Trunk line logistics hybrid heavy truck energy-saving control method considering intelligent driving-human driving mode switching

Abstract

The invention relates to the technical field of intelligent driving, and discloses a trunk line logistics hybrid heavy truck energy-saving control method considering intelligent driving-human driving mode switching. The method comprises the steps of mode switching opportunity optimization based on driving risks and limit energy consumption, intelligent driving predictive vehicle speed planning considering driving scene changes, and intelligent driving-human driving switching hybrid power system energy management hierarchical optimization control. According to the method provided by the invention, the mode switching decision time interval is obtained based on the risk assessment model, the optimization model is established to optimize the switching opportunity, the vehicle speed planning model with the energy consumption cost as the target is established based on the mode switching decision time interval, and the long-period-short-time domain vehicle speed planning rolling optimization strategy is designed by considering the scene interaction strength. According to the method, driving risks, driving scene changes and energy consumption are comprehensively considered, and the problems that a man-machine co-driving intelligent heavy truck is high in driving energy consumption, low in safety, unreasonable in intelligent driving mode vehicle speed planning and the like under urban complex scenes can be solved.

Description

Technical Field

[0001] The present application relates to the field of intelligent driving technology, and in particular to an energy-saving control method for a trunk logistics hybrid heavy truck that takes into account the switching between intelligent driving and human driving modes. Background Art

[0002] In complex urban roads, driving scenarios are changeable. How to achieve efficient and energy-saving driving while ensuring the safe driving of human-machine co-driving trunk logistics hybrid heavy trucks is a severe challenge currently faced. Therefore, it is urgent to study the energy-saving optimization control method of trunk logistics hybrid heavy trucks with intelligent driving-human driving switching for complex and multiple scenarios. By optimizing the intelligent driving-human driving switching that considers the balance between safety and energy consumption, speed planning that adapts to scenario changes, and hybrid system energy management that considers intelligent driving-human driving switching, the safety of human-machine co-driving intelligent heavy trucks can be ensured, and the optimal energy consumption can be achieved to the greatest extent. However, the existing energy-saving control methods for hybrid power systems mainly focus on energy management at the vehicle power system level, which cannot adapt to the complex and changeable characteristics of urban scenarios, and do not consider the energy-saving control of the human-machine co-driving system.

[0003] For example, the invention patent applied for and published on September 20, 2022: Application publication number: CN 115071669A, a hybrid energy management system and method, the method sends vehicle navigation information to the cloud through the vehicle navigation system through the communication unit; the cloud calculates the optimal control synergy value based on the navigation information and transmits it to the communication unit of the vehicle; the communication unit transmits the optimal control synergy value to the vehicle controller; the vehicle controller calculates the distribution power of electric drive and fuel drive based on the optimal control synergy value, and transmits the distribution power to the engine controller and the motor controller for execution; the vehicle controller calculates the distribution relationship between electric drive power and fuel drive power in real time based on the optimal control synergy value, the driver's throttle and brake pedal input and the vehicle hybrid model. This method distributes electric drive and fuel drive power according to the navigation information fitting the vehicle driving condition and speed information, but due to the complex and changeable information of urban driving scenes, it may lead to problems such as poor adaptability of the method and poor energy consumption optimization effect. Summary of the invention

[0004] In order to solve the above problems, the present application provides an energy-saving control method for trunk logistics hybrid heavy-duty trucks that takes into account the switching of intelligent driving and human driving modes, so as to effectively reduce driving energy consumption in scenarios with complex urban road conditions and multi-vehicle interference and interaction.

[0005] In order to achieve the above purpose, the technical solution adopted in this application is as follows:

[0006] The present application provides an energy-saving control method for a trunk logistics hybrid heavy truck considering intelligent driving-human driving mode switching, the method comprising:

[0007] Based on the mode switching process feature data set obtained by driver style recognition and the prediction of vehicle trajectories in the surrounding environment, the risk interval and risk change curve of the mode switching decision are obtained using the driving risk assessment method based on the risk field model. The time interval of the mode switching decision is obtained based on the risk interval and risk change curve of the mode switching decision. The mode switching optimization model based on extreme energy consumption is established by integrating the speed prediction of the human driving mode, and the switching timing is optimized to obtain the optimal switching timing.

[0008] Taking the optimal switching time as the end time point of the intelligent driving mode speed planning, a future vehicle speed planning model considering the dynamic interaction characteristics of the driving scene with energy consumption cost as the optimization target is established; based on the interaction strength of the driving scene, the vehicle speed planning model in different future cycles and time domains is solved by using the rolling optimization of the vehicle speed planning for multi-mode long-cycle-short-time domain to obtain the optimal planned vehicle speed curve;

[0009] According to the optimal planned vehicle speed curve, an approximate optimal SOC trajectory planning model is established and the global optimal SOC reference trajectory is solved; for the intelligent driving mode, the global optimal SOC reference trajectory is used as input, and for the human driving mode, the reference SOC value with the best charging and discharging performance is used as input to perform torque distribution control to minimize fuel consumption. The present application has at least the following beneficial effects:

[0010] (1) This application determines the driving mode switching time interval based on the driver's characteristic data analysis and driving risk prediction, and based on this, establishes an optimization model with the ultimate energy consumption as the goal to optimize the switching timing. The optimal switching timing can reduce the energy consumption of the mode switching process while adapting to the driver's driving style and ensuring driving safety.

[0011] (2) In view of the changing characteristics of driving scenarios in urban environments and taking into account the strength of scene interactions, this application designs a long-period-short-time domain speed planning rolling optimization solution strategy, which makes the speed planning in the intelligent driving mode more reasonable and better able to adapt to the complex urban traffic environment.

[0012] (3) This application considers the characteristics of intelligent driving and human driving modes to establish a hierarchical optimization control method for energy management of a hybrid system with intelligent driving-human driving switching. It adopts hierarchical energy management of upper-level planning and lower-level control for intelligent driving, and directly executes energy management of the control layer for human driving, thereby achieving optimal energy consumption in different driving modes. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A flow chart of an energy-saving control method for a trunk logistics hybrid heavy truck taking into account the switching between intelligent driving and human driving modes according to an embodiment of the present application is shown.

[0014] Figure 2A schematic diagram of generating a driving mode switching decision time interval according to an embodiment of the present application is shown.

[0015] Figure 3 A schematic diagram of the classification of interaction strength levels in complex urban driving scenarios according to an embodiment of the present application is shown.

[0016] Figure 4 A schematic diagram of long-period-short-time domain rolling optimization according to an embodiment of the present application is shown.

[0017] Figure 5 A schematic diagram of a hybrid system energy management hierarchical optimization control method according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0018] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0019] The specific implementation of the present application is further described in detail below in conjunction with the drawings and examples.

[0020] In order to solve the problems of high energy consumption, low safety and unreasonable speed planning of intelligent heavy trucks with human-machine co-driving in complex urban scenarios. The embodiment of the present application provides an energy-saving control method for trunk logistics hybrid heavy trucks that takes into account the switching of intelligent driving-human driving mode. The method can be configured in the vehicle's intelligent terminal (such as the control chip carried by the vehicle) for implementation to achieve intelligent control of the vehicle and thus achieve the purpose of reducing energy consumption. It should be noted that this method can be applied to any vehicle that can perform intelligent driving control, including but not limited to human-machine co-driving intelligent heavy trucks.

[0021] Specifically, see Figure 1 , which is a flow chart of an energy-saving control method for a hybrid heavy-duty truck for trunk logistics considering the switching between intelligent driving and human driving modes. The energy-saving control method for a hybrid heavy-duty truck for trunk logistics considering the switching between intelligent driving and human driving modes includes the following steps S1 to S3, which are described in detail as follows.

[0022] S1. Predictive switching between intelligent driving and human driving modes based on driving risks and energy consumption.

[0023] In this embodiment, the specific process of predictive switching between intelligent driving and human driving modes based on driving risk and energy consumption is as follows: based on the mode switching process feature data set obtained by driver style recognition and the prediction of vehicle trajectories in the surrounding environment, a driving risk assessment method based on a risk field model is used to obtain the risk interval and risk change curve of the mode switching decision; based on the above two, the time interval of the mode switching decision is obtained; finally, the human driving mode speed prediction is integrated to establish a mode switching optimization model based on extreme energy consumption to optimize the switching timing.

[0024] In an exemplary embodiment, the driver mode switching process characteristic data set N a The current driver style is recognized by the perception function unit of the vehicle's built-in intelligent cockpit system. The data set contains the vehicle position coordinates, vehicle speed, environment vehicle position coordinates, and environment vehicle speed, which can be expressed as formula (1):

[0025]

[0026] in, is the driving data of the mode switching process of the jth driver in the driving style of type a, is the driving data of the environment vehicle i, is the driving data of the jth driver.

[0027] In an exemplary embodiment, the surrounding vehicle trajectory prediction F is calculated by a trajectory prediction model based on an environmental attention network EA-Net, and the trajectory prediction result is expressed as formula (2);

[0028]

[0029] Among them, F is the prediction of the surrounding vehicle trajectory, T l is the length of the historical time domain, T g is the interval length between the historical time domain and the future time domain, T p For the future prediction time domain length, f (t) ={x (t) ,y (t)}, f (t) For x (t) is the lateral position coordinate of the vehicle, y (t) is the longitudinal position coordinate of the vehicle.

[0030] In an exemplary embodiment, the specific implementation steps of the driving risk assessment method based on the risk field model include: considering the impact of Z (in this embodiment, Z=8) environmental vehicles covering all directions around the vehicle on the vehicle, calculating the field strength E of the driving risk field generated by the surrounding moving vehicles at the vehicle position j(Equation (3)), and then use the force calculation formula in the field strength to calculate the force on the vehicle, that is, the driving risk risk of the vehicle j (Formula (4)).

[0031]

[0032] Among them, E V_ij is the surrounding vehicle i(x i ,y i ) in the vehicle j(x j ,y j )The risk field strength generated, (x i ,y i ) and (x j ,y j ) are the center of mass coordinates of the surrounding vehicles i and the vehicle j, k1, k2, G are constants greater than 0, R i is the impact factor of road conditions at the surrounding vehicle i, M i is the equivalent mass of the surrounding vehicle i, r ij =(x j -x i ,y j -y i ), represents the distance vector between the surrounding vehicle i and the vehicle j, and its direction is the same as the field strength direction, v i is the speed of surrounding vehicle i, θ i is the speed direction and distance vector r of the surrounding vehicle i ij Angle.

[0033] risk j =E j M j R j exp[-k2v j cos(θ j )](1+D rj ) (4)

[0034] Among them, M j is the equivalent mass of the vehicle, R j is the influencing factor of the road conditions at the vehicle, v j is the vehicle speed, θ j is the direction of the vehicle's speed and the field strength E j The angle between the directions, D rj is the driver risk factor of the vehicle.

[0035] In an exemplary embodiment, the future driving risk change curve is obtained by the following steps: according to the trajectory prediction results, the future position coordinates and speeds of the vehicle and the surrounding eight vehicles are obtained, and the future T of the vehicle is calculated by using the driving risk assessment method based on the risk field model. pThe driving risk at each moment is used to derive the future driving risk change curve.

[0036] In an exemplary embodiment, the risk interval of the mode switching decision is generated by the following steps:

[0037] According to the driving data of the mode switching process of the current driver style category, the driving risk assessment method based on the risk field model is used to calculate the driving risk value of each driver in the driving data set when switching the driving mode. The minimum value is taken as the driving risk value of the mode switching of this type of driving style (Formula (5)), which is the left interval of driving risk for the mode switching decision. The driving risk value of the collision position between the self-vehicle and the surrounding vehicles in the future trajectory is taken as the right interval of driving risk for the mode switching decision.

[0038]

[0039] in, represents the left interval of driving risk for the mode switching decision of a class driver, It represents the driving risk value of the jth driver in category a when switching the driving mode.

[0040] In an exemplary embodiment, Figure 2 As shown in the figure, it is a schematic diagram of generating a driving mode switching decision time interval provided in an embodiment of the present application. The driving mode switching decision time interval is generated by the following steps: The future driving risk change curve that changes over time determines the recommended time interval for the driver to switch driving modes

[0041] In an exemplary embodiment, the human driving mode speed prediction is obtained by using the LSTM recurrent neural network model, and the speed prediction sequence is expressed as Where T e To predict the length of time;

[0042] In an exemplary embodiment, the optimal switching timing T opt The determination is made through the following steps: according to the speed prediction, mode switching time interval constraint and the energy consumption calculation model obtained by fitting, a mode switching optimization model based on the extreme energy consumption is established (Equation (6), Equation (7)), and the optimal switching time T of the driver is determined by solving the model. opt ;

[0043]

[0044] Among them, T is the decision variable switching time, v in is the speed in intelligent driving mode, v huPredict the speed for human driving mode; v(t) is the speed of the vehicle at time t, t0 is the left endpoint of the decision recommendation time interval, J is the energy consumption target value, denergy is the fitted instantaneous energy consumption calculation function, t is time, v is the vehicle speed, and a is the vehicle acceleration.

[0045] S2. Multi-mode intelligent driving predictive speed planning considering changes in driving scenarios.

[0046] In this embodiment, the specific implementation steps of the multi-mode intelligent driving predictive speed planning considering changes in driving scenarios are: considering the dynamic boundary constraints of different driving scenarios, including considering the safety risk constraints of surrounding leading vehicles and considering the traffic efficiency constraints of signal phases, and establishing a future speed planning model that considers the dynamic interaction characteristics of driving scenarios with energy consumption cost as the optimization target; based on the strength of the interaction of driving scenarios, the speed planning model for different future period time domains is solved by using the rolling optimization of the multi-mode long-cycle-short-time domain speed planning to obtain the optimal planned speed curve.

[0047] In an exemplary embodiment, based on the influence of dynamic changes in driving scenarios on the efficiency and SOC of the hybrid power system, a hybrid power system energy consumption model based on a deep neural network considering battery SOC and system efficiency is established, as shown in formula (8):

[0048]

[0049] in, represents the instantaneous energy consumption cost of the powertrain, f DNN It represents the nonlinear mapping function describing the relationship between driving state and optimal energy consumption, v is the vehicle speed, a is the vehicle acceleration, SOC is the battery state of charge, and η is the system efficiency, including the efficiency of the motor, engine and converter.

[0050] In an exemplary embodiment, the steps of establishing a future vehicle speed planning model considering the dynamic interaction characteristics of driving scenarios include: considering the dynamic boundary constraints of different driving scenarios, including considering the safety risk constraints of surrounding leading vehicles and considering the traffic efficiency constraints of signal phases, and establishing a future vehicle speed planning model considering the dynamic interaction characteristics of driving scenarios with energy consumption cost as the optimization target, as shown in equations (9)-(14):

[0051]

[0052] u=a(t),x=[v(t),s(t),SOC(t)] T (10)

[0053]

[0054] Among them, formula (9) represents the energy consumption target, which is the energy consumption integral in the time domain, T0 is the starting time of vehicle speed planning, T opt is the optimal switching time; Formula (10) indicates that the control variable is acceleration, and the state variables are speed, distance and battery power; Formulas (11)-(13) are differential equations of state variables, f NNS Q is a nonlinear mapping function that describes the relationship between driving state and optimal battery current. b is the battery capacity; Formula (14) is the constraint condition, including speed constraint, acceleration constraint, SOC constraint, following distance constraint and traffic light constraint, s p is the distance traveled by the preceding vehicle, δ s For safe distance, It indicates the time that the i-th traffic light has been running in its periodic clock when the vehicle starts. is the travel time required for a vehicle to pass the i-th intersection, expressed as Calculate the passing time of the vehicle in the signal cycle clock, T0 represents the starting point of the vehicle speed planning, x(t) represents the state variable at time t, u(t) represents the control variable at time t, v(t) represents the speed at time t, x represents the state variable set, u represents the control variable set, a(t) represents the acceleration at time t, s(t) represents the position at time t, v(t) represents the speed at time t, a(t) represents the acceleration at time t, SOC(t) represents the battery power at time t, v min (s) represents the minimum speed at the position s on the road, v max (s) represents the maximum speed at position s, a min represents the minimum acceleration of the vehicle, a max Indicates the maximum acceleration of the vehicle, SOC min Indicates the minimum battery charge value, SOC max Maximum battery charge value.

[0055] In an exemplary embodiment, Figure 3 As shown, it is a schematic diagram of the level classification of interaction strength in urban complex driving scenes provided by the embodiment of the present application. The method for determining the interaction strength in urban complex road driving scenes is:

[0056] When there are no environmental vehicles in all eight directions around the vehicle and within the longitudinal front-to-back distance of X meters (in this embodiment, X=150), the driving scene is defined as a non-interaction scene; when there are two or more environmental vehicles (n≥2) in all eight directions around the vehicle and within the longitudinal front-to-back distance of X meters, and the distance between the vehicle and the vehicle in front is within the safe distance range, the driving scene is defined as a strong interaction scene; otherwise, when the conditions of the non-interaction scene and the strong interaction scene are not met, it is defined as a weak interaction scene.

[0057] In an exemplary embodiment, Figure 4 As shown, it is a schematic diagram of long-term-short-term rolling optimization provided by an embodiment of the present application. The rolling optimization solution steps for multi-mode long-term-short-term vehicle speed planning specifically include:

[0058] The optimal switching time is taken as the end time point of the intelligent driving mode speed planning. When time T p Predict the driving mode switching time T opt , then define the time interval I = [T p ,T opt ] is the short-time domain of the intelligent driving mode speed planning; in the alternating driving scenarios with no interaction, weak interaction and strong interaction within the long period H before the short-time domain, a distance-based rolling optimization solution strategy is established with a rolling step of Δs; in the driving scenarios within the short-time domain, a time-based rolling optimization solution strategy is established with a rolling step of ΔT, as shown in formula (15); where H = [T0, T p ];

[0059]

[0060] Among them, d rolling represents the rolling step, Δs represents the rolling distance step, and ΔT represents the rolling time step;

[0061] The rolling optimization solution strategy uses the current time T0 as the starting point and uses the dynamic programming algorithm to solve the optimal vehicle speed planning within the rolling time domain length L; according to the rolling advancement step length d rolling , the rolling time window moves forward, and the optimal speed planning within the next rolling time domain length is re-solved; the time window is continuously rolled and the speed planning is optimized until the optimal time for mode switching T is reached. opt ; Where L = 15s.

[0062] S3, hierarchical optimization control of energy management of heavy-duty truck power-split hybrid system with switching between intelligent driving and human driving.

[0063] In this embodiment, Figure 5 As shown, it is a schematic diagram of the hierarchical optimization control method for energy management of the hybrid system provided in the embodiment of the present application. In the hierarchical optimization control of energy management of the heavy-duty truck power-split hybrid system with intelligent driving-human driving switching, it includes upper-level control and lower-level control. Among them, the upper-level control is for the intelligent driving mode, and according to the planned economic speed curve, an approximate optimal SOC trajectory planning model is established and the global optimal SOC reference trajectory is solved; in the lower-level control, the global optimal SOC reference trajectory is used as input for the intelligent driving mode, and the reference SOC value with the best charging and discharging performance is used as input for the human driving mode, and a deep reinforcement learning model based on deep deterministic policy gradient (DDPG) is used to perform torque distribution control to minimize fuel consumption;

[0064] In an exemplary embodiment, the approximate optimal SOC trajectory planning is determined by the following steps: taking the fuel consumption as the objective function (Equation 16), the battery SOC as the state variable, and the engine power P e and the drive motor torque T m For the control variables (Equations 17 and 18), the optimal SOC trajectory planning model is established considering the power system constraints (Equation 19), and the dynamic programming algorithm is applied to solve the global optimal reference SOC trajectory;

[0065]

[0066] x=SOC(k),u=[P e (k),T m (k)] T (17)

[0067]

[0068] in, is the fuel consumption rate, R d 、V oc , Q b Respectively represent resistance, open circuit voltage and maximum battery capacity; required power M is the vehicle mass, C d is the drag coefficient, ρ is the air density, A is the frontal area, β = fcos(θ) + sin(θ) is the road coefficient of the rolling resistance coefficient f and the road slope θ; P e (k), P m (k), P g (k) represent the engine power, drive motor power and generator power respectively; R r , R r Respectively represent the radius of the ring gear and sun gear; ω g ,ω e ,ω m Represent the angular velocity of the generator, engine and drive motor respectively; P batt Represents battery power, η m , η g Indicates motor efficiency; P batt_Cmax , P batt_Dmax Indicates the maximum power of battery charging and discharging; ω e_min and ω e_max represents the minimum and maximum values ​​of the engine angular velocity, ω m_min and ω m_max Indicates the minimum and maximum angular velocity of the drive motor, ω g_min and ω g_max represents the minimum and maximum values ​​of the generator angular velocity, T e_min and T e_maxIndicates the minimum and maximum values ​​of the engine torque, T m_min and T m_max Indicates the minimum and maximum values ​​of the drive motor torque, T g_min and T g_max Indicates the minimum and maximum values ​​of the generator torque, SOC min and SOC max represents the minimum and maximum constraints of the battery power, and Δt represents the time interval divided into N stages according to the planning time range.

[0069] In an exemplary embodiment, in a deep reinforcement learning model based on deep deterministic policy gradient (DDPG), a state variable state is defined t (Equation 20) is the difference ΔSOC between the actual SOC and the reference SOC at the current moment t , vehicle speed t , acceleration acc t , define the action variable action t (Equation 20) is the engine power P ICEt , the reward function r (Equation 21) is defined as the fuel consumption rate And the global optimal SOC trajectory tracking error square ΔSOC 2 The negative number of the sum; the Critic network estimates the long-term return Q value (cumulative reward) based on the current state and action, and updates the network parameters through the gradient descent algorithm with the goal of minimizing the loss function (Formula 22); the Actor network outputs the optimal action (engine target power) based on the current state, and updates the network parameters through the gradient ascent algorithm with the goal of maximizing the expected long-term return (Formula 23);

[0070] state t =[ΔSOC t ,v t ,acc t ],action t =[P ICEt ] (20)

[0071]

[0072] Among them, state t is the state variable, ΔSOC t is the difference between the actual SOC and the reference SOC at the current moment, v t is the vehicle speed, acc t is the acceleration, ΔSOC t =ΔSOC t -SOC ref ; a and b represent weight factors, action t is the action variable, PICEt is the engine power; r t is the reward function, is the fuel consumption rate, ΔSOC 2 is the square of the global optimal SOC trajectory tracking error.

[0073]

[0074] Among them, γ∈[0,1) is the discount factor, which indicates the decay of future rewards; θ Q is the parameter of the Critic network, μ(state t+1 θ μ ) is the action generated by the Actor network, L(θ Q ) is the loss function, is to find the mathematical expectation, Q(state t ,action t θ Q ) is the Q-value estimate of the Critic network at the current moment.

[0075]

[0076] Among them, action t =μ(state t θ μ ), θ μ are the parameters of the Actor network, is the parameter θ μ Find the gradient, J(θ μ ) is the Q value of the Critic network, is to find the mathematical expectation, is the Q value function of the Critic network relative to the action action t gradient.

[0077] In an exemplary embodiment, torque distribution control is achieved by the following steps:

[0078] According to the optimal engine target power output by the deep reinforcement learning model, the engine speed corresponding to the highest engine efficiency point is determined through the engine's optimal operating point curve and output torque (Equation 24); then use the planetary gear ratio formula (Equation 25) and the hybrid system dynamics equation (Equation 26) to calculate the drive motor and the torque of the generator

[0079]

[0080] Where K = R r / Rs , represents the planetary gear coefficient, which is the ratio of the radius of the sun gear to the ring gear, i is the speed ratio of the main reducer, R tire is the tire radius, I e ,I g ,I m are the rotational inertia of the engine, generator and drive motor respectively, T e 、T g 、T m 、T c 、T s 、T r 、T t They are the torques at the engine, generator, drive motor, planet carrier, sun gear, ring gear and wheel, η represents the engine efficiency diagram, P ICE is the engine power, is the engine angular acceleration, is the angular acceleration of the drive motor, is the generator angular acceleration, is the vehicle acceleration, m is the static mass of the vehicle, M is the equivalent mass of the vehicle, g is the gravitational acceleration, C d is the drag coefficient, ρ is the air density, A is the frontal area, and β=fcos(θ)+sin(θ) is the road coefficient combining the rolling resistance coefficient f and the road slope θ.

[0081] The above implementation modes are only used to illustrate the present application, and are not intended to limit the present application. Ordinary technicians in the relevant technical field may make various changes and modifications without departing from the spirit and scope of the present application. Therefore, all equivalent technical solutions also belong to the scope of the present application, and the scope of patent protection of the present application shall be limited by the claims.

Claims

1. A method for energy-saving control of a trunk logistics hybrid heavy truck considering the switching between intelligent driving and human driving modes, characterized in that: The method comprises: Based on the mode switching process feature data set obtained by driver style recognition and the prediction of vehicle trajectories in the surrounding environment, the risk interval and risk change curve of the mode switching decision are obtained using the driving risk assessment method based on the risk field model. The time interval of the mode switching decision is obtained based on the risk interval and risk change curve of the mode switching decision. The mode switching optimization model based on extreme energy consumption is established by integrating the speed prediction of the human driving mode, and the switching timing is optimized to obtain the optimal switching timing. Taking the optimal switching time as the end time point of the intelligent driving mode speed planning, a future vehicle speed planning model considering the dynamic interaction characteristics of the driving scene with energy consumption cost as the optimization target is established; based on the interaction strength of the driving scene, the vehicle speed planning model in different future cycles and time domains is solved by using the rolling optimization of the vehicle speed planning for multi-mode long-cycle-short-time domain to obtain the optimal planned vehicle speed curve; According to the optimal planned vehicle speed curve, an approximate optimal SOC trajectory planning model is established and the global optimal SOC reference trajectory is solved; for the intelligent driving mode, the global optimal SOC reference trajectory is used as input, and for the human driving mode, the reference SOC value with the best charging and discharging performance is used as input to perform torque distribution control to minimize fuel consumption.

2. The energy-saving control method for trunk logistics hybrid heavy trucks considering intelligent driving-human driving mode switching as claimed in claim 1 is characterized in that: The mode switching process feature data set includes the position coordinates of the ego vehicle, the speed of the ego vehicle, the position coordinates of the environment vehicle, and the speed of the environment vehicle, which is expressed as formula (1); in, is the driving data of the mode switching process of the jth driver in the driving style of type a, is the driving data of the environment vehicle i, is the driving data of the jth driver; The surrounding vehicle trajectory prediction is calculated by the trajectory prediction model based on the environmental attention network EA-Net, and the trajectory prediction result is expressed as formula (2); Among them, F is the prediction of the surrounding vehicle trajectory, T l is the length of the historical time domain, T g is the interval length between the historical time domain and the future time domain, T p For the future prediction time domain length, f (t) ={x (t) ,y (t) }, f (t) For x (t) is the lateral position coordinate of the vehicle, y (t) is the longitudinal position coordinate of the vehicle; The risk range and risk change curve of mode switching decision are obtained by using the driving risk assessment method based on the risk field model, including: According to the influence of Z surrounding vehicles on the ego vehicle in all directions, the field strength E of the driving risk field generated by the surrounding moving vehicles at the ego vehicle position is calculated by formula (3): j : Among them, E V_ij is the surrounding vehicle i(x i ,y i ) in the vehicle j(x j ,y j )The risk field strength generated, (x i ,y i ) and (x j ,y j ) are the center of mass coordinates of the surrounding vehicles i and the vehicle j, k1, k2, G are constants greater than 0, R i is the impact factor of road conditions at the surrounding vehicle i, M i is the equivalent mass of the surrounding vehicle i, r ij =(x j -x i ,y j -y i ), represents the distance vector between the surrounding vehicle i and the vehicle j, and its direction is the same as the field strength direction, v i is the speed of surrounding vehicle i, θ i is the speed direction and distance vector r of the surrounding vehicle i ij The angle of Use the force calculation formula in the field strength to calculate the force on the vehicle as the driving risk risk of the vehicle j , the calculation formula is as follows; risk j =E j M j R j exp[-k2v j cos(θ j )](1+D rj ) (4) Among them, M j is the equivalent mass of the vehicle, R j is the influencing factor of the road conditions at the vehicle, v j is the vehicle speed, θ j is the direction of the vehicle's speed and the field strength E j The angle between the directions, D rj is the driver risk factor of the vehicle; The future position coordinates and speeds of the ego vehicle and surrounding vehicles are obtained based on the trajectory prediction results. Based on equations (3) and (4), the future T p The driving risk at each moment is used to obtain the future driving risk change curve; According to the driving data of the mode switching process of the current driver style category, the driving risk assessment method based on the risk field model is used to calculate the driving risk value of each driver in the driving data set when switching the driving mode. The minimum value is taken as the driving risk value when the mode is switched for this type of driving style, that is, the left interval of driving risk for the mode switching decision; the driving risk value of the collision position between the self-vehicle and the surrounding vehicles in the future trajectory is taken as the right interval of driving risk for the mode switching decision. The calculation process of taking the minimum value as the driving risk value when switching the mode of this type of driving style is as follows: in, represents the left interval of driving risk for the mode switching decision of a class driver, It represents the driving risk value of the jth driver in category a when switching the driving mode.

3. The energy-saving control method for trunk logistics hybrid heavy trucks considering intelligent driving-human driving mode switching as claimed in claim 2 is characterized in that: Based on the risk interval and risk change curve of mode switching decision, the time interval of mode switching decision is obtained. The mode switching optimization model based on extreme energy consumption is established by integrating the speed prediction of human driving mode. The switching timing is optimized to obtain the optimal switching timing, including: Driving risk interval based on driver mode switching decision The future driving risk change curve that changes over time determines the recommended time interval for the driver to switch driving modes The human driving mode speed prediction is obtained by using the LSTM recurrent neural network model, and the speed prediction sequence is expressed as: Where T e To predict the length of time; According to the speed prediction, mode switching time interval constraint and the energy consumption calculation model obtained by fitting, a mode switching optimization model based on extreme energy consumption is established, and the mode switching optimization model based on extreme energy consumption is solved to determine the optimal switching timing T for the driver. opt ; Among them, the mode switching optimization model based on extreme energy consumption is expressed by the following formula: Among them, T is the decision variable switching time, v in is the speed in intelligent driving mode, v hu Predict the speed for human driving mode; v(t) is the speed of the vehicle at time t, t0 is the left endpoint of the decision recommendation time interval, J is the energy consumption target value, denergy is the fitted instantaneous energy consumption calculation function, t is time, v is the vehicle speed, and a is the vehicle acceleration.

4. The energy-saving control method for trunk logistics hybrid heavy trucks considering intelligent driving-human driving mode switching as claimed in claim 1 is characterized in that: Taking the optimal switching time as the end time point of the intelligent driving mode speed planning, a future vehicle speed planning model taking energy consumption cost as the optimization target and considering the dynamic interaction characteristics of the driving scene is established, including: Based on the influence of dynamic changes in driving scenarios on the efficiency and SOC of the hybrid system, a hybrid system energy consumption model based on a deep neural network considering battery SOC and system efficiency is established, as shown in formula (8): in, represents the instantaneous energy consumption cost of the powertrain, f DNN represents the nonlinear mapping function describing the relationship between driving state and optimal energy consumption, v is the vehicle speed, a is the vehicle acceleration, SOC is the battery state of charge, η is the system efficiency, including the motor, engine and converter efficiency; Based on the hybrid system energy consumption model, the safety risk constraints of the surrounding leading vehicles and the traffic efficiency constraints of the signal phase are considered, and the future vehicle speed planning model considering the dynamic interaction characteristics of the driving scene is established with the energy consumption cost as the optimization target, as shown in equations (9)-(14): u=a(t),x=[v(t),s(t),SOC(t)] T (10) Among them, formula (9) represents the energy consumption target, which is the energy consumption integral in the time domain, T0 is the starting time of vehicle speed planning, T opt is the optimal switching time; Formula (10) indicates that the control variable is acceleration, and the state variables are speed, distance and battery power; Formulas (11)-(13) are differential equations of state variables, f NNS Q is a nonlinear mapping function that describes the relationship between driving state and optimal battery current. b is the battery capacity; Formula (14) is the constraint condition, including speed constraint, acceleration constraint, SOC constraint, following distance constraint and traffic light constraint, s p is the distance traveled by the preceding vehicle, δ s For safe distance, It indicates the time that the i-th traffic light has been running in its periodic clock when the vehicle starts. is the travel time required for a vehicle to pass the i-th intersection, expressed as Calculate the passing time of the vehicle in the signal cycle clock, T0 represents the starting point of the vehicle speed planning, x(t) represents the state variable at time t, u(t) represents the control variable at time t, v(t) represents the speed at time t, x represents the state variable set, u represents the control variable set, a(t) represents the acceleration at time t, s(t) represents the position at time t, v(t) represents the speed at time t, a(t) represents the acceleration at time t, SOC(t) represents the battery power at time t, v min (s) represents the minimum speed at the position s on the road, v max (s) represents the maximum speed at position s, a min represents the minimum acceleration of the vehicle, a max Indicates the maximum acceleration of the vehicle, SOC min Indicates the minimum battery charge value, SOC max Maximum battery charge value.

5. The energy-saving control method for trunk logistics hybrid heavy trucks considering intelligent driving-human driving mode switching as claimed in claim 4 is characterized in that: Based on the interaction strength of the driving scene, the future speed planning model considering the dynamic interaction characteristics of the driving scene is solved by using the multi-mode long-period-short-time domain speed planning rolling optimization to obtain the optimal planned speed curve, including: When there are no environmental vehicles in all eight directions around the vehicle and within the longitudinal front and rear distance of X meters, the driving scene is defined as a non-interaction scene; when there are at least two environmental vehicles in all eight directions around the vehicle and within the longitudinal front and rear distance of X meters, and the distance between the vehicle and the front vehicle is within the safe distance range, the driving scene is defined as a strong interaction scene; when the conditions of the non-interaction scene and the strong interaction scene are not met, it is defined as a weak interaction scene; where X is a set constant value; The optimal switching time is taken as the end time point of the intelligent driving mode speed planning. When time T p Predict the driving mode switching time T opt , then define the time interval I = [T p ,T opt ] is the short-time domain of the intelligent driving mode speed planning; in the alternating driving scenarios with no interaction, weak interaction and strong interaction within the long period H before the short-time domain, a distance-based rolling optimization solution strategy is established with a rolling step of Δs; in the driving scenarios within the short-time domain, a time-based rolling optimization solution strategy is established with a rolling step of ΔT, as shown in formula (15); where H = [T0, T p ]; Among them, d rolling represents the rolling step, Δs represents the rolling distance step, and ΔT represents the rolling time step; The rolling optimization solution strategy uses the current time T0 as the starting point and uses the dynamic programming algorithm to solve the optimal vehicle speed planning within the rolling time domain length L; according to the rolling advancement step length d rolling , the rolling time window moves forward, and the optimal speed planning within the next rolling time domain length is re-solved; the time window is continuously rolled and the speed planning is optimized until the optimal time for mode switching T is reached. opt ; Where L = 15s.

6. The energy-saving control method for trunk logistics hybrid heavy trucks considering intelligent driving-human driving mode switching as claimed in claim 1 is characterized in that: According to the optimal planned vehicle speed curve, an approximate optimal SOC trajectory planning model is established and a global optimal SOC reference trajectory is solved, including: Fuel consumption is used as the objective function, battery SOC is used as the state variable, and engine power P is used as the e and the drive motor torque T m The optimal SOC trajectory planning model is established by considering the power system constraints, and the dynamic programming algorithm is applied to solve the global optimal reference SOC trajectory. The objective function is shown in formula (16), the state variables and control variables are shown in formulas (17) and (18), and the power system constraints are shown in formula (19): x=SOC(x),u=[P e (k),T m (k)] T (17) in, is the fuel consumption rate, R b 、V oc , Q b Respectively represent resistance, open circuit voltage and maximum battery capacity; required power M is the equivalent mass of the vehicle, m is the static mass of the vehicle, C d is the drag coefficient, ρ is the air density, A is the frontal area, is the road coefficient combining the rolling resistance coefficient f and the road slope θ, g is the acceleration of gravity; P e (k), P m (k), P g (k) represent the engine power, drive motor power and generator power respectively; R r , R r Respectively represent the radius of the ring gear and sun gear; ω g ,ω e ,ω m Represent the angular velocity of the generator, engine and drive motor respectively; P batt Represents battery power, η m , η g Indicates motor efficiency; P batt_Cmax , P batt_Dmax Indicates the maximum power of battery charging and discharging; ω e_min and ω e_max represents the minimum and maximum values ​​of the engine angular velocity, ω m_min and ω m_max Indicates the minimum and maximum angular velocity of the drive motor, ω g_min and ω g_max represents the minimum and maximum values ​​of the generator angular velocity, T e_min and T e_max Indicates the minimum and maximum values ​​of the engine torque, T m_min and T m_max Indicates the minimum and maximum values ​​of the drive motor torque, T g_min and T g_max Indicates the minimum and maximum values ​​of the generator torque, SOC min and SOC max represents the minimum and maximum constraints of the battery power, and Δt represents the time interval divided into N stages according to the planning time range.

7. The energy-saving control method for trunk logistics hybrid heavy trucks considering intelligent driving-human driving mode switching as claimed in claim 6 is characterized in that: For the intelligent driving mode, the global optimal SOC reference trajectory is used as input, and for the human driving mode, the reference SOC value with the best charging and discharging performance is used as input to perform torque distribution control to minimize fuel consumption, including: For the intelligent driving mode, the global optimal SOC reference trajectory is used as input, and for the human driving mode, the reference SOC value with the best charging and discharging performance is used as input. The deep reinforcement learning model based on deep deterministic policy gradient is used to perform torque distribution control to minimize fuel consumption. In the deep reinforcement learning model based on deep deterministic policy gradient, the state variables are defined as the difference between the actual SOC and the reference SOC at the current moment, the vehicle speed, and the acceleration; the action variable is defined as the engine power; the reward function is defined as the inverse of the sum of the squares of the fuel consumption rate and the global optimal SOC trajectory tracking error; the Critic network estimates the long-term reward Q value based on the current state and action, and updates the network parameters through the gradient descent algorithm with the goal of minimizing the loss function; the Actor network outputs the optimal action based on the current state, and updates the network parameters through the gradient ascent algorithm with the goal of maximizing the expected long-term reward; wherein, the expressions of the state variables and the action variables are as shown in formula (20), the expression of the reward function is as shown in formula (21), the expression of the loss function is as shown in formula (22), and the update formula of the network parameters is as shown in formula (23): state t =[ΔSOC t ,v t ,acc t ],action t =[P ICEt ] (20) Among them, state t is the state variable, ΔSOC t is the difference between the actual SOC and the reference SOC at the current moment, v t is the vehicle speed, acc t is the acceleration, ΔSOC t =ΔSOC t -SOC ref ; a and b represent weight factors, action t is the action variable, P ICEt is the engine power; r t is the reward function, is the fuel consumption rate, ΔSOC 2 is the square of the global optimal SOC trajectory tracking error; Among them, γ∈[0,1) is the discount factor, which indicates the decay of future rewards; θ Q is the parameter of the Critic network, μ(state t+1 θ μ ) is the action generated by the Actor network, L(θ Q ) is the loss function, is to find the mathematical expectation, Q(state t ,action t θ Q ) is the Q value estimate of the Critic network at the current moment; Among them, action t =μ(state t θ μ ), θ μ are the parameters of the Actor network, is the parameter θ μ Find the gradient, J(θ μ ) is the Q value of the Critic network, is to find the mathematical expectation, is the Q value function of the Critic network relative to the action action t The gradient of According to the optimal engine target power output by the deep reinforcement learning model, the engine speed corresponding to the highest engine efficiency point is determined through the optimal operating point curve of the engine and output torque As shown in formula (24); the drive motor is calculated using the planetary gear set speed ratio formula as shown in formula (25) and the hybrid system dynamics equation as shown in formula (26) and the torque of the generator Where K = R r / R s , represents the planetary gear coefficient, which is the ratio of the radius of the sun gear to the ring gear, i is the speed ratio of the main reducer, R tire is the tire radius, I e ,I g ,I m are the rotational inertia of the engine, generator and drive motor respectively, T e 、T g 、T m 、T c 、T s 、T r 、T t They are the torques at the engine, generator, drive motor, planet carrier, sun gear, ring gear and wheel, η represents the engine efficiency diagram, P ICE is the engine power, is the engine angular acceleration, is the angular acceleration of the drive motor, is the vehicle acceleration, m is the static mass of the vehicle, M is the equivalent mass of the vehicle, g is the gravitational acceleration, C d is the drag coefficient, ρ is the air density, A is the frontal area, and β=fcos(θ)+sin(θ) is the road coefficient combining the rolling resistance coefficient f and the road slope θ.

Images