Ramp speed control method and device for electric scooter

Abstract

The invention provides a ramp speed control method and device for an electric scooter, and relates to the technical field of electric scooter speed control, and the method comprises the steps: collecting multi-source data in real time based on a preset multi-source data state space, and obtaining a multi-source feature state vector; inputting the multi-source feature state vector into a near-end strategy optimization algorithm to form a double-layer global strategy network, and outputting a corrected target speed and a real-time speed offset in real time; inputting the corrected target speed of the historical acquisition period and the safety speed limit threshold value into a convolutional neural network model to obtain a target speed prediction control model; inputting the real-time corrected target speed into the target speed prediction control model, and outputting a pre-adjusted target speed and a fault risk coefficient; optimizing a target speed prediction control model according to the fault risk coefficient; and inputting the pre-adjusted target speed and the real-time speed offset into the reinforcement learning strategy to obtain the optimal speed, thereby improving the driving safety.

Description

Technical Field

[0001] The present invention relates to the technical field of electric scooter speed control, and in particular to a method and device for controlling the ramp speed of an electric scooter. Background Art

[0002] With the popularity of electric scooters, the safety and comfort of speed control in complex slope environments have become technical difficulties. According to market research, the current ramp speed control method for electric scooters is generally based on single sensor data or preset rules. It is difficult to integrate multiple sources of information such as slope, battery status, and user operation in real time. This leads to risks such as sudden speed changes, battery overload, or failure of safety speed limits in scenarios such as steep slopes and continuous undulating slopes. For example, some solutions do not consider the dynamic impact of slope change rate and battery health status, and cannot achieve adaptive speed adjustment when the slope changes suddenly or the battery is low; other solutions lack accurate recognition of user operation intentions.

[0003] Furthermore, existing models often use static parameter designs and lack self-optimization mechanisms for high-risk operating conditions (such as high battery temperature and overspeed), making it difficult to ensure long-term prediction accuracy. Therefore, building a multi-source data-driven intelligent control model that can dynamically predict ramp speed, perform risk assessment, and perform adaptive optimization has become a key technical bottleneck in improving the safety of electric scooters.

[0004] The present invention integrates algorithms such as proximal strategy optimization, convolutional neural networks, and reinforcement learning to construct a multi-dimensional collaborative intelligent control framework, effectively solving the problems of insufficient control accuracy and safety in complex scenarios in existing technologies. Therefore, a method and device for controlling the slope speed of an electric scooter are now provided. Summary of the Invention

[0005] In order to solve the above technical problems, the object of the present invention is to provide a method and device for controlling the slope speed of an electric scooter.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a method for controlling the speed on a slope of an electric scooter, the method comprising: Based on a preset multi-source data state space, multi-source data of the electric scooter when it is traveling on a slope is collected in real time, and the multi-source data is input into the preset multi-source data state space to obtain a real-time multi-source feature state vector; Input the multi-source feature state vector into the proximal strategy optimization algorithm to construct a two-layer global strategy network; input the real-time multi-source feature state vector into the preset two-layer global strategy network, and then output the corrected target speed and real-time speed offset in real time; Inputting the corrected target speed of the historical acquisition period and the preset safety speed limit threshold into the convolutional neural network model to construct a target speed prediction control model; inputting the real-time output corrected target speed into the target speed prediction control model to output a pre-adjusted target speed and a fault risk coefficient; optimizing the target speed prediction control model based on the fault risk coefficient to ensure the accuracy of the pre-adjusted target speed; The pre-adjusted target speed and the real-time speed offset are input into the reinforcement learning strategy to obtain the optimal speed of the electric scooter when traveling on a slope.

[0007] Furthermore, the process of constructing the multi-source data state space includes: A data acquisition device is provided, the data acquisition device comprising a plurality of different sensor units and having a corresponding acquisition cycle, wherein the acquisition cycle includes a plurality of acquisition moments; multi-source data of the electric scooter when traveling on a slope is collected according to the acquisition moments; the multi-source data includes: slope, slope change rate, real-time speed, acceleration, battery state of charge, battery temperature, and handle pressure; A multi-source state vector is constructed according to the multi-source data; and feature processing is performed on the multi-source state vector based on an autoencoder to obtain a corresponding multi-source feature state vector.

[0008] Furthermore, the multi-source feature state vector is input into the proximal policy optimization algorithm to construct a two-layer global policy network, which includes: The two-layer global strategy network includes: an upper-layer global strategy network and a lower-layer real-time adjustment network; the multi-source characteristic state vector is input into the upper-layer global strategy network, and then a corrected target speed is output; the corrected target speed is input into the lower-layer real-time adjustment network, and then a real-time speed offset is output.

[0009] Furthermore, according to the upper global policy network, the process of outputting the corrected target speed includes: According to the multi-source characteristic state vector and based on the Softmax function, the optimal speed range of the current ramp type is output; according to the optimal speed range, the target speed is obtained, and the physical parameters of the target speed are corrected to obtain the corrected target speed.

[0010] Furthermore, the process of outputting the real-time speed offset according to the lower-layer real-time adjustment network includes: Obtain the slope, slope change rate, battery state of charge, battery temperature and handle pressure when driving on a slope; Obtain the base speed offset based on the slope and slope change rate; obtain the battery state compensation speed offset based on the battery state of charge and battery temperature; and obtain the user intended speed offset component based on the handle pressure. A real-time speed offset is obtained according to the basic speed offset, the battery state compensation speed offset and the user intended speed offset component.

[0011] Furthermore, the corrected target speed of the historical acquisition period and the preset safety speed limit threshold are input into the convolutional neural network model. The process of constructing the target speed predictive control model includes: Obtain the corrected target speed for several sets of historical acquisition cycles and preset a safe speed limit threshold; Based on the convolutional network neural network, a standard prediction model is constructed; according to the modified target speed, the safe speed limit threshold and the standard prediction model, a target speed prediction control model is constructed.

[0012] Furthermore, the process of inputting the real-time output corrected target speed into the target speed prediction control model and then outputting the pre-adjusted target speed and the fault risk coefficient includes: Obtaining an output vector according to a target speed predictive control model; obtaining a fault risk coefficient according to the output vector, an optimal speed range, and a safe speed limit threshold; According to the target speed prediction control model, a predicted basic target speed is obtained; according to the fault risk coefficient, the predicted basic target speed and the safe speed limit threshold, a pre-adjusted target speed is obtained.

[0013] Furthermore, according to the fault risk coefficient, the process of optimizing the target speed predictive control model includes: Failure risk factor Perform multi-level risk threshold classification to obtain category prediction values; The error between the predicted value and the true label is calculated using a loss function to obtain a loss value; according to The optimizer and the loss value are used to update the model parameters of the target speed predictive control model to obtain updated model parameters.

[0014] Furthermore, the process of obtaining the optimal speed of the electric scooter when traveling on a slope includes: The pre-adjusted target speed and real-time speed offset are input into the multi-source state vector to form an optimized state vector. Based on the reinforcement learning strategy, the optimal speed of the electric scooter in its current state when driving on a slope is obtained.

[0015] The second aspect of the present invention further provides a slope speed control device for an electric scooter, comprising: a data acquisition module, a data processing module, an intelligent control module and an intelligent optimization module; A data acquisition module is used to collect multi-source data when the electric scooter is traveling on a slope; The data processing module is used to construct a multi-source state vector based on multi-source data, and then output the corrected target speed and real-time speed offset; An intelligent control module is configured to construct a target speed prediction control model based on the modified target speed and the safety speed limit threshold, thereby outputting a pre-adjusted target speed and a failure risk factor; and based on a reinforcement learning strategy and the pre-adjusted target speed and the real-time speed offset, obtain the optimal speed of the electric scooter when traveling on a slope; The intelligent optimization module is used to optimize the target speed prediction control model according to the fault risk coefficient, thereby ensuring the accuracy of the pre-adjusted target speed.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. By constructing a multi-source data state space, integrating data collected by the inclination sensor, vehicle speed encoder, BMS unit, and handle pressure sensor, and utilizing autoencoders for feature dimensionality reduction, the system effectively eliminates data redundancy while retaining key features. This design enables the system to capture changes in the ramp environment and vehicle status in real time. Combined with the hierarchical decision-making mechanism of a two-layer global policy network (the upper layer determines the optimal speed range, and the lower layer adjusts the offset in real time), it achieves speed control accuracy and meets real-time requirements.

[0017] 2. The target speed predictive control model, built based on a convolutional neural network model, dynamically outputs the pre-adjusted target speed and fault risk factor. Combined with a reinforcement learning strategy for global optimization of the pre-adjusted speed and offset, the system automatically balances the safe speed limit with the user's operational intent, avoiding the speed fluctuations inherent in traditional rule-based control. Furthermore, continuous optimization of the model through the fault risk factor enhances system robustness over long-term operation, effectively mitigating safety hazards associated with hill driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0019] Figure 1 The figure is a schematic diagram of the steps of a method for controlling the speed on a slope of an electric scooter.

[0020] Figure 2 The figure is a schematic diagram of a module of a hill speed control device for an electric scooter.

[0021] Figure 3 A schematic diagram showing an electric scooter driving on a slope. DETAILED DESCRIPTION

[0022] To make the objectives, technical solutions, and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other implementations obtained by those of ordinary skill in the art without inventive effort are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, a method for controlling the slope speed of an electric scooter includes the following steps: Based on a preset multi-source data state space, multi-source data of the electric scooter when it is traveling on a slope is collected in real time, and the multi-source data is input into the preset multi-source data state space to obtain a real-time multi-source feature state vector; Input the multi-source feature state vector into the proximal strategy optimization algorithm to construct a two-layer global strategy network; input the real-time multi-source feature state vector into the preset two-layer global strategy network, and then output the corrected target speed and real-time speed offset in real time; A training sample set is formed by combining the corrected target speed of a historical acquisition cycle and a preset safety speed limit threshold. The training sample set is input into a convolutional neural network model to construct a target speed prediction control model. The real-time corrected target speed is input into the target speed prediction control model to output a pre-adjusted target speed and a fault risk coefficient. The target speed prediction control model is optimized based on the fault risk coefficient to ensure the accuracy of the pre-adjusted target speed. The pre-adjusted target speed and the real-time speed offset are input into the reinforcement learning strategy to obtain the optimal speed of the electric scooter when traveling on a slope.

[0024] It should be further explained that, in the specific implementation process, the specific process of constructing the multi-source data state space includes: A data acquisition device is preset, which is composed of several different sensor units and has a corresponding acquisition cycle. The acquisition cycle includes several acquisition moments; multi-source data of the electric scooter when it is traveling on a slope is collected according to the acquisition moments.

[0025] It should be further explained that the sensor unit includes an inclination sensor, a vehicle speed encoder, a BMS unit and a handle pressure sensor; the multi-source data includes: slope, slope change rate, real-time speed, acceleration, battery state of charge, battery temperature and handle pressure.

[0026] According to the multi-source data, a multi-source state vector is constructed and recorded as ; The multi-source state vector for: ;in, is the slope; is the slope change rate; is real-time speed; is the acceleration; is the battery charge state; is the battery temperature; is the handle pressure.

[0027] Based on the autoencoder, the multi-source state vector Perform feature processing to obtain the corresponding multi-source feature state vector ; It should be further explained that the multi-source data state space is composed of the data acquisition device and the autoencoder.

[0028] It should be further explained that, in the specific implementation process, based on the proximal policy optimization algorithm and according to the multi-source feature state vector, the specific process of constructing a two-layer global policy network includes: It should be further explained that the two-layer global policy network includes: an upper-layer global policy network and a lower-layer real-time adjustment network.

[0029] The multi-source feature state vector Input to the upper global policy network, and then output the corrected target speed; The modified target speed is input into the lower-layer real-time adjustment network, and then the real-time speed offset is output.

[0030] It should be further explained that, in the specific implementation process, according to the upper-level global policy network, the specific process of outputting the corrected target speed includes: According to the multi-source feature state vector , and based on the Softmax function, output the optimal speed range of the current ramp type ; According to the optimal speed range , get the target speed ; The target speed for: ; It should be further explained that the optimal speed range Maximum safe speed The maximum value of .

[0031] For the target speed Perform physical parameter correction to obtain the corrected target speed ; The corrected target speed for: ;in, is the slope influence coefficient; is the road friction coefficient; Indicates uphill; Indicates downhill.

[0032] It should be further explained that the types of slopes include: gentle upslope, steep upslope, gentle downslope, steep downslope and straight slope.

[0033] For example, the optimal speed range for a slow downhill slope is .

[0034] It should be further explained that, in the specific implementation process, the specific process of outputting the real-time speed offset according to the lower-layer real-time adjustment network includes: Get the slope when driving on a slope , slope change rate , battery state of charge , battery temperature and handle pressure ; According to the slope and slope change rate , get the base speed offset ; The base speed offset for: ;in, is the deceleration coefficient when the downhill slope becomes steeper; is the inherent deceleration coefficient for downhill slope; is the acceleration coefficient when the uphill slope slows down; is the inherent deceleration coefficient for uphill slopes; it is determined by technical personnel; It should be further explained that when going downhill, if the slope becomes steeper or the absolute value of the slope increases, the basic speed offset is reduced; when going uphill, if the slope becomes gentler or the slope decreases, the basic speed offset is increased.

[0035] According to the battery state of charge and battery temperature , obtain battery status compensation speed offset ; The battery status compensation speed offset for: ;in, Compensating for speed offset components for battery state of charge; Compensate the speed offset component for battery temperature; It should be further explained that the battery state of charge compensates for the speed offset component for: ;in, The battery state of charge decreases at a linear rate.

[0036] The battery temperature compensates the speed offset component for: ;in, is the linear battery temperature response coefficient.

[0037] According to the handle pressure , obtain the user's intended velocity offset component The user intended speed offset component for: ;in, is the pressure response coefficient; is the base pressure; is a sign function to ensure that the pressure and velocity change in the same direction.

[0038] According to the base speed offset , Battery status compensation speed offset and the user intended velocity offset component , get the real-time speed offset ; The real-time speed offset for: .

[0039] It should be further explained that, in the specific implementation process, the specific process of building the target speed predictive control model includes: Obtain the corrected target speed for several sets of historical acquisition cycles; Preset safe speed limit threshold; The corrected target speed and safety speed limit threshold of several groups of historical collection cycles are grouped and labeled, and recorded as is a natural number; Will The corrected target speed and safe speed limit threshold of the group history collection period are used as sample data, and is less than and using the sample data to obtain the mean of the sample data, which is recorded as a sample set; the corrected target speed and the safety speed limit threshold of the remaining groups of historical acquisition cycles are used as a test set; and a training sample set is formed based on the sample set and the test set; Based on convolutional network neural network, a standard prediction model is constructed; The training sample set is input into the standard prediction model, the standard prediction model is trained, and the standard prediction model after training is recorded as the target speed prediction control model.

[0040] The real-time output corrected target speed is input into the target speed prediction control model, and then the pre-adjusted target speed is output. and failure risk factor ; The failure risk factor for: ;in, is the weight matrix of the output layer of the target speed prediction control model; is the Sigmoid function, and the failure risk coefficient Map to the interval [0,1]; is the bias vector of the output layer of the target speed prediction control model; is the output vector of the fully connected layer of the target speed prediction control model; It should be further explained that the output vector for: ;in, is the activation function; is the weight matrix of the fully connected layer of the target speed predictive control model; is the bias vector of the fully connected layer of the target speed prediction control model; 、 Pooling layer operation for target speed prediction control model; 、 Convolutional layer operations for target speed prediction control model; For history Corrected target speed for each acquisition cycle; is the safety speed limit threshold, which is the optimal speed range ; Correct target speed in real time.

[0041] The pre-adjusted target speed for: ;in, is a limiting function, ensuring the pre-adjusted target speed within a safe range; is the predicted basic target speed of the target speed prediction control model; The minimum safe speed in the optimal speed range; It should be further explained that the target speed prediction control model predicts the basic target speed for: ;in, It is the output of the fully connected layer of the target speed prediction control model.

[0042] It should be further explained that, in the specific implementation process, the specific process of optimizing the target speed predictive control model according to the fault risk coefficient includes: Failure risk factor Perform multi-level risk threshold classification to obtain category prediction values ; The category prediction value for: ; It should be further explained that the category prediction values ​​include: non-optimized category prediction values, slightly optimized category prediction values, and deeply optimized category prediction values.

[0043] The error between the predicted value and the true label is calculated using the loss function, and the formula is: ;in, is the loss value; is the true label, and its value is or ; Predict the value for the corresponding category for probability; Predict the value for the corresponding category for probability; according to The optimizer updates the model parameters of the target speed predictive control model based on minimizing the loss function. The update formula is: ;in, is the learning rate; are the model parameters of the target speed predictive control model.

[0044] It should be further explained that, in the specific implementation process, based on the reinforcement learning strategy and according to the pre-adjusted target speed and the real-time speed offset, the specific process of obtaining the optimal speed of the electric scooter when traveling on a slope includes: Pre-adjust the target speed , real-time speed offset Input to multiple source state vectors In the above example, the optimized state vector is composed of ; The optimized state vector for: ; Based on the reinforcement learning strategy, the optimal speed of the electric scooter when driving on a slope is obtained. ; The optimal speed for: .

[0045] Furthermore, in one of the preferred embodiments of the present invention, in a steep downhill scenario, the traditional control solution only sets a fixed speed limit based on data from a single slope sensor, and cannot respond in real time to low battery, high temperature, and user operation intentions. This may result in: the speed exceeding the safety threshold (such as 25km / h), causing excessive braking distance or the risk of loss of control; continuous high current discharge when the battery is low, accelerating battery wear; When the user urgently grasps the handle, the speed adjustment lags behind the operation intention.

[0046] For example, Figure 3 As shown in the figure, when the electric scooter is driving on the city road, it encounters a slope of The road surface is dry asphalt, and the battery state of charge ( )for , the battery temperature is , the pressure of the user holding the handle is (Base pressure is ).

[0047] Obtain slope in real time through inclination sensor , slope change rate , the vehicle speed encoder collects real-time speed , acceleration (Downhill acceleration), BMS unit obtains , battery temperature , the handle pressure sensor obtains .

[0048] Based on the upper global strategy network, the Softmax function is used to determine whether the current situation is a "steep downhill" type and output the optimal speed range. Target speed . And combined with the slope influence coefficient , road friction coefficient (Dry road) Corrected to the corrected target speed .

[0049] The deceleration factor when the slope becomes steeper The value is ; Downhill inherent deceleration coefficient The value is ; Acceleration coefficient when the uphill slope slows down The value is ; Uphill inherent deceleration coefficient The value is ; Linear battery state of charge decrease rate The value is ;Linear battery temperature reduction rate The value is .

[0050] Calculate the basic speed offset based on the underlying real-time adjustment network ; Battery status compensation speed offset ; User intended velocity offset component ; Real-time speed offset .

[0051] The historical corrected target speed and the safety speed limit threshold are as follows: enter Model, output pre-adjusted target speed (Failure risk factor γ = 0.6, triggering mild optimization); Fusion pre-adjusted target speed Offset from real-time speed , to obtain the best speed , which satisfies both the safety speed limit ( ), responds to the user's acceleration intention while avoiding battery overload.

[0052] like Figure 2 As shown, a slope speed control device for an electric scooter includes a data acquisition module, a data processing module, an intelligent control module and an intelligent optimization module; A data acquisition module is used to collect multi-source data when the electric scooter is traveling on a slope; The data processing module is used to construct a multi-source state vector based on multi-source data, and then output the corrected target speed and real-time speed offset; An intelligent control module is configured to construct a target speed prediction control model based on the modified target speed and the safety speed limit threshold, thereby outputting a pre-adjusted target speed and a failure risk factor; and based on a reinforcement learning strategy and the pre-adjusted target speed and the real-time speed offset, obtain the optimal speed of the electric scooter when traveling on a slope; The intelligent optimization module is used to optimize the target speed prediction control model according to the fault risk coefficient, thereby ensuring the accuracy of the pre-adjusted target speed.

[0053] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for controlling the slope speed of an electric scooter, characterized in that: The method comprises: Based on a preset multi-source data state space, multi-source data of the electric scooter when it is traveling on a slope is collected in real time, and the multi-source data is input into the preset multi-source data state space to obtain a real-time multi-source feature state vector; Input the multi-source feature state vector into the proximal strategy optimization algorithm to construct a two-layer global strategy network; input the real-time multi-source feature state vector into the preset two-layer global strategy network, and then output the corrected target speed and real-time speed offset in real time; A training sample set is formed by combining the corrected target speed of a historical acquisition cycle and a preset safety speed limit threshold. The training sample set is input into a convolutional neural network model to construct a target speed prediction control model. The real-time corrected target speed is input into the target speed prediction control model to output a pre-adjusted target speed and a fault risk coefficient. The target speed prediction control model is optimized based on the fault risk coefficient to ensure the accuracy of the pre-adjusted target speed. The pre-adjusted target speed and the real-time speed offset are input into the reinforcement learning strategy to obtain the optimal speed of the electric scooter when traveling on a slope.

2. The method according to claim 1, characterized in that The process of constructing a multi-source data state space includes: A data acquisition device is provided, the data acquisition device comprising a plurality of different sensor units and having a corresponding acquisition cycle, wherein the acquisition cycle includes a plurality of acquisition moments; multi-source data of the electric scooter when traveling on a slope is collected according to the acquisition moments; the multi-source data includes: slope, slope change rate, real-time speed, acceleration, battery state of charge, battery temperature, and handle pressure; A multi-source state vector is constructed according to the multi-source data; and feature processing is performed on the multi-source state vector based on an autoencoder to obtain a corresponding multi-source feature state vector.

3. The method according to claim 2, characterized in that The multi-source feature state vector is input into the proximal policy optimization algorithm to construct a two-layer global policy network, which includes: The two-layer global strategy network includes: an upper-layer global strategy network and a lower-layer real-time adjustment network; the multi-source characteristic state vector is input into the upper-layer global strategy network, and then a corrected target speed is output; the corrected target speed is input into the lower-layer real-time adjustment network, and then a real-time speed offset is output.

4. The method according to claim 3, characterized in that According to the upper global policy network, the process of outputting the corrected target speed includes: According to the multi-source characteristic state vector and based on the Softmax function, the optimal speed range of the current ramp type is output; according to the optimal speed range, the target speed is obtained, and the physical parameters of the target speed are corrected to obtain the corrected target speed.

5. The method according to claim 4, characterized in that The process of outputting the real-time velocity offset according to the underlying real-time adjustment network includes: Obtain the slope, slope change rate, battery state of charge, battery temperature and handle pressure when driving on a slope; Obtain the base speed offset based on the slope and slope change rate; obtain the battery state compensation speed offset based on the battery state of charge and battery temperature; and obtain the user intended speed offset component based on the handle pressure. A real-time speed offset is obtained according to the basic speed offset, the battery state compensation speed offset and the user intended speed offset component.

6. The method according to claim 5, characterized in that The corrected target speed of the historical acquisition period and the preset safety speed limit threshold are input into the convolutional neural network model. The process of building the target speed predictive control model includes: Obtain the corrected target speed for several sets of historical acquisition cycles and preset a safe speed limit threshold; The corrected target speeds of the plurality of historical acquisition cycles and the preset safety speed limit thresholds are combined into a training sample set; a standard prediction model is constructed based on a convolutional neural network; The training sample set is input into a standard prediction model for training, thereby constructing a target speed prediction control model.

7. The method according to claim 6, characterized in that The process of inputting the real-time output corrected target speed into the target speed prediction control model and then outputting the pre-adjusted target speed and the fault risk coefficient includes: Obtaining an output vector according to a target speed predictive control model; obtaining a fault risk coefficient according to the output vector, an optimal speed range, and a safe speed limit threshold; According to the target speed prediction control model, a predicted basic target speed is obtained; according to the fault risk coefficient, the predicted basic target speed and the safe speed limit threshold, a pre-adjusted target speed is obtained.

8. The method according to claim 7, characterized in that According to the fault risk coefficient, the process of optimizing the target speed predictive control model includes: Failure risk factor Perform multi-level risk threshold classification to obtain category prediction values; The error between the predicted value and the true label is calculated using a loss function to obtain a loss value; according to The optimizer and the loss value are used to update the model parameters of the target speed predictive control model to obtain updated model parameters.

9. The method according to claim 8, characterized in that The process of obtaining the optimal speed of an electric scooter when traveling on a slope includes: The pre-adjusted target speed and real-time speed offset are input into the multi-source state vector to form an optimized state vector. Based on the reinforcement learning strategy, the optimal speed of the electric scooter in its current state when driving on a slope is obtained.

10. A device for controlling the speed on a slope of an electric scooter, implementing the method for controlling the speed on a slope of an electric scooter according to any one of claims 1 to 9, comprising: Data acquisition module, data processing module, intelligent control module and intelligent optimization module; A data acquisition module is used to collect multi-source data when the electric scooter is traveling on a slope; The data processing module is used to construct a multi-source state vector based on multi-source data, and then output the corrected target speed and real-time speed offset; An intelligent control module is used to construct a target speed prediction control model based on the corrected target speed and the safety speed limit threshold, and then output a pre-adjusted target speed and a fault risk coefficient; Based on the reinforcement learning strategy and according to the pre-adjusted target speed and the real-time speed offset, the optimal speed of the electric scooter when traveling on the slope is obtained; The intelligent optimization module is used to optimize the target speed prediction control model according to the fault risk coefficient, thereby ensuring the accuracy of the pre-adjusted target speed.

Images