constrained optimization algorithms (such as a QP-based power scheduling model) are used to determine the optimal battery output power at each time point, taking into account state parameters such as the battery pack's current state of charge (SOC), cell temperature distribution, and internal resistance trends. Furthermore, within the multi-mode drive architecture, control also determines the output distribution ratio between the front and rear axle motors. For example, in low-grip conditions, power can be tilted toward the front axle to improve traction stability. This entire power output scheduling is handled by an intelligent acceleration control engine, which integrates power prediction, load balancing, thermal safety regulation, and voltage protection mechanisms to form a highly responsive and adaptable power distribution execution platform.

[0048] The intelligent acceleration control engine enters the immediate acceleration control phase, rapidly converting planned power output and torque commands into motor control commands to ensure the vehicle follows the acceleration trajectory stably. Through high-frequency sampling (typically at the 10ms level), it continuously monitors actual vehicle speed, torque response, wheel slip, and drive efficiency, comparing these values ​​against target values ​​in real time. If deviations exceed set thresholds (e.g., torque deviation exceeding ±10Nm or acceleration deviation exceeding ±0.15m / s²), the control commands are immediately corrected, achieving precise closed-loop power regulation. Furthermore, the engine records each transition period between power outputs, known as "acceleration pauses." These pauses are often critical periods of discontinuous power output but offer potential for energy management. These pauses are particularly pronounced in urban traffic, slow-moving traffic, and frequent start-stop situations, occurring an average of two to three times per minute. Effectively utilizing these pauses can help improve vehicle energy efficiency.

[0049] During acceleration intervals, the system automatically switches to energy recovery and battery balancing management mode. For energy recovery, the vehicle's braking motor converts kinetic energy into electrical energy, which is then fed back to the battery. This energy is then precisely distributed based on the current state of charge (SOC) of each cell, achieving active battery balancing. The control system dynamically adjusts the charging rate of each cell, prioritizing cells with lower SOC or lower temperature, thereby reducing the rate of degradation of the overall battery pack's consistency. This process relies on a high-precision cell status monitoring module, achieving an average voltage difference compensation of 1-2% every 10 minutes. Furthermore, based on an intelligent start mode selection strategy, the system integrates driver operating habits, environmental scenario tags, and traffic status predictions to determine the power release rhythm for the next start (e.g., soothing mode, energy-saving mode, or response-first mode), and adjusts engine control parameters in real time. This closed-loop process seamlessly integrates driver behavior understanding, target demand prediction, acceleration trajectory generation, power resource scheduling, and energy recovery and reuse. Through this personalized, environmentally adaptable, and fast-response control, electric vehicles can achieve higher energy efficiency, better driving comfort, and greater safety during the starting and acceleration phases, providing solid technical support for individual intelligent behavior decisions of vehicles in intelligent transportation.

[0050] In this embodiment, an AI-based electric vehicle intelligent starting and acceleration control system is provided, which is used to execute the above-mentioned AI-based electric vehicle intelligent starting and acceleration control method, including: The intelligent start module is used to obtain the owner's driving log, perform intelligent vehicle optimization after startup, and build an intelligent start mode selection strategy; The driving profile module is used to obtain the driver's physiological status data, analyze the evolution of the status change, and build a personalized driving profile; Synchronous simulation module, used to collect battery pack monitoring parameters, perform holographic state perception and vehicle virtual simulation, and build a synchronous digital simulation model of the vehicle; The adaptive power distribution module is used to predict power demand based on the vehicle's synchronized digital simulation model and personalized driving profile, and redistribute battery pack power based on the intelligent start mode selection strategy.

[0051] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0052] The foregoing description is intended only to provide specific embodiments of the present invention, which are intended to enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest manner consistent with the principles and novel features disclosed herein.

Claims

1. An AI-based intelligent starting and acceleration control method for electric vehicles, characterized in that: The following steps are involved: Step S1: Obtain the owner's driving log, perform post-start intelligent vehicle optimization, and build an intelligent start mode selection strategy; Step S2: Acquire the driver's physiological state data, analyze the state change evolution, and build a personalized driving profile; Step S3: Collecting battery pack monitoring parameters, performing holographic state perception and vehicle virtual simulation, and building a vehicle synchronous digital simulation model; Step S4: Power demand prediction is performed based on the vehicle synchronous digital simulation model and personalized driving profile, and battery pack power is redistributed based on the intelligent start mode selection strategy.

2. The AI-based electric vehicle intelligent starting and acceleration control method according to claim 1 is characterized in that: The specific steps of step S1 are: Obtain the owner's driving log; Define a time series behavior cycle and extract periodic vehicle startup behavior from the owner's driving log to obtain the vehicle startup behavior data stream within the cycle; Performing driving behavior analysis after each start of the vehicle start-up behavior data stream to generate driving behavior characteristics; Extracting a vehicle startup timestamp from the vehicle startup behavior data stream; Performing periodic time series fitting on the timestamps to construct a periodic timestamp marker graph; Intelligent vehicle startup optimization is performed based on driving behavior characteristics and cycle timestamp marking graph, and an intelligent startup mode selection strategy is constructed.

3. The AI-based electric vehicle intelligent starting and acceleration control method according to claim 2, characterized in that: The specific steps of performing intelligent vehicle startup optimization based on driving behavior characteristics and periodic time stamp markers and constructing an intelligent startup mode selection strategy are as follows: Identify the owner's real-time vehicle entry time; Perform optimal similarity matching calculation on the owner's real-time vehicle entry time based on the periodic timestamp marker graph to obtain the most similar startup timestamp in history; Analyze the owner's vehicle usage scenario based on the most similar historical startup timestamp to obtain the current owner's vehicle usage scenario; Calculate the travel time delay of the most similar start timestamp in history to obtain the delay parameter of this trip; Intelligently select the vehicle driving mode based on the driving behavior characteristics and the delay parameters of this trip to obtain a preloaded driving mode; Perform intelligent vehicle startup optimization based on preloaded driving modes and build an intelligent startup mode selection strategy.

4. The AI-based electric vehicle intelligent starting and acceleration control method according to claim 1, characterized in that: The specific steps of step S2 are: Based on the vehicle-mounted multimodal sensor array, the driver's heart rate variability, eye movement trajectory, eye blink frequency and closure degree are collected in real time to construct the driver's physiological state data; Performing multi-time point state change evolution analysis on the multidimensional feature vector to obtain multi-time point state evolution characteristics; Mining potential driving intentions and evaluating driving scene urgency based on multi-time point state evolution characteristics to generate potential driving intentions and driving scene urgency; Based on potential driving intentions and the urgency of driving scenarios, personalized real-time portrait modeling is performed to build a personalized driving portrait.

5. The AI-based electric vehicle intelligent starting and acceleration control method according to claim 1, characterized in that: The specific steps of step S3 are: Collect battery pack monitoring parameters; Perform temperature distribution analysis on the multi-dimensional monitoring parameters, calculate voltage differences among multiple battery cells, identify internal resistance changes, and perform holographic state perception to construct a holographic state perception map of the battery pack; Collect the motor rotor position, stator temperature, and magnetic field strength, and perform motor system performance analysis to obtain a real-time performance portrait of the motor system; Obtain vehicle posture, tire pressure, suspension load, and fit chassis system status information; Conduct vehicle dynamics evolution analysis on chassis system status information and construct a vehicle dynamics characteristic map; Conduct vehicle virtual simulation based on the vehicle dynamics characteristic map, real-time portrait of motor system performance, and holographic state perception map of the battery pack to build a synchronous digital simulation model of the vehicle; Collect radar feedback sensor information and real-time monitoring images in front of the vehicle to perform traffic situation prediction and vehicle scene prediction perception, and build a vehicle scene prediction perception model.

6. The AI-based electric vehicle intelligent starting and acceleration control method according to claim 1, characterized in that: The specific steps of collecting radar feedback sensor information and real-time monitoring images in front of the vehicle, performing traffic situation prediction and vehicle scene prediction perception, and building a vehicle scene prediction perception model are as follows: Perform all-round environmental scanning based on the vehicle-mounted laser radar to obtain radar feedback sensor information; Calculate road slope, road friction coefficient, and wind speed and direction based on radar feedback sensor information to obtain environmental parameter characteristics; Obtain real-time monitoring images in front of the vehicle; The real-time monitoring image in front of the vehicle is used to identify the traffic flow state and calculate the signal light change cycle to obtain traffic state information characteristics; Based on the real-time monitoring image in front of the vehicle, deep visual recognition is performed to analyze the dynamic distribution of pedestrians and vehicle displacement changes to obtain the dynamic target distribution characteristics of the scene; Traffic situation prediction is performed based on traffic status information characteristics and scene dynamic target distribution characteristics, and a traffic situation prediction map is constructed; Carry out vehicle scene prediction perception based on traffic situation prediction map and environmental parameter characteristics, and build a vehicle scene prediction perception model.

7. The AI-based electric vehicle intelligent starting and acceleration control method according to claim 1, characterized in that: The specific steps of step S4 are: Calculate the comprehensive vehicle performance index for the vehicle synchronous digital simulation model to generate a dynamic performance baseline; Perform environmental constraint assessment on the vehicle scene prediction perception model, calculate the maximum available acceleration of the scene, and generate the maximum available acceleration parameters of the scene; Perform multi-objective optimization based on the maximum available acceleration parameters and dynamic performance baseline of the scenario to generate the optimal starting torque curve; Perform environmental adaptive power distribution based on the optimal starting torque curve and build a power distribution strategy; Predict power demand based on personalized driving profiles, redistribute battery power based on power distribution strategies, and build an intelligent matching acceleration control engine.

8. The method according to claim 7, characterized in that The specific steps for predicting power demand based on personalized driving profiles, redistributing battery power based on power distribution strategies, and building an intelligent matching acceleration control engine are as follows: Perform power demand prediction based on personalized driving profiles to generate target power demand prediction values ​​for driver licenses; Dynamic acceleration trajectory planning is performed based on the predicted power demand value of the driver's license target to generate an acceleration trajectory that matches the power demand; Based on the acceleration trajectory, the power distribution strategy is used to redistribute the battery power and build an intelligent matching acceleration control engine; Perform instant acceleration control based on the intelligent matching acceleration control engine and calculate the acceleration interval; Braking energy recovery and intelligent battery balancing management are performed during acceleration intervals, and vehicle intelligent starting and acceleration control operations are executed based on the intelligent starting mode selection strategy.

9. An AI-based intelligent starting and acceleration control system for electric vehicles, characterized in that: The method for executing the AI-based intelligent starting and acceleration control method for an electric vehicle according to claim 1 comprises: The intelligent start module is used to obtain the owner's driving log, perform intelligent vehicle optimization after startup, and build an intelligent start mode selection strategy; The driving profile module is used to obtain the driver's physiological status data, analyze the evolution of the status change, and build a personalized driving profile; Synchronous simulation module, used to collect battery pack monitoring parameters, perform holographic state perception and vehicle virtual simulation, and build a synchronous digital simulation model of the vehicle; The adaptive power distribution module is used to predict power demand based on the vehicle's synchronized digital simulation model and personalized driving profile, and redistribute battery pack power based on the intelligent start mode selection strategy.

Images