Intelligent car and intelligent control method thereof

Abstract

The present invention provides an intelligent car and an intelligent control method thereof, belonging to the technical field of elevator cars. The intelligent car comprises a car body, one end of which is provided with a switch slot, two air dispersion plates, the two air dispersion plates are fixedly connected to the switch slot, an exhaust fan, the exhaust fan is connected to the adjacent ends of the two air dispersion plates by rotating an electric roller, and two sets of driving mechanisms are arranged on the upper side of the car body. The exhaust fan can be driven to rotate between the two air dispersion plates by the rotation of the electric roller to perform ventilation treatment in the car body. At the same time, the rotation of the electric roller will also start the driving mechanism, and the sterilization and deodorization liquid such as spices in the spray pipe will be sprayed out through the diffusion nozzle through the driving mechanism. At the same time, the sprayed liquid is further diffused through the rotation of the exhaust fan, thereby achieving the effect of sterilization and deodorization, and improving the user experience.

Description

Technical Field

[0001] The present invention belongs to the technical field of elevator cars, and in particular relates to an intelligent car and an intelligent control method thereof. Background Art

[0002] The elevator car is the passenger or cargo transport part of the elevator. It is a box-shaped structure with four sides closed, designed to move vertically up and down in the elevator shaft. The interior of the car is usually equipped with lighting, ventilation systems, mirrors, handrails, floors, control panels and other facilities to ensure the comfort and safety of passengers.

[0003] As an important part of the elevator, the elevator car is directly related to the safety and riding experience of passengers. However, there are still many problems in the use of existing elevator cars, such as:

[0004] 1. The ventilation system in traditional elevator cars is not smooth, which causes the air to be stuffy or have a strange smell, especially when there are many people, which affects the riding experience of passengers;

[0005] 2. When the elevator stops, the car is not flush with the floor, and passengers need to step in and out, which is particularly inconvenient for the elderly and people with limited mobility. It will also affect the entry and exit of devices such as carts in the car, affecting the use of the car. Summary of the invention

[0006] The purpose of the present invention is to provide an intelligent car and an intelligent control method thereof, aiming to solve the problems in the prior art of insufficient car ventilation and unevenness between the car and the floor.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] An intelligent car, comprising:

[0009] A car body, one end of which is provided with a switch slot;

[0010] Two air dispersion plates, both of which are fixedly connected to the switch slot;

[0011] An exhaust fan, the exhaust fan being rotatably connected to the adjacent ends of the two air dispersion plates through an electric roller;

[0012] Two sets of driving mechanisms, both sets of driving mechanisms are arranged on the upper side of the car body; and

[0013] Two groups of spraying mechanisms, each group of the spraying mechanisms consists of a liquid storage tank, two spraying pipes and two transmission pipes, the liquid storage tank is fixedly connected to one end of the car body, the two transmission pipes are fixedly connected to one end of the car body, and the two transmission pipes are communicated with the liquid storage tank, the two spraying pipes are fixedly connected to the upper inner wall of the car body, and the two spraying pipes are fixedly connected to the two transmission pipes respectively.

[0014] As a preferred solution of the present invention, each group of the driving mechanism consists of two lower pressure blocks, a screw, a screw nut, two external blocks, two lower pressure rods, two lower pressure tubes, two I-shaped wheels and a transmission belt. The two lower pressure tubes are respectively fixedly connected to the upper ends of the two spray pipes, the screw is rotatably connected to the upper end of the car body, the two I-shaped wheels are respectively fixedly connected to the circumferential surfaces of the electric roller and the screw, the transmission belt is transmission-connected to the circumferential surfaces of the two I-shaped wheels, the screw nut is threadedly connected to the circumferential surface of the screw, the two external blocks are respectively fixedly connected to the two ends of the screw nut, the two lower pressure rods are respectively fixedly connected to the lower ends of the two external blocks, and the two lower pressure blocks are respectively fixedly connected to the lower ends of the two lower pressure rods.

[0015] As a preferred solution of the present invention, a plurality of vertical rods are fixedly connected to the upper end of the car body, a top plate is fixedly connected to the upper ends of the plurality of vertical rods, a plurality of stabilizing holes are opened at the upper end of the lead screw nut, and the lead screw nut is slidably connected to the circumferential surfaces of the plurality of vertical rods through the plurality of stabilizing holes.

[0016] As a preferred solution of the present invention, an anti-slip disk is fixedly connected to the upper end of the electric roller, and the diameter of the anti-slip disk is larger than that of the electric roller.

[0017] As a preferred solution of the present invention, a replenishing groove is provided at one end of the liquid storage tank, and a sealing plate is rotatably connected to one end of the liquid storage tank via a rotating shaft, and a hand grip groove is provided at one end of the sealing plate.

[0018] As a preferred solution of the present invention, the upper and lower inner walls of the two switch grooves are provided with sealing grooves, a first electric guide rail is fixedly connected between the inner walls on both sides of the plurality of sealing grooves, a first slider is slidably connected to the circumferential surfaces of the plurality of first electric guide rails, and the plurality of first sliders are respectively fixedly connected to the two switch doors.

[0019] As a preferred solution of the present invention, a movable groove is opened at one end of the car body, an extension plate is slidably connected in the movable groove, both ends of the extension plate are rotatably connected to gears through a rotating shaft, two racks are fixedly connected to the lower inner wall of the movable groove, and the two gears are respectively meshed with the two racks.

[0020] As a preferred solution of the present invention, two connecting plates are fixedly connected to one end of the car body, and a plurality of infrared calibrators are arranged at one end of the two connecting plates.

[0021] As a preferred solution of the present invention, one end of the two spray pipes is fixedly connected to a diffusion nozzle, the two diffusion nozzles are both conical, and the two diffusion nozzles are both located at the lower side of the discharge fan.

[0022] As a preferred solution of the present invention, ejection grooves are provided on both side inner walls of the movable groove, and second electric guide rails are fixedly connected to both side inner walls of the two ejection grooves, and second sliders are slidably connected to the circumferential surfaces of the two second electric guide rails, and the two second sliders are rotatably connected to two rotating shafts respectively.

[0023] The intelligent control method of an intelligent car of the present invention achieves a balance between energy saving and comfort through an intelligent control algorithm. The specific implementation process is as follows:

[0024] Step 1: Data collection;

[0025] Real-time collection of environmental parameters including temperature, humidity, CO2 concentration and energy consumption data; recording frequency of use, number of passengers and instant energy consumption each time the elevator is used; collecting passenger feedback data on environmental comfort;

[0026] Step 2: Use transfer learning to process and analyze time series data;

[0027]

[0028] Among them, Indicates source domain dataset, subscript or superscript represents the source domain; Indicates source domain data samples, including feature vectors and the corresponding labels ; Indicates Feature vectors of source domain data samples; Indicates Labels of source domain data samples; Represents the number of samples in the source domain dataset;

[0029]

[0030] in, Indicates the target domain dataset, subscript or superscript represents the target domain; Indicates target domain data samples, including feature vectors and the corresponding labels ; Indicates Feature vectors of target domain data samples; Indicates The labels of target domain data samples; Represents the number of samples in the target domain dataset;

[0031] Train a deep neural network model on the source domain data:

[0032]

[0033] in, Represents the source domain function output; Represents input data; represents the parameters of the source domain neural network; Represents a neural network, through the parameters and input Perform forward propagation and obtain the output;

[0034] Fine-tune the pre-trained model on the target domain data to adapt it to the specific environment of the elevator:

[0035]

[0036] in, It is a function of the target domain, representing the forward propagation process of a neural network; Represents input data; represents the parameters of the target domain neural network, Indicates the change in parameter, used to change the source domain parameter Generate target domain parameters , represents a neural network;

[0037] Step 3: Define the transfer learning fine-tuning objective function;

[0038] In the fine-tuning phase, by adjusting the parameters of the source model To adapt to the target task;

[0039]

[0040] in, is a function of the target domain, represents the source domain function output, Represents input data; represents the parameters of the source domain neural network; represents the parameters of the target domain neural network, Indicates the change in parameter;

[0041]

[0042] in, represents the sample-level loss function, is a function of the target domain, represents the parameters of the target domain neural network, Indicates the target domain data The input features of samples, Indicates the target domain data The labels of the samples;

[0043] Define the following fine-tuning objective function;

[0044]

[0045] in, represents the overall loss function; Represents the sample-level loss function, which evaluates the difference between the predicted value and the actual value; Indicates the number of target domain data samples; represents the parameters of the target domain neural network, represents the parameters of the source domain neural network, is a function of the target domain, Indicates the target domain data The input features of samples, Indicates the target domain data The labels of the samples; Represents the regularization coefficient, which controls the degree of deviation between the fine-tuning parameters and the pre-training parameters; Represents the regularization term to prevent the model from overfitting;

[0046] Use gradient descent to update the parameters to minimize the loss function:

[0047]

[0048] in, Indicates the number of target domain data samples; Representation function In the target domain, samples The predicted value of Indicates that it is a sample The true label of Indicates samples, is a function In the parameters Next pair of samples The gradient of represents the regularization coefficient, represents the parameters of the source domain neural network; Represents the loss function About parameters The gradient of

[0049]

[0050] in, represents the parameters of the target domain neural network, is the learning rate, which controls the step size of each update; Represents the loss function About parameters The gradient of

[0051] Step 4: Use reinforcement learning to find the optimal actions to take in different states;

[0052] Use the historical operation data of the car to train the DQN model and find the optimal actions to take in different states;

[0053] Defining states is a vector containing multiple variables;

[0054]

[0055] in, Indicates the frequency of use of the car; Indicates the current time; Indicates the number of passengers in the car; Indicates the current energy consumption in the car;

[0056] Defining actions To adjust the power of each system, set the following operations:

[0057]

[0058] in, Indicates adjustment of the exhaust fan power; Indicates the power of adjusting the driving mechanism; Indicates adjustment of the power of lighting facilities; Indicates the power adjustment of air conditioning facilities;

[0059] Designing the reward function To minimize energy consumption while taking into account passenger comfort;

[0060]

[0061] in, represents the adjusted energy consumption; Indicates the degree of reduction in passenger comfort; Indicates the degree of improvement in passenger comfort; , , Represents the weight coefficient, which is used to balance the relationship between energy consumption and comfort;

[0062] Q Function Indicates in status Take Action The total expected return is:

[0063]

[0064] in, is in state Take action The state-action value function of Indicates expected value; is a discount factor that balances the importance of current rewards with future rewards; Represents the discount factor Step by time The power of is the time step Instant rewards; Indicates the starting state is , the starting action is ; and Represents the time step status and actions;

[0065] The core of DQN is to update the Q function to estimate Take action The expected utility of . The update rule of the Q function is as follows:

[0066]

[0067] in, represents the new Q value, Indicates the current Q value; Is the current state Take Action Immediate rewards after is a discount factor that weighs immediate rewards against future rewards; Yes Status After action The new state to which it is transferred later; Is in the new state One of all possible actions under is in the next state Select Action The maximum Q value when are the parameters of the target network, is the learning rate;

[0068] Use experience replay and the target network to update the Q function;

[0069]

[0070] in, represents the target Q value; Is the current state Take Action Immediate rewards after is a discount factor that weighs immediate rewards against future rewards; Yes Status After action The new state to which it is transferred later; Is in the new state One of all possible actions under is in the next state Select Action The maximum Q value when are the parameters of the target network, is the target Q network for the new state and actions The estimated Q value of

[0071] The loss function is the mean square error between the current Q value and the target Q value:

[0072]

[0073] in, is the loss function; Indicates expected value; is the target Q value; is the current Q network pair state and actions Q value estimation of ; are the parameters of the current Q network;

[0074] Update the parameters of the Q network using gradient descent:

[0075]

[0076] in, are the parameters of the current Q network; is the learning rate, which determines the step size of each update; is the loss function Relative to the parameters gradient.

[0077] Compared with the prior art, the present invention has the following beneficial effects:

[0078] 1. The rotation of the electric roller can drive the exhaust fan to rotate between the two air dispersion plates to ventilate the car body. At the same time, the rotation of the electric roller will also start the driving mechanism, which will spray the fragrance and other sterilization and deodorization liquids in the spray tube through the diffusion nozzle. At the same time, the sprayed liquid is further diffused through the rotation of the exhaust fan, thereby achieving the effect of sterilization and deodorization and improving the user experience.

[0079] 2. After the car body and the floor are parallel, the parallelism of the car body and the floor can be improved through the detection of the infrared calibrator and the floor. After the car body stops, the second electric guide rail is started to slide the second slider on its circumferential surface and drive the extension plate to move. During the movement, the gear will mesh with the rack, driving the extension plate to rotate until it contacts the ground and stops. The paving of the extension plate can facilitate the entry and exit of personnel and facilities.

[0080] 3. When the screw nut moves, the stabilizing hole will synchronously slide on the circumferential surface of the vertical rod, thereby improving the vertical sliding stability of the screw nut and preventing the screw nut from rotating during movement, thereby improving the stability and rationality of the use of the device.

[0081] 4. Through transfer learning, the pre-trained air quality prediction model can quickly adapt to the specific environment of the car, reducing the data and time required to retrain the model in the target domain. The model fine-tuned by transfer learning can more accurately predict the air quality parameters in the car, such as temperature, humidity, and CO2 concentration, thereby providing more refined environmental control. Since the model has been pre-trained in a similar environment, transfer learning can accelerate the model deployment and optimization process, enabling the system to respond to environmental changes more quickly. The DQN model can dynamically adjust the power settings of each system based on real-time environmental data and historical operating data, and automatically learn the optimal energy-saving strategy. Through continuous exploration and updating, DQN finds the global optimal solution for energy consumption and passenger comfort, avoiding the local optimal problem in traditional control methods. Through reinforcement learning optimization, the energy consumption of exhaust fans, drive mechanisms, lighting, and air conditioning has been significantly reduced, and the energy-saving effect is obvious. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0083] Figure 1 It is the main stereoscopic view in the present invention;

[0084] Figure 2 It is a side perspective view of the present invention;

[0085] Figure 3It is the first main sectional stereogram in the present invention;

[0086] Figure 4 It is a second main sectional stereoscopic view of the present invention;

[0087] Figure 5 It is the third main sectional stereoscopic view of the present invention;

[0088] Figure 6 It is a side cutaway stereogram of the present invention;

[0089] Figure 7 It is a top-down perspective view of the present invention;

[0090] Figure 8 For the present invention Figure 7 A partial enlarged view of the middle A;

[0091] Fig. 9 It is a flow chart of the intelligent control algorithm in the present invention.

[0092] In the figure: 1. Car body; 101. Switch slot; 102. Switch door; 103. Liquid storage tank; 104. Sealing plate; 105. Handle slot; 106. Air dispersion plate; 107. Connecting plate; 108. Infrared calibrator; 109. Transmission pipe; 110. Spray pipe; 111. Diffusion nozzle; 112. Exhaust fan; 2. I-shaped wheel; 201. Transmission belt; 202. Top plate; 203. Vertical rod; 204. Wire Rod nut; 205, external block; 206, pressure rod; 207, pressure tube; 208, electric roller; 209, anti-slip disk; 210, pressure block; 211, screw rod; 3, sealing groove; 301, first electric guide rail; 302, first slider; 303, moving groove; 304, ejection groove; 305, second electric guide rail; 306, second slider; 307, gear; 308, rack; 309, extension plate. DETAILED DESCRIPTION

[0093] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0094] Example 1

[0095] See also Figure 1-8 , the present invention provides the following technical solutions:

[0096] An intelligent car, comprising:

[0097] A car body 1, one end of which is provided with a switch slot 101;

[0098] Two air dispersion plates 106, both of which are fixedly connected to the switch slot 101;

[0099] The exhaust fan 112 is rotatably connected to the adjacent ends of the two air dispersion plates 106 through the electric roller 208;

[0100] Two sets of driving mechanisms, both of which are arranged on the upper side of the car body 1, each set of driving mechanisms consists of two pressing blocks 210, a screw rod 211, a screw rod nut 204, two external blocks 205, two pressing rods 206, two pressing pipes 207, two I-shaped wheels 2 and a transmission belt 201. The two pressing pipes 207 are respectively fixedly connected to the upper ends of the two spray pipes 110, the screw rod 211 is rotatably connected to the upper end of the car body 1, and the two I-shaped wheels 2 are respectively connected to the upper ends of the two spray pipes 110. The two external blocks 205 are respectively fixedly connected to the circumferential surfaces of the motorized roller 208 and the screw rod 211, the transmission belt 201 is drivingly connected to the circumferential surfaces of the two I-shaped wheels 2, the screw rod nut 204 is threadedly connected to the circumferential surface of the screw rod 211, the two external blocks 205 are respectively fixedly connected to the two ends of the screw rod nut 204, the two lower pressure rods 206 are respectively fixedly connected to the lower ends of the two external blocks 205, and the two lower pressure blocks 210 are respectively fixedly connected to the lower ends of the two lower pressure rods 206; and

[0101] There are two groups of spraying mechanisms, each group of spraying mechanisms consists of a liquid storage tank 103, two spraying pipes 110 and two transmission pipes 109. The liquid storage tank 103 is fixedly connected to one end of the car body 1, and the two transmission pipes 109 are fixedly connected to one end of the car body 1, and the two transmission pipes 109 are communicated with the liquid storage tank 103, and the two spraying pipes 110 are fixedly connected to the upper inner wall of the car body 1, and the two spraying pipes 110 are fixedly connected to the two transmission pipes 109 respectively.

[0102] In a specific embodiment of the present invention, when the car body 1 is used daily, the electric roller 208 is started, and the electric roller 208 drives the exhaust fan 112 to rotate in the two air dispersion plates 106, so that the gas in the car body 1 and the external gas are exchanged, so as to reduce the odor in the car body 1. At the same time, when the electric roller 208 rotates, the rotational force of the electric roller 208 can be transmitted to the screw rod 211 through the two I-shaped wheels 2 and the transmission belt 201, so that the screw rod 211 rotates, and then the screw rod nut 201 connected with the thread on its circumferential surface is 4 will descend, and then the external block 205 and the pressing rod 206 will descend. At this time, the descending of the pressing rod 206 will drive the pressing block 210 to descend, squeeze the fragrance or sterilization and deodorization liquid in the spray pipe 110, and spray it out through the spray pipe 110. The sprayed liquid can also expand the area through the discharge fan 112 to remove the odor in the car body 1, thereby improving the user experience. The liquid in the liquid storage tank 103 can be transferred to the spray pipe 110 through the transmission pipe 109, so as to facilitate the timely replenishment of the liquid in the spray pipe 110.

[0103] For details, please refer to Figure 1-9 The upper end of the car body 1 is fixedly connected to a plurality of vertical rods 203, the upper ends of the plurality of vertical rods 203 are fixedly connected to a top plate 202, a plurality of stabilizing holes are opened at the upper end of the lead screw nut 204, and the lead screw nut 204 is slidably connected to the circumferential surfaces of the plurality of vertical rods 203 through the plurality of stabilizing holes, and the upper end of the electric roller 208 is fixedly connected to an anti-slip disk 209, and the diameter of the anti-slip disk 209 is larger than that of the electric roller 208.

[0104] In this embodiment: when the screw nut 204 moves, it can slide on the circumferential surface of the vertical rod 203 through the stabilizing hole, thereby improving the vertical sliding stability of the screw nut 204, preventing the screw nut 204 from rotating with the screw 211, and improving the rationality of the use of the screw nut 204. The anti-detachment disk 209 can stabilize the two I-shaped wheels 2 on the circumferential surface of the electric roller 208, preventing the I-shaped wheels 2 from detaching from the circumferential surface of the electric roller 208.

[0105] For details, please refer to Figure 1-9 A replenishing groove is provided at one end of the liquid storage tank 103, and a sealing plate 104 is rotatably connected to one end of the liquid storage tank 103 through a rotating shaft, a hand-grip groove 105 is provided at one end of the sealing plate 104, and sealing grooves 3 are provided on the upper and lower inner walls of the two switch grooves 101, and a first electric guide rail 301 is fixedly connected between the inner walls on both sides of the multiple sealing grooves 3, and the circumferential surfaces of the multiple first electric guide rails 301 are slidably connected to the first sliders 302, and the multiple first sliders 302 are respectively fixedly connected to the two switch doors 102.

[0106] In this embodiment: the liquid in the liquid storage tank 103 can be replenished through the replenishing groove, and the replenishing groove can be closed through the sealing plate 104 to prevent the liquid inside the liquid storage tank 103 from overflowing, and the hand buckle groove 105 facilitates the replenishment and pulling of the sealing plate 104, and the first slider 302 can be made to slide by starting the first electric guide rail 301, and drive the two switch doors 102 to move to the closer ends, so as to facilitate the closing and opening of the car body 1, thereby improving the rationality of the use of this device.

[0107] For details, please refer to Figure 1-9 A movable groove 303 is provided at one end of the car body 1, and an extension plate 309 is slidably connected in the movable groove 303. Both ends of the extension plate 309 are rotatably connected to gears 307 through a rotating shaft. Two racks 308 are fixedly connected to the lower inner wall of the movable groove 303, and the two gears 307 are respectively engaged with the two racks 308. Two connecting plates 107 are fixedly connected to one end of the car body 1, and multiple infrared calibrators 108 are provided at one end of the two connecting plates 107. Push-out grooves 304 are provided on the inner walls on both sides of the movable groove 303. Second electric guide rails 305 are fixedly connected to the inner walls on both sides of the two push-out grooves 304. Second sliders 306 are slidably connected to the circumferential surfaces of the two second electric guide rails 305, and the two second sliders 306 are respectively rotatably connected to the two rotating shafts.

[0108] In this embodiment: a plurality of infrared calibrators 108 fixedly connected at one end of the connecting plate 107 can emit infrared rays to improve the parallelism of the docking between the car body 1 and the floor. After the car body 1 stops, the second electric guide rail 305 is started to slide the second slider 306 on its circumferential surface, and drive the rotating shaft and the extension plate 309 to move outside the car body 1 through the movable groove 303. At the same time, the gear 307 will engage with the rack 308, drive the rack 308, the rotating shaft and the extension plate 309 to rotate, and cause the extension plate 309 to tilt until it rests on the ground of the floor, so as to facilitate the entry and exit of personnel and equipment.

[0109] For details, please refer to Figure 1-9 One end of each of the two spray pipes 110 is fixedly connected with a diffusion nozzle 111 . Both of the two diffusion nozzles 111 are conical and are located at the lower side of the discharge fan 112 .

[0110] In this embodiment: the area over which the spray pipe 110 sprays liquid can be further expanded by using a plurality of diffusion nozzles 111, thereby improving the sterilization and deodorization effects.

[0111] The working principle and use process of the present invention are as follows: when the car body 1 is used daily, the electric roller 208 is started, and the electric roller 208 drives the exhaust fan 112 to rotate in the two air dispersion plates 106, so that the gas in the car body 1 and the external gas are exchanged, so as to reduce the odor in the car body 1. At the same time, when the electric roller 208 rotates, the rotational force of the electric roller 208 can be transmitted to the screw rod 211 through the two I-shaped wheels 2 and the transmission belt 201, so that the screw rod 211 rotates, and then the screw rod nut 2 connected to the threaded surface of the screw rod 211 is rotated. 04 will descend, and then the external block 205 and the downward pressure rod 206 will descend. At this time, the descent of the downward pressure rod 206 will drive the downward pressure block 210 to descend, squeeze the fragrance or sterilization and deodorization liquid in the spray pipe 110, and spray it through the spray pipe 110. The sprayed liquid can also expand the area through the discharge fan 112 to remove the odor in the car body 1, thereby improving the user experience. The liquid in the liquid storage tank 103 can be transferred to the spray pipe 110 through the transmission pipe 109, which is convenient for timely replenishment of the liquid in the spray pipe 110.

[0112] In this embodiment: an intelligent control method for an intelligent car, the balance between energy saving and comfort is achieved through an intelligent control algorithm, and the specific implementation process is as follows:

[0113] Step 1: Data collection;

[0114] Real-time collection of environmental parameters including temperature, humidity, CO2 concentration and energy consumption data; recording frequency of use, number of passengers and instant energy consumption each time the elevator is used; collecting passenger feedback data on environmental comfort;

[0115] Step 2: Use transfer learning to process and analyze time series data;

[0116]

[0117] in, Indicates source domain dataset, subscript or superscript represents the source domain; Indicates source domain data samples, including feature vectors and the corresponding labels ; Indicates Feature vectors of source domain data samples; Indicates Labels of source domain data samples; Represents the number of samples in the source domain dataset;

[0118]

[0119] in, Indicates the target domain dataset, subscript or superscript represents the target domain; Indicates target domain data samples, including feature vectors and the corresponding labels ; Indicates Feature vectors of target domain data samples; Indicates The labels of target domain data samples; Represents the number of samples in the target domain dataset;

[0120] Train a deep neural network model on the source domain data:

[0121]

[0122] in, Represents the source domain function output; Represents input data; represents the parameters of the source domain neural network; Represents a neural network, through the parameters and input Perform forward propagation and obtain the output;

[0123] Fine-tune the pre-trained model on the target domain data to adapt it to the specific environment of the elevator:

[0124]

[0125] in, It is a function of the target domain, representing the forward propagation process of a neural network; Represents input data; represents the parameters of the target domain neural network, Indicates the change in parameter, used to change the source domain parameter Generate target domain parameters , represents a neural network;

[0126] Step 3: Define the transfer learning fine-tuning objective function;

[0127] In the fine-tuning phase, by adjusting the parameters of the source model To adapt to the target task;

[0128]

[0129] in, is a function of the target domain, represents the source domain function output, Represents input data; represents the parameters of the source domain neural network; represents the parameters of the target domain neural network, Indicates the change in parameter;

[0130]

[0131] in, represents the sample-level loss function, is a function of the target domain, represents the parameters of the target domain neural network, Indicates the target domain data The input features of samples, Indicates the target domain data The labels of the samples;

[0132] Define the following fine-tuning objective function;

[0133]

[0134] in, represents the overall loss function; Represents the sample-level loss function, which evaluates the difference between the predicted value and the actual value; Indicates the number of target domain data samples; represents the parameters of the target domain neural network, represents the parameters of the source domain neural network, is a function of the target domain, Indicates the target domain data The input features of samples, Indicates the target domain data The labels of the samples; Represents the regularization coefficient, which controls the degree of deviation between the fine-tuning parameters and the pre-training parameters; Represents the regularization term to prevent the model from overfitting;

[0135] Use gradient descent to update the parameters to minimize the loss function:

[0136]

[0137] in, Indicates the number of target domain data samples; Representation function In the target domain, samples The predicted value of Indicates that it is a sample The true label of Indicates samples, is a function In the parameters Next pair of samples The gradient of represents the regularization coefficient, represents the parameters of the source domain neural network; Represents the loss function About parameters The gradient of

[0138]

[0139] in, represents the parameters of the target domain neural network, is the learning rate, which controls the step size of each update; Represents the loss function About parameters The gradient of

[0140] Step 4: Use reinforcement learning to find the optimal actions to take in different states;

[0141] Use the historical operation data of the car to train the DQN model and find the optimal actions to take in different states;

[0142] Defining states is a vector containing multiple variables;

[0143]

[0144] in, Indicates the frequency of use of the car; Indicates the current time; Indicates the number of passengers in the car; Indicates the current energy consumption in the car;

[0145] Defining actions To adjust the power of each system, set the following operations:

[0146]

[0147] in, Indicates adjustment of the exhaust fan power; Indicates the power of adjusting the driving mechanism; Indicates adjustment of the power of lighting facilities; Indicates the power adjustment of air conditioning facilities;

[0148] Designing the reward function To minimize energy consumption while taking into account passenger comfort;

[0149]

[0150] in, represents the adjusted energy consumption; Indicates the degree of reduction in passenger comfort; Indicates the degree of improvement in passenger comfort; , , Represents the weight coefficient, which is used to balance the relationship between energy consumption and comfort;

[0151] Q Function Indicates in status Take Action The total expected return is:

[0152]

[0153] in, is in state Take action The state-action value function of Indicates expected value; is a discount factor that balances the importance of current rewards with future rewards; Represents the discount factor Step by time The power of is the time step Instant rewards; Indicates the starting state is , the starting action is ; and Represents the time step status and actions;

[0154] The core of DQN is to update the Q function to estimate Take action The expected utility of . The update rule of the Q function is as follows:

[0155]

[0156] in, represents the new Q value, Indicates the current Q value; Is the current state Take Action Immediate rewards after is a discount factor that weighs immediate rewards against future rewards; Yes Status After action The new state to which it is transferred later; Is in the new state One of all possible actions under is in the next state Select Action The maximum Q value when are the parameters of the target network, is the learning rate;

[0157] Use experience replay and the target network to update the Q function;

[0158]

[0159] in, represents the target Q value; Is the current state Take Action Immediate rewards after is a discount factor that weighs immediate rewards against future rewards; Yes Status After action The new state to which it is transferred later; Is in the new state One of all possible actions under is in the next state Select Action The maximum Q value when are the parameters of the target network, is the target Q network for the new state and actions The estimated Q value of

[0160] The loss function is the mean square error between the current Q value and the target Q value:

[0161]

[0162] in, is the loss function; Indicates expected value; is the target Q value; is the current Q network pair state and actions Q value estimation of ; are the parameters of the current Q network;

[0163] Update the parameters of the Q network using gradient descent:

[0164]

[0165] in, are the parameters of the current Q network; is the learning rate, which determines the step size of each update; is the loss function Relative to the parameters gradient.

[0166] Numerical analysis case:

[0167] The source domain dataset (commercial building) contains 10,000 samples, each of which includes temperature, humidity, CO2 concentration, and air quality index AQI. The target domain dataset (car) contains 200 samples, each of which includes temperature, humidity, CO2 concentration, and AQI.

[0168] λ=0.01: Regularization coefficient to prevent overfitting, learning rate α=0.001. Through transfer learning, the loss value of the model gradually decreases, indicating that the fine-tuned model is more suitable for the target domain data and can better predict the air quality in the cabin.

[0169] Current state s=(20,12:00,15,3.0);

[0170] The available action sets are:

[0171] a1=(80,60,70,90);

[0172] a2=(60,50,50,70);

[0173] a3=(40,40,30,50);

[0174] Calculate the Q value of each action: Q(s,a1)=30 Q(s,a2)=40 Q(s,a3)=50;

[0175] The optimal action is a3=(40,40,30,50);

[0176] By selecting the optimal action a3 and adjusting the power of the exhaust fan, drive mechanism, lighting and air conditioning, the following effects can be achieved:

[0177] Exhaust fan power 40; drive mechanism power 40; lighting power 30; air conditioning power 50.

[0178] The result is a reduction in immediate energy consumption and improved passenger comfort, achieving a balance between energy saving and comfort.

[0179] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent car, characterized in that: include: A car body (1), wherein a switch slot (101) is provided at one end of the car body (1); Two air dispersion plates (106), both of which are fixedly connected to the switch slot (101); An exhaust fan (112), the exhaust fan (112) being rotatably connected to adjacent ends of two air dispersion plates (106) via a motorized roller (208); Two sets of driving mechanisms, both sets of driving mechanisms are arranged on the upper side of the car body (1); and Two groups of spraying mechanisms, each group of the spraying mechanisms is composed of a liquid storage tank (103), two spraying pipes (110) and two transmission pipes (109), the liquid storage tank (103) is fixedly connected to one end of the car body (1), the two transmission pipes (109) are fixedly connected to one end of the car body (1), and the two transmission pipes (109) are communicated with the liquid storage tank (103), the two spraying pipes (110) are fixedly connected to the upper inner wall of the car body (1), and the two spraying pipes (110) are fixedly connected to the two transmission pipes (109) respectively; Each group of the driving mechanisms comprises two downward pressing blocks (210), a screw rod (211), a screw rod nut (204), two external blocks (205), two downward pressing rods (206), two downward pressing pipes (207), two I-shaped wheels (2) and a transmission belt (201); the two downward pressing pipes (207) are respectively fixedly connected to the upper ends of the two spray pipes (110); the screw rod (211) is rotatably connected to the upper end of the car body (1); the two I-shaped wheels (2) are respectively fixedly connected to the electric rollers (201); 8) and the circumferential surface of the screw rod (211), the transmission belt (201) is transmission-connected to the circumferential surface of the two I-shaped wheels (2), the screw rod nut (204) is threadedly connected to the circumferential surface of the screw rod (211), the two external blocks (205) are respectively fixedly connected to the two ends of the screw rod nut (204), the two lower pressure rods (206) are respectively fixedly connected to the lower ends of the two external blocks (205), and the two lower pressure blocks (210) are respectively fixedly connected to the lower ends of the two lower pressure rods (206); The upper end of the car body (1) is fixedly connected to a plurality of vertical rods (203), the upper ends of the plurality of vertical rods (203) are fixedly connected to a top plate (202), the upper end of the lead screw nut (204) is provided with a plurality of stabilizing holes, and the lead screw nut (204) is slidably connected to the circumferential surfaces of the plurality of vertical rods (203) through the plurality of stabilizing holes.

2. The intelligent elevator car according to claim 1, characterized in that: The upper end of the electric roller (208) is fixedly connected to an anti-slip disk (209), and the diameter of the anti-slip disk (209) is larger than that of the electric roller (208).

3. The intelligent car according to claim 2, characterized in that: One end of the liquid storage box (103) is provided with a supplementing groove, and one end of the liquid storage box (103) is rotatably connected to a sealing plate (104) via a rotating shaft, and one end of the sealing plate (104) is provided with a hand grip groove (105).

4. The intelligent car according to claim 3, characterized in that: The upper and lower inner walls of the two switch slots (101) are each provided with a sealing slot (3); a first electric guide rail (301) is fixedly connected between the inner walls on both sides of the plurality of sealing slots (3); a first slider (302) is slidably connected to the circumferential surfaces of the plurality of first electric guide rails (301); and the plurality of first sliders (302) are respectively fixedly connected to the two switch doors (102).

5. The intelligent car according to claim 4, characterized in that: A movable groove (303) is provided at one end of the elevator car body (1), an extension plate (309) is slidably connected in the movable groove (303), both ends of the extension plate (309) are rotatably connected to gears (307) via a rotating shaft, and two racks (308) are fixedly connected to the lower inner wall of the movable groove (303), and the two gears (307) are respectively meshed with the two racks (308).

6. The intelligent car according to claim 5, characterized in that: Two connection plates (107) are fixedly connected to one end of the car body (1), and a plurality of infrared calibrators (108) are provided at one end of the two connection plates (107).

7. The intelligent car according to claim 6, characterized in that: One end of each of the two spray pipes (110) is fixedly connected to a diffusion nozzle (111); both of the two diffusion nozzles (111) are conical, and both of the two diffusion nozzles (111) are located at the lower side of the discharge fan (112).

8. The intelligent car according to claim 7, characterized in that: Both inner walls of the movable groove (303) are provided with ejection grooves (304), and the inner walls of both sides of the two ejection grooves (304) are fixedly connected with second electric guide rails (305), and the circumferential surfaces of the two second electric guide rails (305) are slidably connected with second sliders (306), and the two second sliders (306) are rotatably connected to the two rotating shafts respectively.

9. The intelligent control method of an intelligent car according to claim 1, characterized in that ,Through intelligent algorithms, a balance between energy saving and comfort is achieved. The specific implementation process is as follows: Step 1: Collect environmental parameters including temperature, humidity, CO2 concentration and energy consumption data in real time; record the frequency of use, number of passengers and instant energy consumption each time the elevator is used; collect feedback data on environmental comfort from passengers; Step 2: Use transfer learning to process and analyze time series data; in, Indicates source domain dataset, subscript or superscript represents the source domain; Indicates source domain data samples, including feature vectors and the corresponding labels ; Indicates Feature vectors of source domain data samples; Indicates Labels of source domain data samples; Represents the number of samples in the source domain dataset; in, Indicates the target domain dataset, subscript or superscript represents the target domain; Indicates target domain data samples, including feature vectors and the corresponding labels ; Indicates Feature vectors of target domain data samples; Indicates The labels of target domain data samples; Represents the number of samples in the target domain dataset; Train a deep neural network model on the source domain data: in, Represents the source domain function output; Represents input data; represents the parameters of the source domain neural network; Represents a neural network, through the parameters and input Perform forward propagation and obtain the output; Fine-tune the pre-trained model on the target domain data to adapt it to the specific environment of the elevator: in, It is a function of the target domain, representing the forward propagation process of a neural network; Represents input data; represents the parameters of the target domain neural network, Indicates the change in parameter, used to change the source domain parameter Generate target domain parameters , represents a neural network; Step 3: Define the transfer learning fine-tuning objective function; In the fine-tuning phase, by adjusting the parameters of the source model To adapt to the target task; in, is a function of the target domain, represents the source domain function output, Represents input data; represents the parameters of the source domain neural network; represents the parameters of the target domain neural network, Indicates the change in parameter; in, represents the sample-level loss function, is a function of the target domain, represents the parameters of the target domain neural network, Indicates the target domain data The input features of samples, Indicates the target domain data The labels of the samples; Define the following fine-tuning objective function; in, represents the overall loss function; Represents the sample-level loss function, which evaluates the difference between the predicted value and the actual value; Indicates the number of target domain data samples; represents the parameters of the target domain neural network, represents the parameters of the source domain neural network, is a function of the target domain, Indicates the target domain data The input features of samples, Indicates the target domain data The labels of the samples; Represents the regularization coefficient, which controls the degree of deviation between the fine-tuning parameters and the pre-training parameters; Represents the regularization term to prevent the model from overfitting; Use gradient descent to update the parameters to minimize the loss function: in, Indicates the number of target domain data samples; Representation function In the target domain, samples The predicted value of Indicates that it is a sample The true label of Indicates samples, is a function In the parameters Next pair of samples The gradient of represents the regularization coefficient, represents the parameters of the source domain neural network; Represents the loss function About parameters The gradient of in, represents the parameters of the target domain neural network, is the learning rate, which controls the step size of each update; Represents the loss function About parameters The gradient of Step 4: Use reinforcement learning to find the optimal actions to take in different states; Use the historical operation data of the car to train the DQN model and find the optimal actions to take in different states; Defining states is a vector containing multiple variables; in, Indicates the frequency of use of the car; Indicates the current time; Indicates the number of passengers in the car; Indicates the current energy consumption in the car; Defining actions To adjust the power of each system, set the following operations: in, Indicates adjustment of the exhaust fan power; Indicates the power of adjusting the driving mechanism; Indicates adjustment of the power of lighting facilities; Indicates the power adjustment of air conditioning facilities; Designing the reward function To minimize energy consumption while taking into account passenger comfort; in, represents the adjusted energy consumption; Indicates the degree of reduction in passenger comfort; Indicates the degree of improvement in passenger comfort; , , Represents the weight coefficient, which is used to balance the relationship between energy consumption and comfort; Q Function Indicates in status Take Action The total expected return is: in, is in state Take action The state-action value function of Indicates expected value; is a discount factor that balances the importance of current rewards with future rewards; Represents the discount factor Step by time The power of is the time step Instant rewards; Indicates the starting state is , the starting action is ; and Represents the time step status and actions; The core of DQN is to update the Q function to estimate Take action The expected utility of Q function is as follows: in, represents the new Q value, Indicates the current Q value; Is the current state Take Action Immediate rewards after is a discount factor that weighs immediate rewards against future rewards; Yes Status After action The new state to which it is transferred later; Is in the new state One of all possible actions under is in the next state Select Action The maximum Q value when are the parameters of the target network, is the learning rate; Use experience replay and the target network to update the Q function; in, represents the target Q value; Is the current state Take Action Immediate rewards after is a discount factor that weighs immediate rewards against future rewards; Yes Status After action The new state to which it is transferred later; Is in the new state One of all possible actions under is in the next state Select Action The maximum Q value when are the parameters of the target network, is the target Q network for the new state and actions The estimated Q value of The loss function is the mean square error between the current Q value and the target Q value: in, is the loss function; Indicates expected value; is the target Q value; is the current Q network pair state and actions Q value estimation of ; are the parameters of the current Q network; Update the parameters of the Q network using gradient descent: in, are the parameters of the current Q network; is the learning rate, which determines the step size of each update; is the loss function Relative to the parameters gradient.

Images