Unmanned vehicle endurance optimization method and system

Abstract

The invention provides an unmanned vehicle endurance optimization method and system, and belongs to the technical field of unmanned driving. According to the invention, the wind speed is monitored by using ultrasonic or laser radar, the vehicle attitude is evaluated by using the inertial measurement unit, the road condition is analyzed by using the radar and the laser radar, and the road characteristics are identified by using the camera and the infrared sensor cooperatively, so that real-time environmental parameters are comprehensively obtained; then, predicting a future power demand based on historical data and an LSTM model, and optimizing energy distribution in combination with a physical driving force model; in the dynamic driving mode, it is ensured that energy recovery is safe and efficient by adjusting the working state of a motor and feedback current control; and finally, the fuzzy logic controller dynamically generates a control instruction according to the state and demand of the vehicle, so that real-time optimization of energy distribution is realized. According to the method, the endurance mileage of the unmanned vehicle is remarkably improved, the adaptability and economy of the unmanned vehicle under various driving conditions are enhanced, and a foundation is laid for wide application of the unmanned technology.

Description

Technical Field

[0001] The present invention belongs to the field of unmanned driving technology, and specifically relates to a method and system for optimizing the range of an unmanned driving vehicle. Background Art

[0002] As an important part of future intelligent transportation, the endurance of driverless vehicles is one of the key factors affecting practical applications. Currently, driverless vehicles mainly rely on batteries, but the battery capacity is limited. How to effectively manage energy and extend the vehicle's range has become the focus of technical research. Traditional methods are often based on fixed driving strategies and energy allocation schemes, which are difficult to adapt to complex and changing driving environments and dynamically changing road conditions. Therefore, they have limitations in energy saving and range optimization.

[0003] Currently, the main challenges facing vehicle range optimization include the uncertainty of environmental factors, such as wind speed, slope changes, different road conditions, etc., which will significantly affect the vehicle's energy consumption and driving efficiency. In addition, the real-time prediction accuracy of power demand is not high, and the traditional power distribution strategy lacks foresight and flexibility for future needs, resulting in inefficient energy utilization. At the same time, energy recovery mechanisms, especially brake energy recovery, are often not fully optimized in actual operations, failing to maximize recovery efficiency and wasting potential energy resources. Therefore, a more intelligent and dynamic range optimization method is needed to adapt to the range challenges of unmanned vehicles in complex environments. Summary of the invention

[0004] Based on the above technical problems, the present invention provides a method and system for optimizing the endurance of unmanned vehicles, which monitors environmental parameters in real time, uses LSTM to predict power demand, dynamically adjusts the driving mode to recover energy, and calculates the optimal control strategy to allocate energy, aiming to improve endurance.

[0005] The present invention provides a method for optimizing the cruising range of an unmanned vehicle, the method comprising:

[0006] Step S1: Real-time monitoring of environmental parameters around the vehicle;

[0007] Step S2: constructing a vehicle driving force model and performing power demand prediction;

[0008] Step S3: dynamically adjusting the vehicle's driving mode to perform energy recovery according to the prediction results;

[0009] Step S4: Calculate the optimal control strategy and perform energy distribution.

[0010] Optionally, the constructing of the vehicle driving force model and performing power demand prediction specifically includes:

[0011] F total=m·a+F grav +F air +F roll +F fric

[0012] F grav =m·g·(sin(θ)-μ dymamic ·cos(θ)

[0013] F air =1 / 2×C d ×ρ×A×(v+v wind ) 2 ×cos(σ)

[0014] F roll =C roll ·NF fric =μ dymamic ·N

[0015] In the formula, F total is the total driving force requirement; a is the vehicle acceleration, m is the vehicle mass; F grav is the gravity component, g is the acceleration due to gravity, θ is the slope angle, μ dymamic is the dynamic friction coefficient; F air is the air resistance, C d is the air resistance coefficient of the vehicle, ρ is the air density, A is the frontal windward area of ​​the vehicle, v is the speed of the vehicle relative to the ground, wind is the relative wind speed, and σ is the angle between the vehicle's forward direction and the wind direction; F roll is the rolling resistance, C roll is the rolling resistance coefficient, N is the vertical force of the wheel on the ground; F fric is the road friction;

[0016] Apply LSTM to predict power demand, including:

[0017] Collect historical driving data, including vehicle status, environmental parameters, and corresponding driving force requirements; clean the data and process missing values;

[0018] Define the LSTM network structure, including input layer, hidden layer, and output layer; the input layer receives a set of feature vectors, including the current environmental parameters and vehicle status, as well as historical information of the previous time steps; build a multi-layer LSTM network, with a fully connected layer after the LSTM network to map the hidden state to the power demand prediction value t is the time step, i is the number of future time steps predicted, a linear activation function is used, and the power demand is a continuous value;

[0019] Divide the training set, validation set and test set, use the back propagation algorithm combined with gradient descent to train the model; set the loss function to mean square error; adjust the network parameters;

[0020] Use the trained LSTM model to input the current environmental monitoring data and vehicle status to predict the power demand curve for the next few time steps; combine the vehicle driving force model, use the LSTM prediction results as input, and refine the power distribution strategy;

[0021] Using the prediction results, the vehicle's driving strategy is adjusted through particle swarm optimization to maximize the efficiency of kinetic energy recovery, including:

[0022] Obtaining the power demand forecast value on the future driving path from the LSTM network

[0023] Based on the prediction results, an optimization model is constructed with the goal of minimizing external energy consumption while maximizing kinetic energy recovery; let the total recovered energy be E rec , the external energy consumption is E exp , then the objective function is expressed as maximizing E rec -E exp or minimize -E rec +E exp ;

[0024] Particle swarm optimization is used to solve and initialize candidate solutions for driving strategies. Each particle represents a driving speed adjustment scheme and / or route planning scheme, and the position and speed of the particle represent the specific parameters of the driving strategy. For each driving strategy, the estimated kinetic energy recovery and energy consumption in the entire prediction interval are calculated to evaluate the pros and cons of the strategy. Based on the individual optimal solution pBest and the global optimal solution gBest of the particle, the position and speed of each particle are updated. The above steps are repeated until the number of iterations reaches the preset number or the set threshold has converged.

[0025] The optimal driving strategy or a group of strategies close to the optimal one is selected to perform corresponding speed adjustment and / or route planning.

[0026] Optionally, dynamically adjusting the vehicle driving mode according to the prediction result to perform energy recovery specifically includes:

[0027] The recovered kinetic energy is expressed as

[0028] P 回收 =η·m·g·sin(θ)·v+1 / 2·ρ·C d ·A·v 3

[0029] Where P 回收is the recovery power; η is the recovery efficiency; m is the vehicle mass; g is the acceleration of gravity; θ is the slope angle; C d is the air resistance coefficient of the vehicle; ρ is the air density; A is the frontal windward area of ​​the vehicle; v is the speed of the vehicle relative to the ground;

[0030] Reduce the air resistance coefficient C d , expressed as

[0031] C d =f(α, β, wind, φ)

[0032] Where α and β are vehicle posture angles; φ is the wind direction angle;

[0033] During vehicle braking, the motor operating point is dynamically adjusted according to the current vehicle speed, braking force, and battery status inputs based on the real-time braking conditions, expressed as

[0034] P rec =η max ·V batt I opt ·f(I b , V batt ,θ control )

[0035] Where P rec is the recovery power; η max is the maximum efficiency; V batt is the battery voltage; I opt To optimize the current; I b is the braking instantaneous current; θ control is the control parameter dynamically adjusted according to the working conditions; f is the function of the motor characteristics;

[0036] The PID controller is used to adjust the feedback current in real time. The control equation is expressed as

[0037] e(t)=I target -I actual (t)

[0038]

[0039] Where, e(t) is the error signal; I target is the target current; I actual (t) is the actual measured current; u(t) is the control output; K p , K i , K d They are the proportional, integral and differential coefficients of the PID controller respectively;

[0040] According to the current state of the battery and the real-time demand for braking energy recovery, the output power is adjusted, expressed as Pcharge =min(P rec , P max_charge )

[0041] Where P charge is the charging power; P max_charge For maximum safe charging power.

[0042] Optionally, the calculating of the optimal control strategy and performing energy distribution specifically includes:

[0043] By predicting the vehicle state in the next few time steps and calculating the corresponding energy allocation instruction u in real time * (k), expressed as

[0044]

[0045] In the formula, u * (k) is the control variable at time k; y(I|k) is the output at future time I predicted based on the current state; r(I) is the target reference value; Q and R are weight matrices that adjust the prediction error and the cost of the control action, respectively;

[0046] The allocation strategy is optimized and the optimization strategy is generated according to the control strategy results and real-time environmental information, which is expressed as

[0047]

[0048] Where P opt (t) is the optimized driving force or energy distribution strategy at time t; P req (t+Δt) is the predicted power demand after Δt in the future; P(t+Δt) is the power output predicted according to the current strategy; α and β are weight factors, representing the priority of power demand matching and energy loss penalty, respectively; E loss (P(t) is the additional energy loss due to improper energy distribution;

[0049] Dynamically adjust the energy allocation strategy. The fuzzy logic controller generates control instructions based on the vehicle status, battery, and predicted driving force requirements, and converts them into specific control signals. The fuzzy reasoning process is expressed as

[0050]

[0051] Where u(t) is the control output; μA j ,μB j The input variables x are j ,y j Membership function; C j is the conclusion part of the corresponding rule; N is the total number of rules.

[0052] The present invention also provides an unmanned vehicle range optimization system, the system comprising:

[0053] Environmental parameter monitoring module, used to monitor the environmental parameters around the vehicle in real time;

[0054] The driving model building module is used to build a vehicle driving force model and predict power demand;

[0055] The driving energy recovery module is used to dynamically adjust the vehicle's driving mode and perform energy recovery based on the prediction results;

[0056] The control strategy generation module is used to calculate the optimal control strategy and perform energy distribution.

[0057] Optionally, the driving model construction module specifically includes:

[0058] F total =m·a+F grav +F air +F rol l+F fric

[0059] F grav =m·g·(sin(θ)-μ dymamic ·cos(θ)

[0060] F air =1 / 2×C d ×ρ×A×(v+v wind ) 2 ×cos(σ)

[0061] F roll =C roll ·NF fric =μ dymamic ·N

[0062] In the formula, F total is the total driving force requirement; a is the vehicle acceleration, m is the vehicle mass; F grav is the gravity component, g is the acceleration due to gravity, θ is the slope angle, μ dymamic is the dynamic friction coefficient; F air is the air resistance, C d is the air resistance coefficient of the vehicle, ρ is the air density, A is the frontal windward area of ​​the vehicle, v is the speed of the vehicle relative to the ground, wind is the relative wind speed, and σ is the angle between the vehicle's forward direction and the wind direction; F roll is the rolling resistance, C roll is the rolling resistance coefficient, N is the vertical force of the wheel on the ground; F fric is the road friction;

[0063] Apply LSTM to predict power demand, including:

[0064] Collect historical driving data, including vehicle status, environmental parameters, and corresponding driving force requirements; clean the data and process missing values;

[0065] Define the LSTM network structure, including input layer, hidden layer, and output layer; the input layer receives a set of feature vectors, including the current environmental parameters and vehicle status, as well as historical information of the previous time steps; build a multi-layer LSTM network, with a fully connected layer after the LSTM network to map the hidden state to the power demand prediction value t is the time step, i is the number of future time steps predicted, a linear activation function is used, and the power demand is a continuous value;

[0066] Divide the training set, validation set and test set, use the back propagation algorithm combined with gradient descent to train the model; set the loss function to mean square error; adjust the network parameters;

[0067] Use the trained LSTM model to input the current environmental monitoring data and vehicle status to predict the power demand curve for the next few time steps; combine the vehicle driving force model, use the LSTM prediction results as input, and refine the power distribution strategy;

[0068] Using the prediction results, the vehicle's driving strategy is adjusted through particle swarm optimization to maximize the efficiency of kinetic energy recovery, including:

[0069] Obtaining the power demand forecast value on the future driving path from the LSTM network

[0070] Based on the prediction results, an optimization model is constructed with the goal of minimizing external energy consumption while maximizing kinetic energy recovery; let the total recovered energy be E rec , the external energy consumption is E exp , then the objective function is expressed as maximizing E rec -E exp or minimize -E rec +E exp ;

[0071] Particle swarm optimization is used to solve and initialize candidate solutions for driving strategies. Each particle represents a driving speed adjustment scheme and / or route planning scheme, and the position and speed of the particle represent the specific parameters of the driving strategy. For each driving strategy, the estimated kinetic energy recovery and energy consumption in the entire prediction interval are calculated to evaluate the pros and cons of the strategy. Based on the individual optimal solution pBest and the global optimal solution gBest of the particle, the position and speed of each particle are updated. The above steps are repeated until the number of iterations reaches the preset number or the set threshold has converged.

[0072] The optimal driving strategy or a group of strategies close to the optimal one is selected to perform corresponding speed adjustment and / or route planning.

[0073] Optionally, the driving energy recovery module specifically includes:

[0074] The recovered kinetic energy is expressed as

[0075] P 回收 =η·m·g·sin(θ)·v+1 / 2·ρ·C d ·A·v 3

[0076] Where P 回收 is the recovery power; η is the recovery efficiency; m is the vehicle mass; g is the acceleration of gravity; θ is the slope angle; C d is the air resistance coefficient of the vehicle; ρ is the air density; A is the frontal windward area of ​​the vehicle; v is the speed of the vehicle relative to the ground;

[0077] Reduce the air resistance coefficient C d , expressed as

[0078] C d =f(α, β, wind, φ)

[0079] Where α and β are vehicle posture angles; φ is the wind direction angle;

[0080] During vehicle braking, the motor operating point is dynamically adjusted according to the current vehicle speed, braking force, and battery status inputs based on the real-time braking conditions, expressed as

[0081] P rec =η max ·V batt I opt ·f(I b , V batt ,θ control )

[0082] Where P rec is the recovery power; η max is the maximum efficiency; V batt is the battery voltage; I opt To optimize the current; I b is the braking instantaneous current; θ control is the control parameter dynamically adjusted according to the working conditions; f is the function of the motor characteristics;

[0083] The PID controller is used to adjust the feedback current in real time. The control equation is expressed as

[0084] e(t)=I target -I actual(t)

[0085]

[0086] Where, e(t) is the error signal; I target is the target current; I actual (t) is the actual measured current; u(t) is the control output; K p , K i , K d They are the proportional, integral and differential coefficients of the PID controller respectively;

[0087] According to the current state of the battery and the real-time demand for braking energy recovery, the output power is adjusted, which is expressed as

[0088] P charge =min(P rec , P max_charge )

[0089] Where P charge is the charging power; P max_charge For maximum safe charging power.

[0090] Optionally, the control strategy generation module specifically includes:

[0091] By predicting the vehicle state in the next few time steps and calculating the corresponding energy allocation instruction u in real time * (k), expressed as

[0092]

[0093] In the formula, u * (k) is the control variable at time k; y(I|k) is the output at future time I predicted based on the current state; r(I) is the target reference value; Q and R are weight matrices that adjust the prediction error and the cost of the control action, respectively;

[0094] The allocation strategy is optimized and the optimization strategy is generated according to the control strategy results and real-time environmental information, which is expressed as

[0095]

[0096] Where P opt (t) is the optimized driving force or energy distribution strategy at time t; P req (t+Δt) is the predicted power demand after Δt in the future; P(t+Δt) is the power output predicted according to the current strategy; α and β are weight factors, representing the priority of power demand matching and energy loss penalty, respectively; E loss (P(t) is the additional energy loss due to improper energy distribution;

[0097] Dynamically adjust the energy allocation strategy. The fuzzy logic controller generates control instructions based on the vehicle status, battery, and predicted driving force requirements, and converts them into specific control signals. The fuzzy reasoning process is expressed as

[0098]

[0099] Where u(t) is the control output; μA j ,μB j The input variables x are j ,y j Membership function; C i is the conclusion part of the corresponding rule; N is the total number of rules.

[0100] Compared with the prior art, the present invention has the following beneficial effects:

[0101] The present invention combines the LSTM model with the vehicle physical model to achieve high-precision prediction of future power demand, so that the energy management strategy has forward-looking and dynamic adjustment capabilities, and can effectively cope with various driving scenarios; the efficient recovery of vehicle kinetic energy during downhill gliding and braking is utilized, combined with the dynamic adjustment of the motor operating point, to significantly improve the energy recovery efficiency and extend the vehicle's endurance; the introduction of the fuzzy logic controller allows the energy allocation strategy to be adjusted in real time according to multi-dimensional inputs, balancing immediate needs with future expectations, and improving the flexibility and adaptability of the strategy; through the dynamic power control of the DC / DC converter, the battery safety during the charging process is ensured, overcharging and over-discharging are avoided, and the battery life is extended. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 A flow chart of a method for optimizing the range of an unmanned vehicle according to the present invention;

[0103] Figure 2 This is a structural diagram of an unmanned vehicle range optimization system of the present invention. DETAILED DESCRIPTION

[0104] The present invention is further described below in conjunction with specific implementation cases and drawings, but the present invention is not limited to these embodiments.

[0105] Example 1

[0106] like Figure 1 As shown, the present invention discloses a method for optimizing the cruising range of an unmanned vehicle, the method comprising:

[0107] Step S1: Real-time monitoring of environmental parameters around the vehicle.

[0108] Step S2: Construct a vehicle driving force model and perform power demand prediction.

[0109] Step S3: According to the prediction results, dynamically adjust the vehicle's driving mode to perform energy recovery.

[0110] Step S4: Calculate the optimal control strategy and perform energy distribution.

[0111] The following is a detailed discussion of each step:

[0112] Step S1: Real-time monitoring of environmental parameters around the vehicle, including but not limited to wind speed, slope change, road conditions, etc.

[0113] Step S1 specifically includes:

[0114] Wind speed sensors use ultrasonic technology or LiDAR technology to monitor the wind speed and direction changes around the vehicle in real time; ultrasonic wind speed sensors calculate wind speed by emitting ultrasonic pulses and measuring their round-trip time in the air, because wind speed affects the propagation speed of sound waves. LiDAR can not only measure wind speed but also provide wind direction information by emitting laser pulses and analyzing the time difference and intensity distribution of the return signal. The collected wind speed data is processed by filtering and correction algorithms to eliminate errors caused by instantaneous eddy currents and the equipment's own vibrations. The data collection frequency is not less than 10Hz to ensure real-time and accuracy.

[0115] The inertial measurement unit (IMU) integrates a three-axis accelerometer and gyroscope, which can measure the linear acceleration and angular velocity of the vehicle along three axes (X, Y, and Z). The accelerometer is used to detect the acceleration and deceleration of the vehicle's linear motion, and the gyroscope is used to determine the vehicle's rotation angle and angular velocity, and then calculate the vehicle's absolute posture; it continuously monitors the vehicle's pitch angle and roll angle to determine whether the vehicle is going uphill, downhill, or tilting.

[0116] The road condition sensor uses radar sensors to send electromagnetic waves and receive reflected waves, analyzing the time delay and frequency change of the echo to determine the distance and material characteristics of the road ahead. LiDAR (laser radar) emits laser pulses and receives reflected light to generate high-resolution three-dimensional images that accurately depict the road profile and surface features. Vibration sensors are installed in the vehicle suspension system and indirectly evaluate the roughness and wetness of the road by monitoring the vibration frequency and amplitude when the tire contacts the ground.

[0117] In this embodiment, the camera and the infrared sensor work together. The camera captures visual images and analyzes visual features such as road color and texture, while the infrared sensor detects road temperature and indirectly reflects the road humidity condition (a wet road reflects less infrared light than a dry road). The image processing algorithm analyzes camera data, identifies road material (such as asphalt, concrete, gravel) and surface condition (wet, dry, icy, waterlogged, etc.), and the infrared data assists in determining the degree of road wetness.

[0118] Step S2: Construct a vehicle driving force model and perform power demand prediction.

[0119] Step S2 specifically includes:

[0120] F total =m·a+F grav +F air +F roll +F fric

[0121] F grav =m·g·(sin(θ)-μ dymamic ·cos(θ)

[0122] F air =1 / 2×C d ×ρ×A×(v+v wind) 2 ×cos(σ)

[0123] F roll =C roll ·NF fric =μ dymamic ·N

[0124] In the formula, F tota , is the total driving force requirement; m·a is the force caused by inertia during vehicle acceleration or deceleration, a is the vehicle acceleration, and m is the vehicle mass; F grav is the gravity component, considering the influence of slope and friction on the driving force, g is the acceleration of gravity, θ is the slope angle, μ dymamic is the dynamic friction coefficient, which reflects the additional resistance of the vehicle going uphill and downhill under different road conditions; when the vehicle is stationary or at low speed, μ dymamic Replaced by the static friction coefficient μ static ; F air is the air resistance, C d is the air resistance coefficient of the vehicle, which depends on the shape of the vehicle, ρ is the air density, which is affected by factors such as temperature, humidity and altitude, A is the frontal area of ​​the vehicle, v is the speed of the vehicle relative to the ground, wind is the relative wind speed (the vector difference between wind speed and vehicle speed), and σ is the angle between the vehicle's forward direction and the wind direction; F roll is the rolling resistance, C roll is the rolling resistance coefficient, N is the vertical force of the wheel on the ground; F fric The friction of the road surface takes into account factors such as wetness and slipperiness.

[0125] In this embodiment, a model is constructed to describe the movement state of the vehicle and the required power consumption under different environmental conditions (such as wind speed, slope, road conditions); it usually includes the interaction between the physical properties of the vehicle (such as mass, drag coefficient, tire friction, etc.) and environmental parameters, and can calculate the driving force or energy consumption required for vehicle driving based on the input real-time environmental data.

[0126] Apply LSTM to predict power demand, including:

[0127] Collect historical driving data, including vehicle status (such as speed, acceleration), environmental parameters (wind speed, slope, road conditions) and corresponding driving force requirements; clean the data, process missing values, standardize or normalize the numerical range to ensure data quality; based on environmental monitoring data, extract meaningful features, such as wind speed fluctuation period, slope slope change rate, etc., and use time-frequency analysis techniques (such as wavelet transform) to enhance the representativeness of the features.

[0128] Define the LSTM network structure, including input layer, hidden layer (multiple LSTM units are stacked to capture long-term dependencies), and output layer; the input layer receives a set of feature vectors, including the current environmental parameters and vehicle status, as well as historical information of the previous time steps; build a multi-layer LSTM network, each layer contains multiple LSTM units, LSTM units use memory cells to store long-term dependency information, forget gates, input gates, and output gates dynamically control the flow of information, which is suitable for capturing the time series characteristics of power demand; finally, a fully connected layer is connected after the LSTM network to map the hidden state to the power demand prediction value is the time step, i is the number of future time steps to be predicted, a linear activation function is used, and the power requirement is a continuous value.

[0129] The training set, validation set and test set are divided, and the model is trained using the back-propagation algorithm combined with gradient descent. An appropriate loss function (mean square error MSE) is set to measure the gap between the predicted value and the actual power demand. The network parameters (such as the number of units, the number of layers, the learning rate, etc.) and regularization methods are adjusted to prevent overfitting and improve generalization ability.

[0130] Use the trained LSTM model, input the current environmental monitoring data and vehicle status, and predict the power demand curve for the next few time steps; combine the vehicle driving force model, use the LSTM prediction results as input, and further refine the power distribution strategy, such as optimizing energy recovery and distribution timing.

[0131] In this embodiment, the LSTM model learns the patterns and trends of power demand based on historical data and can predict future power demand. However, this prediction is based on historical statistical laws and may not fully consider the constraints of physical laws, such as friction, wind resistance and other factors that directly affect the driving force. In order to achieve more precise control and optimization, the prediction results of LSTM are As a reference input or constraint of the driving force model, specifically:

[0132] The driving force demand predicted by LSTM is used as the expected driving force target of the driving force model. On this basis, the driving force model will calculate the actual required driving force F according to the current specific environmental parameters (such as actual wind speed, road friction coefficient, etc.) adjusted , ensuring that physical conditions and actual driving scenarios are met.

[0133] Based on the LSTM prediction and the adjusted driving force calculated by the driving force model, energy distribution can be planned more reasonably. For example, if the LSTM predicts that there will be a large slope rise next, and the driving force model calculation shows that a large amount of driving force is required, the energy management system can schedule in advance to ensure that the battery has enough power to support the climb or increase energy recovery on the previous flat section to prepare for future needs.

[0134] Through this integration, the predictive power of LSTM is combined with the accurate execution of the physical model, which not only takes advantage of data-driven prediction, but also ensures the physical feasibility and efficiency of the control strategy. In this way, the vehicle's control system can not only predict future needs, but also make real-time adjustments that conform to the laws of physics based on these predictions, thereby achieving optimal energy utilization and driving performance under various driving conditions.

[0135] The prediction results are further used to adjust the vehicle's driving strategy (such as speed regulation and route planning) through an optimization algorithm (particle swarm optimization PSO) to maximize the efficiency of kinetic energy recovery, including:

[0136] First, obtain the power demand forecast value on the future driving path from the LSTM network This includes the amount of driving effort the vehicle is expected to require at different time steps, as well as potential energy recovery opportunities based on environmental conditions (such as an upcoming downhill slope).

[0137] Based on the prediction results, a mathematical optimization model is constructed with the goal of minimizing external energy consumption (such as battery power consumption) while maximizing kinetic energy recovery; let the total recovered energy be E rec , the external energy consumption is E exp , then the objective function can be expressed as maximizing E rec -E exp or minimize -E rec +E exp , while meeting constraints such as power requirements, speed limits, and safety.

[0138] Particle swarm optimization (PSO) is used to solve the above optimization problem. A group of particles (i.e., candidate solutions of driving strategies) are initialized. Each particle represents a possible driving speed adjustment scheme and / or route planning scheme. The position and speed of the particle represent the specific parameters of the driving strategy. For each particle (driving strategy), its expected kinetic energy recovery and energy consumption in the entire prediction interval are calculated to evaluate the pros and cons of the strategy. Based on the individual optimal solution pBest and the global optimal solution gBest of the particle, the position and speed of each particle are updated to simulate the foraging behavior of bird flocks in nature and guide the search to converge to a better solution. The above steps are repeated until the stopping criterion is met (such as the number of iterations reaches the preset number or the set threshold has converged).

[0139] The optimal driving strategy (or a group of strategies close to the optimal one) is selected to guide the vehicle to perform corresponding speed adjustment and / or route planning, so as to achieve the predicted efficient kinetic energy recovery in actual driving.

[0140] In this embodiment, if a long downhill slope is predicted ahead, the system will slow down in advance and maximize energy recovery during gravity gliding.

[0141] Step S3: According to the prediction results, dynamically adjust the vehicle's driving mode to perform energy recovery.

[0142] Step S3 specifically includes:

[0143] If the prediction result shows that the vehicle is about to enter a downhill slope, the motor drive power is reduced or stopped to allow the vehicle to slide down naturally due to gravity. At the same time, the motor is converted into a generator mode to efficiently recover the kinetic energy generated by sliding. A slight braking force is applied to control the vehicle speed to avoid safety risks caused by sliding down too fast. At the same time, the motor is used as a generator to maximize the recovery of the kinetic energy generated by the vehicle due to gravity, which is expressed as P 回收 =η·m·g·sin(θ)·v+1 / 2·ρ·Cd·A·v 3

[0144] Where P 回收 is the recovery power; η is the recovery efficiency; m is the vehicle mass; g is the acceleration of gravity; θ is the slope angle; C d is the air resistance coefficient of the vehicle; ρ is the air density; A is the frontal area of ​​the vehicle; v is the speed of the vehicle relative to the ground.

[0145] In an environment with high wind resistance, the air resistance coefficient C is reduced by adjusting the active aerodynamic components at the front of the vehicle (such as variable fenders, air intake grille) and the body posture. d , while optimizing the driving direction and speed to reduce additional energy consumption, the wind resistance adjustment is calculated in real time based on the wind speed and direction, expressed as

[0146] C d =f(α, β, wind, φ)

[0147] Where α and β are vehicle attitude angles; φ is the wind direction angle.

[0148] During vehicle braking, the motor’s operating point is dynamically adjusted based on the vehicle’s current speed, braking force, and battery status, according to real-time braking conditions, to ensure that the motor operates within the maximum power point tracking (MPPT) region, expressed as

[0149] P rec =η max ·V batt I opt ·f(I b , V batt ,θ control )

[0150] Where P rec is the recovery power; η max is the maximum efficiency; V batt is the battery voltage; I opt To optimize the current; I b is the braking instantaneous current; θ control is the control parameter dynamically adjusted according to the working conditions; f is the function of the motor characteristics.

[0151] Adopt PID controller or fuzzy logic control to adjust the feedback current in real time so that the motor works near the efficiency peak to ensure high efficiency conversion. The control equation is expressed as

[0152] e(t)=I target -I actual (t)

[0153]

[0154] Where, e(t) is the error signal; I target is the target current; I actual (t) is the actual measured current; u(t) is the control output; K p , K i , K d are the proportional, integral and differential coefficients of the PID controller respectively.

[0155] The DC / DC converter automatically adjusts the output power according to the current state of the battery and the real-time demand for braking energy recovery to ensure that the charging rate does not exceed the safety limit of the battery, which is expressed as

[0156] P charge =min(P rec , P max_charge )

[0157] Where P charge is the charging power; P max_charge For maximum safe charging power.

[0158] In this embodiment, the dynamic driving mode includes downhill section processing and wind resistance optimization, maximizing the use of the motor as a generator to recover the kinetic energy generated by the vehicle's gravity; optimizing the motor controller to improve the power generation efficiency of the motor under different braking intensities, involving precise adjustment of the motor back electromotive force and current feedback control, and instantaneous power matching to reduce energy loss and increase recovery; real-time monitoring of the motor feedback current and precise matching with the battery charging status to ensure efficient operation of the motor while protecting the battery from damage, and achieving fine energy management through dynamic power control of the DC / DC converter.

[0159] Step S4: Calculate the optimal control strategy and perform energy distribution.

[0160] Step S4 specifically includes:

[0161] By predicting the vehicle state in the next few time steps and calculating the corresponding energy allocation instruction u in real time * (k) to achieve accurate matching of energy flow and vehicle power demand, expressed as

[0162]

[0163] In the formula, u * (k) is the control variable (energy allocation instruction) at the kth moment; y(I|k) is the output (energy state) at the future moment I predicted based on the current state; r(I) is the target reference value (desired energy allocation); Q and R are weight matrices that adjust the prediction error and the cost of the control action, respectively.

[0164] The above formula means predicting the future state y(I|k) based on the current state and comparing it with the target reference value r(I). By minimizing the prediction error and controlling the action cost, the strategy of using the overall energy is implemented.

[0165] The above allocation strategy is optimized and the optimization strategy is generated according to the control strategy results and real-time environmental information, which is expressed as

[0166]

[0167] Where P opt (t) is the optimized driving force or energy distribution strategy at time t; P req(t+Δt) is the predicted power demand after Δt in the future; P(t+Δt) is the power output predicted according to the current strategy; α and β are weight factors, representing the priority of power demand matching and energy loss penalty, respectively; E loss (P(t) is the additional energy loss due to improper energy distribution.

[0168] Dynamically adjust the energy allocation strategy. The fuzzy logic controller generates control instructions based on factors such as the current vehicle state, battery SOC, and predicted driving force demand. The fuzzy rule set defines logical relationships such as "if the SOC is low and the driving force demand is large, increase the energy recovery strength", which is converted into specific control signals. The fuzzy reasoning process is expressed as

[0169]

[0170] Where u(t) is the control output; μA j ,μB j The input variables x are j ,y j Membership function; C i is the conclusion part of the corresponding rule; N is the total number of rules.

[0171] In this embodiment, firstly, by predicting the future state and calculating the optimal control strategy, a goal orientation is provided for the subsequent steps; it uses the vehicle dynamics model and the estimated future state to ensure that the energy allocation strategy is forward-looking, which is the planning stage in the entire intelligent management framework. Then, on this basis, according to the predicted results and real-time environmental information, a more refined optimization strategy is generated, which is the concretization and implementation of the strategy, focusing on the balance between immediate needs and future expectations, and reflecting the dynamic adjustment and real-time response characteristics of the strategy; finally, the introduction of energy distribution further enhances the execution efficiency and adaptability of the strategy. For example, fuzzy logic control can make decisions quickly based on multi-dimensional inputs and adjust the real-time distribution of energy; this part is the key to realize the transformation from strategy to actual action, ensuring that the control instructions are both accurate and flexible and can cope with complex and changeable driving conditions.

[0172] Example 2

[0173] like Figure 2 As shown, the present invention discloses an unmanned vehicle endurance optimization system, the system comprising:

[0174] The environmental parameter monitoring module 10 is used to monitor the environmental parameters around the vehicle in real time.

[0175] The driving model building module 20 is used to build a vehicle driving force model and perform power demand prediction.

[0176] The driving energy recovery module 30 is used to dynamically adjust the driving mode of the vehicle according to the prediction result to perform energy recovery.

[0177] The control strategy generation module 40 is used to calculate the optimal control strategy and perform energy distribution.

[0178] As an optional implementation mode, the driving model building module 20 of the present invention specifically includes:

[0179] F total =m·a+F grav +F air +F roll +F fric

[0180] F grav =m·g·(sin(θ)-μ dymamic ·cos(θ)

[0181] F air =1 / 2×C d ×ρ×A×(v+v wind ) 2 ×cos(ρ)

[0182] F roll =C roll ·NF fric =μ dymamic ·N

[0183] In the formula, F total is the total driving force requirement; a is the vehicle acceleration, m is the vehicle mass; F grav is the gravity component, g is the acceleration due to gravity, θ is the slope angle, μ dymamic is the dynamic friction coefficient; F air is the air resistance, C d is the air resistance coefficient of the vehicle, ρ is the air density, A is the frontal windward area of ​​the vehicle, v is the speed of the vehicle relative to the ground, wind is the relative wind speed, and σ is the angle between the vehicle's forward direction and the wind direction; F roll is the rolling resistance, C roll is the rolling resistance coefficient, N is the vertical force of the wheel on the ground; F fric The friction of the road surface.

[0184] Apply LSTM to predict power demand, including:

[0185] Collect historical driving data, including vehicle status, environmental parameters, and corresponding driving force requirements; clean the data and process missing values.

[0186] Define the LSTM network structure, including input layer, hidden layer, and output layer; the input layer receives a set of feature vectors, including the current environmental parameters and vehicle status, as well as historical information of the previous time steps; build a multi-layer LSTM network, with a fully connected layer after the LSTM network to map the hidden state to the power demand prediction value t is the time step, i is the number of future time steps predicted, a linear activation function is used, and the power demand is a continuous value.

[0187] Divide the training set, validation set and test set, use the back propagation algorithm combined with gradient descent to train the model; set the loss function to mean square error; adjust the network parameters.

[0188] Use the trained LSTM model, input the current environmental monitoring data and vehicle status, and predict the power demand curve for the next few time steps; combine the vehicle driving force model, use the LSTM prediction results as input, and refine the power distribution strategy.

[0189] Using the prediction results, the vehicle's driving strategy is adjusted through particle swarm optimization to maximize the efficiency of kinetic energy recovery, including:

[0190] Obtaining the power demand forecast value on the future driving path from the LSTM network

[0191] Based on the prediction results, an optimization model is constructed with the goal of minimizing external energy consumption while maximizing kinetic energy recovery; let the total recovered energy be E rec , the external energy consumption is E exp , then the objective function is expressed as maximizing E rec -E exp or minimize -E rec +E exp .

[0192] Particle swarm optimization is used to solve the candidate solutions of the driving strategy. Each particle represents a driving speed adjustment scheme and / or route planning scheme, and the position and speed of the particle represent the specific parameters of the driving strategy. For each driving strategy, the estimated kinetic energy recovery and energy consumption in the entire prediction interval are calculated to evaluate the pros and cons of the strategy. Based on the individual optimal solution pBest and the global optimal solution gBest of the particle, the position and speed of each particle are updated. The above steps are repeated until the number of iterations reaches the preset number or the set threshold has converged.

[0193] The optimal driving strategy or a group of strategies close to the optimal one is selected to perform corresponding speed adjustment and / or route planning.

[0194] As an optional implementation, the driving energy recovery module 30 of the present invention specifically includes:

[0195] The recovered kinetic energy is expressed as

[0196] P 回收 =η·m·g·sin(θ)·v+1 / 2·ρ·C d ·A·v 3

[0197] Where P 回收 is the recovery power; η is the recovery efficiency; m is the vehicle mass; g is the acceleration of gravity; θ is the slope angle; C d is the air resistance coefficient of the vehicle; ρ is the air density; A is the frontal area of ​​the vehicle; v is the speed of the vehicle relative to the ground.

[0198] Reduce the air resistance coefficient C d , expressed as

[0199] C d =f(α, β, wind, φ)

[0200] Where α and β are vehicle attitude angles; φ is the wind direction angle.

[0201] During vehicle braking, the motor operating point is dynamically adjusted according to the current vehicle speed, braking force, and battery status inputs based on the real-time braking conditions, expressed as

[0202] P rec =η max ·V batt I opt ·f(I b , V batt ,θ control )

[0203] Where P rec is the recovery power; η max is the maximum efficiency; V batt is the battery voltage; I opt To optimize the current; I b is the braking instantaneous current; θ control is the control parameter dynamically adjusted according to the working conditions; f is the function of the motor characteristics.

[0204] The PID controller is used to adjust the feedback current in real time. The control equation is expressed as

[0205] e(t)=I target -I actual (t)

[0206]

[0207] Where, e(t) is the error signal; I target is the target current; I actual(t) is the actual measured current; u(t) is the control output; K p , K i , K d are the proportional, integral and differential coefficients of the PID controller respectively.

[0208] According to the current state of the battery and the real-time demand for braking energy recovery, the output power is adjusted, which is expressed as

[0209] P charge =min(P rec , P max_charge )

[0210] Where P charge is the charging power; P max_charge For maximum safe charging power.

[0211] As an optional implementation mode, the control strategy generation module 40 of the present invention specifically includes:

[0212] By predicting the vehicle state in the next few time steps and calculating the corresponding energy allocation instruction u in real time * (k), expressed as

[0213]

[0214] In the formula, u * (k) is the control variable at the kth moment; y(I|k) is the output at the future moment I predicted based on the current state; r(I) is the target reference value; Q and R are weight matrices, which adjust the prediction error and the cost of the control action respectively.

[0215] The allocation strategy is optimized and the optimization strategy is generated according to the control strategy results and real-time environmental information, which is expressed as

[0216]

[0217] Where P opt (t) is the optimized driving force or energy distribution strategy at time t; P req (t+Δt) is the predicted power demand after Δt in the future; P(t+Δt) is the power output predicted according to the current strategy; α and β are weight factors, representing the priority of power demand matching and energy loss penalty, respectively; E loss (P(t) is the additional energy loss due to improper energy distribution.

[0218] Dynamically adjust the energy allocation strategy, and the fuzzy reasoning process is expressed as

[0219]

[0220] Where u(t) is the control output; μA j ,μB j The input variables x are j ,y j Membership function; C j is the conclusion part of the corresponding rule; N is the total number of rules.

[0221] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for optimizing the range of an unmanned vehicle, characterized in that: The method comprises: Step S1: real-time monitoring of environmental parameters around the vehicle; Step S2: constructing a vehicle driving force model and performing power demand prediction; Step S3: dynamically adjusting the vehicle's driving mode to perform energy recovery according to the prediction results; Step S4: Calculate the optimal control strategy and perform energy distribution.

2. The method for optimizing the range of an unmanned vehicle according to claim 1, characterized in that: The constructing of the vehicle driving force model and performing power demand prediction specifically includes: F total =m·a+F grav +F air +F roll +F fric F grav =m·g·(sin(θ)-μ) dymamic ·cos(θ)) F air =1 / 2×C d ×ρ×A×(v+v wind ) 2 ×cos(σ) F roll =C roll ·N F fric =μ dymamic ·N In the formula, F total is the total driving force requirement; a is the vehicle acceleration, m is the vehicle mass; F grav is the gravity component, g is the acceleration due to gravity, θ is the slope angle, μ dymamic is the dynamic friction coefficient; F air is the air resistance, C d is the air resistance coefficient of the vehicle, ρ is the air density, A is the frontal windward area of ​​the vehicle, v is the speed of the vehicle relative to the ground, wind is the relative wind speed, and σ is the angle between the vehicle's forward direction and the wind direction; F roll is the rolling resistance, C roll is the rolling resistance coefficient, N is the vertical force of the wheel on the ground; F fric is the road friction; Apply LSTM to predict power demand, including: Collect historical driving data, including vehicle status, environmental parameters, and corresponding driving force requirements; clean the data and process missing values; Define the LSTM network structure, including input layer, hidden layer, and output layer; the input layer receives a set of feature vectors, including the current environmental parameters and vehicle status, as well as historical information of the previous time steps; build a multi-layer LSTM network, with a fully connected layer after the LSTM network to map the hidden state to the power demand prediction value is the time step, i is the number of future time steps predicted, a linear activation function is used, and the power demand is a continuous value; Divide the training set, validation set and test set, use the back propagation algorithm combined with gradient descent to train the model; set the loss function to mean square error; adjust the network parameters; Use the trained LSTM model to input the current environmental monitoring data and vehicle status to predict the power demand curve for the next few time steps; combine the vehicle driving force model, use the LSTM prediction results as input, and refine the power distribution strategy; Using the prediction results, the vehicle's driving strategy is adjusted through particle swarm optimization to maximize the efficiency of kinetic energy recovery, including: Obtaining the power demand forecast value on the future driving path from the LSTM network Based on the prediction results, an optimization model is constructed with the goal of minimizing external energy consumption while maximizing kinetic energy recovery; let the total recovered energy be E rec , the external energy consumption is E exp , then the objective function is expressed as maximizing E rec -E exp or minimize -E rec +E exp ; Particle swarm optimization is used to solve and initialize candidate solutions for driving strategies. Each particle represents a driving speed adjustment scheme and / or route planning scheme, and the position and speed of the particle represent the specific parameters of the driving strategy. For each driving strategy, the estimated kinetic energy recovery and energy consumption in the entire prediction interval are calculated to evaluate the pros and cons of the strategy. Based on the individual optimal solution pBest and the global optimal solution gBest of the particle, the position and speed of each particle are updated. The above steps are repeated until the number of iterations reaches the preset number or the set threshold has converged. The optimal driving strategy or a group of strategies close to the optimal one is selected to perform corresponding speed adjustment and / or route planning.

3. The method for optimizing the range of an unmanned vehicle according to claim 1, characterized in that: The method of dynamically adjusting the vehicle driving mode according to the prediction result to perform energy recovery specifically includes: The recovered kinetic energy is expressed as P 回收 =η·m·g·sin(θ)·v+1 / 2·ρ·C d ·A·v 3 Where P 回收 is the recovery power; η is the recovery efficiency; m is the vehicle mass; g is the gravitational acceleration; θ is the slope angle; C d is the air resistance coefficient of the vehicle; ρ is the air density; A is the frontal windward area of ​​the vehicle; v is the speed of the vehicle relative to the ground; Reduce the air resistance coefficient C d , expressed as C d =f(α,β,wind,φ) Where α and β are vehicle posture angles; φ is the wind direction angle; During vehicle braking, the motor operating point is dynamically adjusted according to the current vehicle speed, braking force, and battery status inputs based on the real-time braking conditions, expressed as P rec =the max ·V batt ·I opt ·f(I b ,V batt ,the control ) Where P rec is the recovery power; η max is the maximum efficiency; V batt is the battery voltage; I opt To optimize the current; I b is the braking instantaneous current; θ control is the control parameter dynamically adjusted according to the working conditions; f is the function of the motor characteristics; The PID controller is used to adjust the feedback current in real time. The control equation is expressed as e(t)=I target -I actual (t) Where, e(t) is the error signal; I target is the target current; I actual (t) is the actual measured current; u(t) is the control output; K p , K i , K d They are the proportional, integral and differential coefficients of the PID controller respectively; According to the current state of the battery and the real-time demand for braking energy recovery, the output power is adjusted, which is expressed as P charge =min(P rec ,P max_charge ) Where P charge is the charging power; P max_charge For maximum safe charging power.

4. The method for optimizing the range of an unmanned vehicle according to claim 1, characterized in that: The calculating of the optimal control strategy and the energy distribution specifically include: By predicting the vehicle state in the next few time steps, the corresponding energy allocation instruction u is calculated in real time * (k), expressed as In the formula, u * (k) is the control variable at time k; y(I|k) is the output at future time I predicted based on the current state; r(I) is the target reference value; Q and R are weight matrices that adjust the prediction error and the cost of the control action, respectively; The allocation strategy is optimized and the optimization strategy is generated according to the control strategy results and real-time environmental information, which is expressed as Where P opt (t) is the optimized driving force or energy distribution strategy at time t; P req (t+Δt) is the predicted power demand after Δt in the future; P(t+Δt) is the power output predicted according to the current strategy; α and β are weight factors, representing the priority of power demand matching and energy loss penalty, respectively; E loss (P(t) is the additional energy loss due to improper energy distribution; Dynamically adjust the energy allocation strategy. The fuzzy logic controller generates control instructions based on the vehicle status, battery, and predicted driving force requirements, and converts them into specific control signals. The fuzzy reasoning process is expressed as Where u(t) is the control output; μA j ,μB j The input variables x are j ,y j Membership function; C j is the conclusion part of the corresponding rule; N is the total number of rules.

5. An unmanned vehicle range optimization system, characterized in that: The system comprises: Environmental parameter monitoring module, used to monitor the environmental parameters around the vehicle in real time; The driving model building module is used to build a vehicle driving force model and predict power demand; The driving energy recovery module is used to dynamically adjust the vehicle's driving mode and perform energy recovery based on the prediction results; The control strategy generation module is used to calculate the optimal control strategy and perform energy distribution.

6. The unmanned vehicle endurance optimization system according to claim 5, characterized in that: The driving model construction module specifically includes: F total =m·a+F grav +F air +F roll +F fric F grav =m·g·(sin(θ)-μ) dymamic ·cos(θ)) F air =1 / 2×C d ×ρ×A×(v+v wind ) 2 ×cos(σ) F roll =C roll ·N F fric =μ dymamic ·N In the formula, F total is the total driving force requirement; a is the vehicle acceleration, m is the vehicle mass; F grav is the gravity component, g is the acceleration due to gravity, θ is the slope angle, μ dymamic is the dynamic friction coefficient; F air is the air resistance, C d is the air resistance coefficient of the vehicle, ρ is the air density, A is the frontal windward area of ​​the vehicle, is the speed of the vehicle relative to the ground, wind is the relative wind speed, and σ is the angle between the vehicle's forward direction and the wind direction; F roll is the rolling resistance, C roll is the rolling resistance coefficient, N is the vertical force of the wheel on the ground; F fric is the road friction; Apply LSTM to predict power demand, including: Collect historical driving data, including vehicle status, environmental parameters, and corresponding driving force requirements; clean the data and process missing values; Define the LSTM network structure, including input layer, hidden layer, and output layer; the input layer receives a set of feature vectors, including the current environmental parameters and vehicle status, as well as historical information of the previous time steps; build a multi-layer LSTM network, with a fully connected layer after the LSTM network to map the hidden state to the power demand prediction value is the time step, i is the number of future time steps predicted, a linear activation function is used, and the power demand is a continuous value; Divide the training set, validation set and test set, use the back propagation algorithm combined with gradient descent to train the model; set the loss function to mean square error; adjust the network parameters; Use the trained LSTM model to input the current environmental monitoring data and vehicle status to predict the power demand curve for the next few time steps; combine the vehicle driving force model, use the LSTM prediction results as input, and refine the power distribution strategy; Using the prediction results, the vehicle's driving strategy is adjusted through particle swarm optimization to maximize the efficiency of kinetic energy recovery, including: Obtain the power demand prediction value F on the future driving path from the LSTM network t+i ; Based on the prediction results, an optimization model is constructed with the goal of minimizing external energy consumption while maximizing kinetic energy recovery; let the total recovered energy be E rec , the external energy consumption is E exp , then the objective function is expressed as maximizing E rec -E exp or minimize -E rec +E exp ; Particle swarm optimization is used to solve and initialize candidate solutions for driving strategies. Each particle represents a driving speed adjustment scheme and / or route planning scheme, and the position and speed of the particle represent the specific parameters of the driving strategy. For each driving strategy, the estimated kinetic energy recovery and energy consumption in the entire prediction interval are calculated to evaluate the pros and cons of the strategy. Based on the individual optimal solution pBest and the global optimal solution gBest of the particle, the position and speed of each particle are updated. The above steps are repeated until the number of iterations reaches the preset number or the set threshold has converged. The optimal driving strategy or a group of strategies close to the optimal one is selected to perform corresponding speed adjustment and / or route planning.

7. The unmanned vehicle range optimization system according to claim 5, characterized in that: The driving energy recovery module specifically includes: The recovered kinetic energy is expressed as P 回收 =η·m·g·sin(θ)·v+1 / 2·ρ·C d ·A·v 3 Where P 回收 is the recovery power; η is the recovery efficiency; m is the vehicle mass; g is the gravitational acceleration; θ is the slope angle; C d is the air resistance coefficient of the vehicle; ρ is the air density; A is the frontal windward area of ​​the vehicle; v is the speed of the vehicle relative to the ground; Reduce the air resistance coefficient C d , expressed as C d =f(α,β,wind,φ) Where α and β are vehicle posture angles; φ is the wind direction angle; During vehicle braking, the motor operating point is dynamically adjusted according to the current vehicle speed, braking force, and battery status inputs based on the real-time braking conditions, expressed as P rec =the max ·V batt ·I opt ·f(I b ,V batt ,the control ) Where P rec is the recovery power; η max is the maximum efficiency; V batt is the battery voltage; I opt To optimize the current; I b is the braking instantaneous current; θ control is the control parameter dynamically adjusted according to the working conditions; f is the function of the motor characteristics; The PID controller is used to adjust the feedback current in real time. The control equation is expressed as e(t)=I target -I actual (t) Where, e(t) is the error signal; I target is the target current; I actual (t) is the actual measured current; u(t) is the control output; K p , K i , K d They are the proportional, integral and differential coefficients of the PID controller respectively; According to the current state of the battery and the real-time demand for braking energy recovery, the output power is adjusted, which is expressed as P charge =min(P rec ,P max_charge) Where P charge is the charging power; P max_charge For maximum safe charging power.

8. The unmanned vehicle range optimization system according to claim 5, characterized in that: The control strategy generation module specifically includes: By predicting the vehicle state in the next few time steps, the corresponding energy allocation instruction u is calculated in real time * (k), expressed as In the formula, u * (k) is the control variable at time k; y(I|k) is the output at future time I predicted based on the current state; r(I) is the target reference value; Q and R are weight matrices that adjust the prediction error and the cost of the control action, respectively; The allocation strategy is optimized and the optimization strategy is generated according to the control strategy results and real-time environmental information, which is expressed as Where P opt (t) is the optimized driving force or energy distribution strategy at time t; P req (t+Δt) is the predicted power demand after Δt in the future; P(t+Δt) is the power output predicted according to the current strategy; α and β are weight factors, representing the priority of power demand matching and energy loss penalty, respectively; E loss (P(t) is the additional energy loss due to improper energy distribution; Dynamically adjust the energy allocation strategy. The fuzzy logic controller generates control instructions based on the vehicle status, battery, and predicted driving force requirements, and converts them into specific control signals. The fuzzy reasoning process is expressed as Where u(t) is the control output; μA j ,uB j The input variables x are j ,y j Membership function; C j is the conclusion part of the corresponding rule; N is the total number of rules.

Images