Ventilation system, learning device, and inference device

Abstract

The present disclosure relates to a ventilation system for maintaining an air environment in a parking place by supplying the air from the outside and exhausting the air to the outside. The ventilation system includes an air supply blower for supplying the external air from an air supply port to a parking place, an air exhaust blower for exhausting the air in the parking place to the outside from an air exhaust port, a conveyance blower for conveying an air flow from the air supply blower to the air exhaust blower, an imaging unit for imaging the inside of the parking place and outputting image data, and a control unit for detecting a movement of a vehicle on the basis of the image data output from the imaging unit and changing an air volume of the conveyance blower.

Description

[[Technical Field]]

[0001] The present disclosure relates to a ventilation system, a learning device, and an inference device. [[Background Art]]

[0002] In a parking lot installed underground or indoors, ventilation via duct piping does not sufficiently ventilate the center or intricate areas of the parking lot, and air containing exhaust gas may stagnate in these areas. For this reason, ductless ventilation systems including conveying blowers that convey air from an air supply blower to an exhaust blower are widely used. Patent Document 1 discloses a ventilation system including: a CO sensor installed near an exhaust blower and configured to output a signal indicating a CO gas concentration in the vicinity thereof; and a blower control device that controls the operation of at least any one of the air supply blower, the exhaust blower, and the conveying blower based on an output of the CO sensor, wherein the ventilation system is characterized in that a plurality of conveying blowers are combined to connect air flows, thereby generating an air flow directed from the air supply blower to the exhaust blower. [[Prior Art Literature]] [[Patent Literature]]

[0003] [[Patent Document 1]] Japanese Patent No. 6807373 [[Summary of the Invention]] [[Problem to be Solved by the Invention]]

[0004] When a CO sensor is installed in the vicinity of an exhaust blower, a time difference occurs before exhaust gas generated in the parking lot reaches the CO sensor near the exhaust blower. Therefore, exhaust gas cannot be discharged quickly by controlling the blower based on the output of the CO sensor. Accordingly, an object of the present disclosure is to provide a ventilation system capable of quickly discharging exhaust gas in a parking lot. [[Means for Solving the Problem]]

[0005] The ventilation system disclosed herein comprises an air supply fan that supplies outside air to the parking lot from an air intake, an exhaust fan that discharges air from the parking lot to the outside from an exhaust port, and a transport fan that transports airflow from the air supply fan to the exhaust fan. The ventilation system comprises an imaging unit that captures images of the inside of the parking lot and outputs image data, and based on the image data output from the imaging unit Vehicles moving within the parking lot Detect, The size of the detected moving vehicle is further detected, and according to the size of the vehicle... Control to change the airflow of the conveyor fan. department Prepare further. [Effects of the Invention]

[0006] According to the ventilation system described herein, based on the image data output from the imaging unit Vehicles moving within the parking lot Detect, The size of the detected moving vehicle is further detected, and according to the size of the vehicle... By changing the airflow of the transport blower, exhaust fumes can be quickly discharged within the parking lot. [Brief explanation of the drawing]

[0007] [Figure 1] This is a schematic diagram showing a ventilation system 100 according to Embodiment 1, which is installed in a parking lot. [Figure 2] This is a schematic diagram showing the signal communication connection in the ventilation system 100. [Figure 3] This is a block diagram showing the functional configuration of the control unit 30 of the ventilation system 100. [Figure 4] This is a block diagram showing the hardware configuration of the control unit 30 of the ventilation system 100. [Figure 5] A flowchart illustrating the operation of the control unit 30 of the ventilation system 100. [Figure 6] This is a schematic diagram showing multiple areas set up within a parking lot where a ventilation system 100 is installed. [Figure 7] Figure 5 is a flowchart illustrating the operation of step S4. [Figure 8] This is a schematic diagram showing a ventilation system 110 according to Embodiment 2, which is installed in a parking lot. [Figure 9] Figure 8 is a schematic diagram showing the airflow system generated by the ventilation system 110. [Figure 10] This is a schematic diagram showing another example of the arrangement of the ventilation system 110 in a parking lot. [Figure 11] This is another schematic diagram showing the airflow system generated by the ventilation system 110 in Figure 10. [Figure 12] This is a schematic diagram showing a ventilation system 120 according to Embodiment 3, which is installed in a parking lot. [Figure 13] This is a schematic diagram showing a ventilation system 130 according to Embodiment 4, which is installed in a parking lot. [Figure 14] This is a block diagram showing the configuration of the learning device 200 according to Embodiment 4. [Figure 15] This is a flowchart showing the operation of the learning device 200. [Figure 16] This is a block diagram showing the configuration of the inference device 300 according to Embodiment 4. [Figure 17] This is a flowchart showing the operation of the inference device 300 and the ventilation system 130. [Modes for carrying out the invention]

[0008] Embodiments of this disclosure will be described below with reference to the drawings. In the drawings and the following description, identical and substantially identical components will be denoted by the same reference numerals, and the descriptions of components denoted by the same reference numerals will not be repeated.

[0009] Embodiment 1. FIG. 1 is a schematic diagram showing a ventilation system 100 according to Embodiment 1 of the present disclosure. The ventilation system 100 is installed in a highly airtight parking garage such as an underground parking garage or an indoor parking garage, performs ventilation inside the parking garage, and discharges exhaust gas out of the parking garage. The ventilation system 100 includes an air supply blower 5, an exhaust blower 6, and a plurality of conveying blowers 90. The planar shape of the parking garage is substantially rectangular, and the air supply port 1 and the exhaust port 2 are disposed at diagonal positions of the parking garage. The air supply port 1 and the air supply blower 5 are disposed near the entrance 8, and the air supply blower 5 supplies outside air into the parking garage through the air supply port 1. The exhaust port 2 and the exhaust blower 6 are disposed at a corner near the exit 9, and the exhaust blower 6 discharges air from the parking garage to the outside through the exhaust port 2. The plurality of conveying blowers 90 are installed inside the parking garage, and convey air from the air supply blower 5 to the exhaust blower 6 to promote ventilation inside the parking garage.

[0010] In the figure, an example is shown of a parking garage provided with parking spaces 7 for 26 vehicles and a vehicle movement route 10 that branches into two from the vicinity of the entrance 8 and merges again near the exit 9. One of the two branched vehicle movement routes 10 is referred to as a vehicle movement route 10-1, and the other is referred to as a vehicle movement route 10-2, respectively. Conveying blowers 90-1 and 90-3 among the plurality of conveying blowers 90 are installed along the vehicle movement route 10-1, and conveying blowers 90-2 and 90-4 among the plurality of conveying blowers 90 are installed along the vehicle movement route 10-2. The conveying blowers 90-1 and 90-2 are installed adjacent to each other, and the conveying blowers 90-3 and 90-4 are disposed at diagonal positions of the parking garage.

[0011] The conveying blowers 90-1 and 90-3 generate airflows 12-1 and 12-3, respectively. The conveying blower 90-1 is installed on the windward side of the conveying blower 90-3, and the conveying blower 90-3 is disposed at a position that the airflow 12-1 reaches. Therefore, the conveying blowers 90-1 and 90-3 form an airflow directed from the air supply port 1 to the exhaust port 2 via the vehicle movement path 10-1. The conveying blowers 90-2 and 90-4 generate airflows 12-2 and 12-4, respectively. The conveying blower 90-2 is installed on the windward side of the conveying blower 90-4, and the conveying blower 90-4 is disposed at a position that the airflow 12-2 reaches. Therefore, the conveying blowers 90-2 and 90-4 form an airflow directed from the air supply port 1 to the exhaust port 2 via the vehicle movement path 10-2. Note that, hereinafter, unless otherwise specifically described, the conveying blower 90 refers to each of the conveying blowers 90-1 to 90-4.

[0012] The ventilation system 100 includes an imaging unit 20. The imaging unit 20 is installed in a parking lot, images the interior of the parking lot, and outputs image data thereof. The imaging unit 20 is disposed at a position where at least the vehicle movement paths 10-1 and 10-2 can be imaged. Specifically, the imaging unit 20 includes two cameras 20-1 and 20-2. The cameras 20-1 and 20-2 are respectively installed near the conveying blowers 90-3 and 90-4, and are disposed at diagonal positions of the parking lot relative to each other. The cameras 20-1 and 20-2 output image data of images obtained by imaging in respective directions that the cameras face.

[0013] The ventilation system 100 includes a control unit 30. There are no restrictions on the installation location, but the control unit 30 should be installed, for example, adjacent to the conveying fan 90-2. Figure 2 is a schematic diagram showing the signal communication connections in the ventilation system 100. The control unit 30 is connected to each of the supply fan 5, exhaust fan 6, conveying fans 90-1 to 90-4, camera 20-1, and camera 20-2 by wire or wireless. To simplify the wiring work, it is desirable that the control unit 30, supply fan 5, exhaust fan 6, conveying fans 90-1 to 90-4, camera 20-1, and camera 20-2 are each connected to the network 140 by wireless. Since the control unit 30 is adjacent to the conveying fan 90-2, it may also be connected to the conveying fan 90-2 by wire. The control unit 30 may also be installed inside the conveying fan 90-2.

[0014] The control unit 30 controls the operation of the supply air blower 5, the exhaust air blower 6, and the transport air blower 90. Regarding the control of the transport air blower 90, the control unit 30 detects the movement of the vehicle based on the image data output from the imaging unit 20 and changes the airflow rate of at least one of the transport air blowers 90-1 to 90-4. The airflow rate of the blower is the amount of airflow generated by the blower.

[0015] Figure 3 is a block diagram showing the functional configuration of the control unit 30. The control unit 30 includes a vehicle movement detection unit 31, a vehicle shape detection unit 32, and a blower control unit 33. The vehicle movement detection unit 31 detects the movement of a vehicle based on image data received from the imaging unit 20. The vehicle movement detection unit 31 also detects the position of the moving vehicle and outputs vehicle position information indicating the vehicle's position. The vehicle shape detection unit 32 detects the shape of the moving vehicle based on image data received from the imaging unit 20. The vehicle shape detection unit 32 outputs vehicle shape information indicating the size of the vehicle.

[0016] When the vehicle movement detection unit 31 detects vehicle movement, the blower control unit 33 identifies at least one of the transport blowers 90-1 to 90-4 based on the vehicle position information output from the vehicle movement detection unit 31. The blower control unit 33 also determines the airflow rate of the identified transport blower based on the vehicle shape information.

[0017] The blower control unit 33 transmits an airflow control command to the identified transport blower. The airflow control command includes airflow information specifying the airflow determined based on the vehicle shape information. Upon receiving the airflow control command, the transport blower changes its operation to the specified airflow based on the airflow information.

[0018] A blower has a motor and a fan, and generates airflow by rotating the fan with the motor. The airflow of the blower changes by changing the operating parameters of the blower (motor speed, power supplied to the motor, or power frequency, etc.). Therefore, airflow information is information for setting the operating parameters of the blower. In the case of a blower in which the airflow can be set in multiple stages, the airflow information may also be information that identifies an operating mode that represents one of the multiple configurable stages. Furthermore, the airflow information may also be information that represents the difference between the airflow before the change and the airflow after the change.

[0019] Figure 4 is a block diagram showing the hardware configuration of the control unit 30. As shown in Figure 4(a), the control unit 30 consists of a processor 70, a memory 71, and a bus 72. The processor 70 processes according to a program to realize the functions of the vehicle movement detection unit 31, the vehicle shape detection unit 32, and the blower control unit 33. The program is stored in the memory 71 and provided to the processor 70 via the bus 72. The processor 70 is, for example, a central processing unit (CPU), and the memory 71 is a storage device that includes volatile memory such as RAM (Random Access Memory) and non-volatile memory such as ROM (Read Only Memory) and flash memory.

[0020] Image data from the imaging unit 20 is received by an interface circuit (not shown) and stored in memory 71 via bus 72. The processor 70 reads the image data from memory 71 and processes it. The airflow control command generated by the processor 70 is transmitted via bus 72 to another interface circuit (not shown) and then to the transport control unit.

[0021] As shown in Figure 4(a), the configuration having a processor 70 and memory 71 can be implemented using a microcomputer or the like, which can also be used to control other electrical or electronic equipment. However, it is not limited to this, and the control unit 30 can also be implemented using a dedicated processing circuit 73, as shown in Figure 4(b).

[0022] In the ventilation system 100, the control unit 30 increases the airflow of one or more transport blowers when it detects vehicle movement from image data. Figure 5 is a flowchart showing the operation of the control unit 30 for this control. The start is when the ventilation system 100 is started after being installed in the parking lot, and the control unit 30 instructs the supply blower 5, exhaust blower 6, and transport blower 90 to operate at predetermined airflow rates. The transport blower 90 is controlled, for example, to operate at the minimum airflow rate within the range of configurable airflow rates.

[0023] In step S1, the vehicle movement detection unit 31 and the vehicle shape detection unit 32 each acquire image data from the imaging unit 20, that is, from both cameras 20-1 and 20-2. In step S2, the vehicle movement detection unit 31 detects vehicle movement by detecting changes in the vehicle's position in the moving image based on the image data. If there is no moving vehicle in the image data (No), the operation flow returns to step S1. If there is a moving vehicle (Yes), in step S3, the vehicle movement detection unit 31 detects the position of the moving vehicle.

[0024] Figure 6 is a schematic diagram showing multiple areas set up in the parking lot shown in Figure 1. Four areas A-1 to A-4 are set up on the vehicle movement path 10. Areas A-1 and A-3 are set up on the vehicle movement path 10-1, and area A-1 is upwind of the airflow generated by conveyor fans 90-1 and 90-3 compared to area A-3. Areas A-1 and A-3 are corresponding to conveyor fans 90-1 and 90-3, respectively. Areas A-2 and A-4 are set up on the vehicle movement path 10-2, and area A-2 is upwind of the airflow generated by conveyor fans 90-2 and 90-4 compared to area A-4. Areas A-2 and A-4 are corresponding to conveyor fans 90-2 and 90-4, respectively.

[0025] As one method of detecting the vehicle's position, the vehicle movement detection unit 31 detects from the image data which of areas A-1 to A-4 the vehicle is moving through. The vehicle's movement in areas A-1 and A-3 is determined from the image data of camera 20-1, and the vehicle's movement in areas A-2 and A-4 is determined from the image data of camera 20-2. The vehicle movement detection unit 31 generates vehicle position information indicating the area the vehicle is moving through. The detection of vehicle movement and the area the vehicle is moving through from the image data is performed using known image recognition techniques. The image data can be moving images with a frame rate low enough to detect the vehicle's movement.

[0026] Next, in step S4, the blower control unit 33 identifies the conveying blower 90-1 to 90-4 whose airflow should be increased based on the vehicle position information output from the vehicle movement detection unit 31.

[0027] Figure 7 is a flowchart showing the operation of step S4 in more detail. In step S4-1, the blower control unit 33 determines whether or not a vehicle is present in area A-1. If a vehicle is present in area A-1, exhaust gas tends to accumulate in area A-1, so it is necessary to increase the airflow in area A-1. Furthermore, in order to promote exhaust outside the parking lot, it is desirable to increase the airflow not only in area A-1 but also in area A-3 located downstream. Therefore, if it is determined that a vehicle is present in area A-1 (Yes), in step S4-2, the blower control unit 33 identifies the transport blowers 90-1 and 90-3. Step S4-2 is performed regardless of whether or not other vehicles are present in area A-3. After that, the operation flow moves on to step S4-5.

[0028] If it is determined in step S4-1 that no vehicle exists (No), then in step S4-3, the blower control unit 33 determines whether or not a vehicle exists in area A-3. If it is determined that a vehicle exists (Yes), then in step S4-4, the blower control unit 33 identifies the transport blower 90-3. In this case, since there is no need to increase the airflow in area A-1, the transport blower 90-1 is not identified. The operation flow then proceeds to step S4-5. Similarly, if it is determined in step S4-3 that no vehicle exists (No), the operation flow also proceeds to step S4-5.

[0029] In step S4-5, the blower control unit 33 determines whether or not a vehicle is present in area A-2. If a vehicle is present in area A-2, exhaust gas tends to accumulate in area A-2, so it is necessary to increase the airflow in area A-2. Furthermore, to promote exhaust outside the parking lot, it is desirable to increase the airflow not only in area A-2 but also in area A-4 located downstream. Therefore, if it is determined that a vehicle is present in area A-2 (Yes), in step S4-6, the blower control unit 33 identifies the transport blowers 90-2 and 90-4. Step S4-6 is performed regardless of whether or not another vehicle is present in area A-4. After that, the operation flow proceeds to step S5.

[0030] If it is determined in step S4-5 that no vehicle exists (No), then in step S4-7, the blower control unit 33 determines whether or not a vehicle exists in area A-4. If it is determined that a vehicle exists (Yes), then in step S4-8, the blower control unit 33 identifies the transport blower 90-4. In this case, since there is no need to increase the airflow in area A-2, the transport blower 90-2 is not identified. The operation flow then proceeds to step S5. Similarly, if it is determined in step S4-7 that no vehicle exists (No), the operation flow also proceeds to step S5.

[0031] Returning to Figure 5, in step S5, the vehicle shape detection unit 32 determines the size of each of the vehicles located in areas A-1 to A-4. The vehicle shape detection unit 32 detects the shape of the vehicles using existing image recognition technology and determines the size of the vehicles from that shape. In step S6, the blower control unit 33 receives information about the size of the vehicles from the vehicle shape detection unit 32 and determines the airflow rate of the transport blowers identified in step S4 according to the size of the vehicles. The determined airflow rate is greater than the airflow rate of the transport blowers 90-1 to 90-4 at the start of operation of the ventilation system 100. Since larger vehicles produce a larger volume of exhaust gas, it is desirable to increase the airflow rate of the transport blowers as the vehicle size increases. For example, the blower control unit 33 may have multiple thresholds set for the size of the vehicles, and the blower control unit 33 may determine the airflow rate so that it increases each time the determined size of a vehicle exceeds a threshold.

[0032] In step S7, the blower control unit 33 transmits an airflow control command to the conveyor blower identified in step S4 and controls the conveyor blower to change to the airflow determined in step S6. Upon receiving the airflow control command, the conveyor blower changes its airflow based on the command. If the vehicle is in area A-1, conveyor blowers 90-1 and 90-3 receive the airflow control command. If the vehicle is in area A-2, conveyor blowers 90-2 and 90-4 receive the airflow control command. If the vehicle is in area A-3, conveyor blower 90-3 receives the airflow control command. If the vehicle is in area A-4, conveyor blower 90-4 receives the airflow control command. In all cases, the conveyor blower that receives the airflow control command increases its airflow.

[0033] If multiple vehicles are located in the same area, the blower control unit 33 should control the airflow of the transport blower corresponding to that area so that it becomes the largest airflow among the airflows determined for each of the multiple vehicles. Alternatively, the blower control unit 33 can also control the airflow of the transport blower so that it becomes the largest airflow, further increased according to the number of vehicles.

[0034] In step S8, the blower control unit 33 determines whether a predetermined time has elapsed since the airflow rate of the conveyor blower was changed in step S7. If the airflow rate has been changed, the conveyor blower will continue to operate at the airflow rate changed in step S7 until the predetermined time has elapsed (No). If the predetermined time has elapsed since the airflow rate was changed (Yes), in step S9, the blower control unit 33 controls the airflow rate of the conveyor blower to return it to the airflow rate before the change, i.e., the airflow rate at the start of operation of the ventilation system 100. The conveyor blower receives an airflow rate control command from the blower control unit 33 and returns the airflow rate to its original value based on that airflow rate control command. The operation flow then returns to step S1, and the control unit 30 repeats the operations of steps S1 to S9. Each of steps S1 to S9 can be rephrased as an operation performed by the control unit 30.

[0035] As described above, according to the ventilation system 100 of this embodiment, the control unit 30 detects the movement of the vehicle based on image data and changes the airflow rate of at least one transport blower. Because image data is used, the control unit 30 can quickly detect that the vehicle is generating exhaust gas and quickly discharge the exhaust gas.

[0036] Furthermore, according to the ventilation system 100, the control unit 30 detects the position of the moving vehicle from the image data, identifies at least one of the multiple transport blowers 90-1 to 90-4 according to the detected position, and increases the airflow of the identified transport blower. In areas where there is not much exhaust gas generation, it is not necessary to increase the airflow of the transport blower. In other words, excessive ventilation can be suppressed, thus saving electricity. This leads to a reduction in the running costs of the ventilation system 100.

[0037] The control unit 30 identifies a conveying blower that generates airflow at the detected location of the vehicle, thereby quickly identifying areas in the parking lot where exhaust gases need to be removed. The control unit 30 also identifies conveying blowers located further downstream as conveying blowers that increase airflow, thus efficiently guiding the exhaust gases out of the parking lot.

[0038] Since the imaging unit 20, which captures images, is less expensive than the CO sensor that detects CO concentration, controlling the airflow of the conveying blower based on image data also contributes to reducing the initial cost of the ventilation system 100.

[0039] The following are some variations of Embodiment 1.

[0040] Modification (1): In step S2, the vehicle movement detection unit 31 detects a change in the vehicle's position in the moving image, but the movement of the vehicle may also be detected by detecting the vehicle's position. If a vehicle is present in areas A-1 to A-4 set on the vehicle movement path 10, it can be assumed that the vehicle is moving along the vehicle movement path 10. Therefore, in step S2, the vehicle movement detection unit 31 determines from the image data whether or not the vehicle is present in any of areas A-1 to A-4. If it is determined that a vehicle is present in any of areas A-1 to A-4, the vehicle movement detection unit 31 is deemed to have detected the movement of the vehicle. Then, in step S3, the vehicle movement detection unit 31 generates vehicle position information indicating the area (any of areas A-1 to A-4) in which the vehicle is present. On the other hand, if the vehicle is not present in the image, or if it is present but not in any of areas A-1 to A-4, the vehicle movement detection unit 31 returns to step S1, assuming that there is no moving vehicle. The image data used to detect the movement of the vehicle at this time may be a still image.

[0041] Modification (2): Step S4 is performed by the blower control unit 33, but it may also be performed by the vehicle movement detection unit 31. In this case, the blower control unit 33 receives information from the vehicle movement detection unit 31 to identify the transport blower and transmits an airflow control command to the identified transport blower. Also, step S6 is performed by the blower control unit 33, but it may also be performed by the vehicle shape detection unit 32. In this case, the blower control unit 33 includes the airflow determined by the vehicle movement detection unit 31 as airflow information in the airflow control command.

[0042] Modification (3): The control unit 30 does not necessarily require the vehicle shape detection unit 32. The blower control unit 33 can also control the transport blower to increase the airflow by a predetermined amount regardless of the size of the vehicle. This reduces the processing load on the control unit 30.

[0043] Modification (4): The number and location of transport blowers are changed according to the size of the parking lot. In some parking lots, only one transport blower may be installed. In this case, when the control unit 30 detects vehicle movement, step S4 for identifying the transport blower is unnecessary, and it is sufficient to increase the airflow of the single transport blower. The number of cameras is also changed according to the size of the parking lot. If the parking lot is small, one camera may suffice as the imaging unit 20.

[0044] Embodiment 2. Figure 8 is a schematic diagram of the ventilation system 110 according to Embodiment 2. The ventilation system 110 includes an air supply fan 5, an exhaust fan 6, transport fans 90-1 to 90-4, cameras 20-1 and 20-2, and a control unit 30. The difference from Embodiment 1 is the method of identifying the transport fans in the control unit 30; the other configurations and operations are the same as in Embodiment 1.

[0045] The control unit 30 has a parking map of the parking lot where the ventilation system 110 is installed. The parking map is pre-stored in the memory of the control unit 30. In the parking map, for example, a two-dimensional coordinate system is set on the plane of the parking lot, and the positions of the transport blowers 90-1 to 90-4, cameras 20-1 and 20-2, and exhaust port 2 are pre-stored in the control unit 30 in the form of two-dimensional coordinates.

[0046] As shown in Figure 9, the air supplied from the supply air blower 5 is broadly divided into two systems: an airflow that goes from the conveying blower 90-1 through the conveying blower 90-3 to the exhaust blower 6, and an airflow that goes from the conveying blower 90-2 through the conveying blower 90-4 to the exhaust blower 6.

[0047] The straight line connecting conveyor fan 90-1 and conveyor fan 90-3 is defined as airflow path 91A and corresponds to conveyor fan 90-1. Airflow path 91A is set by the coordinates of two points: conveyor fan 90-1 and conveyor fan 90-3. The straight line connecting conveyor fan 90-3 and exhaust port 2 is defined as airflow path 91C and corresponds to conveyor fan 90-3. Airflow path 91C is set by the coordinates of two points: conveyor fan 90-3 and exhaust port 2. The straight line connecting conveyor fan 90-2 and conveyor fan 90-4 is defined as airflow path 91B and corresponds to conveyor fan 90-2. Airflow path 91B is set by the coordinates of two points: conveyor fan 90-2 and conveyor fan 90-4. The straight line connecting the conveying fan 90-4 and the exhaust port 2 is defined as the airflow path 91D and corresponds to the conveying fan 90-4. The airflow path 91D is determined by the coordinates of the two points: the conveying fan 90-4 and the exhaust port 2.

[0048] The parking lot map also includes information indicating the direction of airflow in air passages 91A to 91D. Therefore, the control unit 30 can determine from the parking lot map that conveyor blower 90-1 is located upstream of conveyor blower 90-3, and that conveyor blower 90-2 is located upstream of conveyor blower 90-4.

[0049] If vehicle 11 is present in the parking lot, the vehicle movement detection unit 31 detects the position of vehicle 11 as a two-dimensional coordinate system based on image data acquired from cameras 20-1 and 20-2. The vehicle movement detection unit 31 outputs this detected position of vehicle 11 as vehicle position information. This position detection by the vehicle movement detection unit 31 corresponds to step S3 in Figure 5.

[0050] Here, the shortest distance from vehicle 11 to airflow path 91A is defined as airflow path distance 92A. The shortest distance from vehicle 11 to airflow path 91B is defined as airflow path distance 92B. The shortest distance from vehicle 11 to airflow path 91C is defined as airflow path distance 92C. The shortest distance from vehicle 11 to airflow path 91D is defined as airflow path distance 92D. The blower control unit 33 calculates the airflow path distances 92A to 92D based on the vehicle position information received from the vehicle movement detection unit 31. The blower control unit 33 identifies the airflow path with the shortest airflow path distance among the airflow paths 91A to 91D calculated by the vehicle movement detection unit 31, and identifies the conveyor blower corresponding to that identified airflow path. If there is another conveyor blower downstream of that conveyor blower, that conveyor blower is also identified. This identification of the conveyor blower by the blower control unit 33 corresponds to step S4 in Figure 5.

[0051] To understand the airflow distance, Figure 8 shows the airflow distances 92A to 92D when the vehicle 11 is parked in the parking space 7. However, since step S3 is performed after step S2, the airflow distances 92A to 92D are not calculated when the vehicle 11 is parked. However, step S3 may be performed before step S2. That is, the vehicle movement detection unit 31 detects the position of the vehicle 11 and calculates the airflow distances 92A to 92D, and then detects the movement of the vehicle 11. When the movement of the vehicle 11 is detected, in step S4, the blower control unit 33 identifies the transport blower based on the airflow distances 92A to 92D relative to the vehicle 11. The operation of the ventilation system 110 from this point onward is the same as step S5 of Embodiment 1, so its explanation is omitted.

[0052] Thus, the ventilation system 110 is characterized by having multiple airflow paths set for each of the airflows generated by the transport blowers 90-1 to 90-4, and the control unit 30 identifies the transport blower corresponding to the airflow path closest to the vehicle as one whose airflow should be increased.

[0053] The following are some examples of modifications of Embodiment 2.

[0054] Modification (1): Similar to modification (1) of Embodiment 1, the control unit 30 may detect vehicle movement based on the vehicle's position. For example, if the vehicle is not present in any of the parking spaces 7, it can be assumed that the vehicle is moving along the vehicle movement path 10. Therefore, the parking map is assumed to contain information about the area of ​​all parking spaces 7. In step S3, the vehicle movement detection unit 31 detects the vehicle's position regardless of vehicle movement and generates vehicle position information as two-dimensional coordinates. The vehicle movement detection unit 31 then determines whether the detected vehicle position is included in the parking spaces 7. If it determines that the vehicle is included in any of the parking spaces 7, the vehicle movement detection unit 31 returns to step S1, assuming the vehicle is not moving. If it determines that the vehicle's position is not included in any of the parking spaces 7, the vehicle movement detection unit 31 moves to step S4, assuming the vehicle 11 is a moving vehicle. Vehicle movement detection is performed in step S3, and step S2 is unnecessary.

[0055] Modification (2): The ventilation system 110 is not limited to the case where the air intake 1 and exhaust 2 are located diagonally opposite each other, as shown in Figure 8. The ventilation system 110 may be located at two adjacent corners of the four corners of the parking lot's planar shape, for example, as shown in Figure 10, where the air intake 1 is located at the upper left corner and the exhaust 2 is located at the lower left corner. In that case, the conveying fan 90-1 is installed near the conveying fan 90-3 and blows air toward the conveying fan 90-2. Thus, as shown in Figure 11, the air supplied from the air intake fan 5 is broadly divided into two systems: an airflow from the conveying fan 90-3 toward the exhaust fan 6, and an airflow from the conveying fan 90-1 toward the exhaust fan 6 via the conveying fans 90-2 and 90-4.

[0056] Modification (3): Each of the modifications (2) to (4) of Embodiment 1 also applies to the ventilation system 110 of Embodiment 2.

[0057] Embodiment 3. Figure 12 is a schematic diagram of the ventilation system 120 according to Embodiment 3. The ventilation system 120 includes an air supply fan 5, an exhaust fan 6, transport fans 90-1 to 90-4, cameras 20-1 and 20-2, and a control unit 30. The difference from Embodiment 1 is the method of identifying the transport fans in the control unit 30; the other configurations and operations are the same as in Embodiment 1.

[0058] The installation locations of the air supply fan 5, exhaust fan 6, transport fans 90-1 to 90-4, camera 20-1, camera 20-2, and control unit 30 are the same as in Figure 8. When viewing the parking lot from above, the upper right corner of Figure 12 is called corner A, the upper left corner is called corner B, the lower right corner is called corner C, and the lower left corner is called corner D. The air supply port 1 and air supply fan 5 are located near corner A. Transport fans 90-1 and 90-2 are located in the passage near corner A. Transport fan 90-1 blows air toward corner B. Transport fan 90-2 blows air toward corner C. Transport fan 90-3 is located in the passage near corner B and blows air toward corner D. Transport fan 90-3 is located in the passage near corner C and blows air toward corner D. An exhaust port 2 and an exhaust fan 6 are located near corner D.

[0059] Camera 20-1 is positioned near corner B, and camera 20-2 is positioned near corner C. Therefore, the left side of the image captured by camera 20-1 becomes the right side of the image captured by camera 20-2, and the right side of the image captured by camera 20-1 becomes the left side of the image captured by camera 20-2. Cameras 20-1 and 20-2 have the same resolution.

[0060] The straight line 94A connecting camera 20-1 and camera 20-2 is a hypothetical line added for convenience. Similarly, the straight line 94B passing midway between camera 20-1 and camera 20-2, parallel to the parking lot floor, and perpendicular to straight line 94A is also a hypothetical line added for convenience. The area tangent to straight lines 94A and 94B, and located to the left of camera 20-1 in front of straight line 94B, is referred to as area 93A. The area tangent to straight lines 94A and 94B, and located to the right of camera 20-1 in front of straight line 94B, is referred to as area 93C. The area tangent to straight lines 94A and 94B, and located to the right of camera 20-2 in front of straight line 94B, is referred to as area 93B. The area adjacent to lines 94A and 94B, and located to the left of line 94B as viewed from camera 20-2, is called area 93D. For convenience in the diagram, areas 93A to 93D are shown as being separate from lines 94A and 94B, but as mentioned above, they are areas adjacent to lines 94A and 94B.

[0061] The vehicle movement detection unit 31 detects the vehicle's position in one of the areas 93A to 93D. First, the vehicle movement detection unit 31 detects the vehicle's position in the image captured by either camera 20-1 or 20-2. The vehicle's position in the image is detected as left-right information, indicating whether the vehicle is in the right half or left half of the image. Let's assume that one of the cameras is camera 20-2. If the vehicle is in the right half of the image captured by camera 20-2, it is determined that the vehicle is in one of the areas 93A or 93B, which are to the right of the straight line 94A. If the vehicle is in the left half of the image captured by camera 20-2, it is determined that the vehicle is in one of the areas 93C or 93D, which are to the left of the straight line 94A.

[0062] Vehicles in the image of camera 20-2 are also captured by the other camera 20-1. The vehicle movement detection unit 31 then compares the size of the vehicles captured by cameras 20-1 and 20-2, respectively. If the vehicle captured by camera 20-1 is larger, it indicates that the vehicle is closer to camera 20-1 from the straight line 94A. On the other hand, if the vehicle captured by camera 20-2 is larger, it indicates that the vehicle is closer to camera 20-2 from the straight line 94A.

[0063] Therefore, the vehicle movement detection unit 31 detects that the vehicle is in area 93A if it detects that the vehicle captured by camera 20-2 is to the right of the straight line 94A and that camera 20-1 captured a larger image than camera 20-2. The vehicle movement detection unit 31 detects that the vehicle is in area 93C if it detects that the vehicle captured by camera 20-2 is to the left of the straight line 94A and that camera 20-1 captured a larger image than camera 20-2. The vehicle movement detection unit 31 detects that the vehicle is in area 93B if it detects that the vehicle captured by camera 20-2 is to the right of the straight line 94A and that camera 20-2 captured a larger image than camera 20-1. The vehicle movement detection unit 31 detects that the vehicle is in area 93D if it detects that the vehicle captured by camera 20-2 is to the left of the straight line 94A and that camera 20-2 captured a larger image than camera 20-1. The vehicle movement detection unit 31 outputs the detected area as vehicle position information. The detection of the area by this vehicle movement detection unit 31 corresponds to step S3 in Figure 5.

[0064] In step S4, the blower control unit 33 identifies the conveying blowers whose airflow should be increased based on the vehicle position information output from the vehicle movement detection unit 31. If the vehicle position information indicates area 93A, the blower control unit 33 identifies conveying blower 90-1 and the downstream conveying blower 90-3 that generate airflow in area 93A. If the vehicle position information indicates area 93C, the blower control unit 33 identifies conveying blower 90-3 that generates airflow in area 93C. If the vehicle position information indicates area 93B, the blower control unit 33 identifies conveying blower 90-2 and the downstream conveying blower 90-4 that generate airflow in area 93B. If the vehicle position information indicates area 93D, the blower control unit 33 identifies conveying blower 90-4 that generates airflow in area 93D. This identification of conveying blowers by the blower control unit 33 corresponds to step S4 in Figure 5. The operation of the ventilation system 120 from this point onward is the same as in step S5 of Embodiment 1, so its explanation will be omitted.

[0065] As described above, the ventilation system 120 is characterized by comparing the size of vehicles captured by two cameras 20-1 and 20-2, and identifying a transport blower that should have its airflow increased according to the comparison results.

[0066] Furthermore, variations (2) to (4) of Embodiment 1 and variation (2) of Embodiment 2 also apply to the ventilation system 120 of Embodiment 3.

[0067] Embodiment 4. The learning device 200 and inference device 300 according to Embodiment 4 will now be described. Both the learning device 200 and the inference device 300 are learning and inference devices related to a ventilation system having a conveyor fan. The learning device 200 acquires learning data, determines the control method of the conveyor fan from the learning data using machine learning, and generates a trained model. Based on the trained model, the inference device 300 infers control information for controlling the conveyor fan from the input data.

[0068] Parking lots come in various shapes, and the arrangement of parking spaces 7 within them also varies. Therefore, rather than controlling the conveying blowers with a uniform control method for multiple parking lots, it is desirable to control the conveying blowers with a control method suited to each parking lot in order to ensure efficient ventilation. A learning device 200 is used for this purpose.

[0069] Figure 13 is a schematic diagram showing a ventilation system 130 related to learning by the learning device 200. The ventilation system 130 is the ventilation system 100 of Embodiment 1, further equipped with environmental sensors 40. The environmental sensors 40 are installed on pillars and other structures in the parking lot and measure the concentration of air pollutants. To understand the air quality of the entire parking lot, multiple environmental sensors 40 are installed at different locations in the parking lot, including the central and peripheral areas. The concentration of pollutants measured by the environmental sensors 40 is used for learning by the learning device 200.

[0070] The gases that need to be vented as exhaust from a parking lot are CO and NOx. Therefore, each environmental sensor 40 measures the concentration of CO and NOx as pollutants. CO is slightly lighter than air and moves upward towards the ceiling over time. NOx is heavier than air and moves downward towards the floor over time. Since the concentration around the mouth and nose is the most problematic when a person is standing, it is desirable to install each environmental sensor 40 at a position 1 to 2 m above the floor of the parking lot.

[0071] The NOx and CO sensors used as each environmental sensor 40 are expensive, making it impractical to permanently install them in a parking lot. In this embodiment, multiple environmental sensors 40 are loaned out for a certain period, during which time the learning device 220 performs learning. When the learning period ends, the multiple environmental sensors 40 are collected and loaned out to ventilation systems installed in other parking lots for learning. Using the environmental sensors 40 sequentially in multiple parking lots in this way is advantageous in terms of initial cost.

[0072] The learning device 200 and the inference device 300 will be described below, divided into the learning phase and the application phase. <Learning Phase>

[0073] Figure 14 is a block diagram showing the configuration of the learning device 200. The learning device 200 comprises a data acquisition unit 21 and a model generation unit 22. The data acquisition unit 21 acquires learning data from the ventilation system 130. The learning data includes airflow data indicating the airflow of each of the transport blowers 90-1 to 90-4 and image data of images taken inside the parking lot.

[0074] Image data consists of images captured by cameras 20-1 and 20-2, respectively. Airflow data represents the airflow of each conveyor blower, specifically represented by data indicating the operating parameters or operating modes of each conveyor blower that determine its airflow. For conveyor blowers whose airflow has been changed by the control unit 30 based on the image data, the airflow data changes to data indicating the changed airflow. The learning data acquired by the data acquisition unit 21 is output to the inference unit 42.

[0075] The model generation unit 22 learns airflow control commands to control the airflow of the conveyor fan based on the training data. In other words, the model generation unit 22 generates a trained model that infers optimal airflow control information from the training data acquired from the ventilation system 110.

[0076] The learning device 200 is connected to the ventilation system 130. The learning device 200 may be connected to, for example, a network 140 (see Figure 2) and acquire training data via the network 140. In this case, the learning device 200 may reside on a cloud server. Alternatively, the learning device 200 may be built into the ventilation system 130.

[0077] The model generation unit 22 can use any known learning algorithm, such as supervised learning, unsupervised learning, or reinforcement learning. As an example, let's explain the case where reinforcement learning is applied. In reinforcement learning, an agent (an acting entity) in a given environment observes the current state (environmental parameters) and decides what action to take. The environment changes dynamically as a result of the agent's actions, and the agent is given a reward according to the changes in the environment. The agent repeats this process and learns the action strategy that yields the most rewards through a series of actions. Representative reinforcement learning methods include Q-learning and TD-learning. For example, in the case of Q-learning, the general update formula for the action-value function Q(s,a) is expressed in equation 1.

[0078]

number

[0079] In Math 1, st represents the state of the environment at time t, and at represents the action at time t. The action at changes the state to st+1. rt+1 represents the reward received for this change in state, γ represents the discount rate, and α represents the learning rate. γ is in the range of 0 < γ ≤ 1, and α is in the range of 0 < α ≤ 1. The airflow from the transport fans 90-1 to 90-4 becomes the action at, the image captured inside the parking lot becomes the state st, and the best action at for state st at time t is learned.

[0080] The update formula, represented by equation 1, increases the action value Q of action a if the action value Q of action a with the highest Q value at time t+1 is greater than the action value Q of action a performed at time t, and decreases the action value Q if the opposite is true. In other words, the action value function Q(s,a) is updated so that the action value Q of action a at time t approaches the best action value at time t+1. As a result, the best action value in a given environment is sequentially propagated to the action values ​​in previous environments.

[0081] When a trained model is generated by reinforcement learning in this manner, the model generation unit 22 comprises a reward calculation unit 22a and a function update unit 22b. The reward calculation unit 22a calculates a reward based on the training data. The reward calculation unit 22a calculates a reward r based on a reward criterion. The reward criterion here is the CO concentration and NOX concentration measured by a plurality of environmental sensors 40, and the respective concentration data is acquired by the data acquisition unit 21. For example, if the measured CO concentration and NOX concentration decrease as a result of operating the conveyor blowers 90-1 to 90-4 at the airflow rate indicated by the airflow rate data, the reward calculation unit 22a increases the reward r (for example, gives a reward of "1"). On the other hand, if the measured CO concentration and NOX concentration increase as a result of operating the conveyor blowers 90-1 to 90-4 at the airflow rate indicated by the airflow rate data, the reward r is reduced (for example, gives a reward of "-1").

[0082] The function update unit 22b updates the function for determining the airflow control information for controlling the transport blowers 90-1 to 90-4 according to the reward calculated by the reward calculation unit 22a, and outputs it to the learned model storage unit 23. For example, in the case of Q-learning, the action-value function Q(st,at) represented by Equation 1 is used as the function for calculating the airflow control information.

[0083] The learning process described above is repeatedly executed. The trained model storage unit 23 stores the action-value function Q(st,at) updated by the function update unit 122b, i.e., the trained model. The trained model storage unit 23 is located outside the learning device 200, but it may also be a storage unit built into the learning device 200.

[0084] Next, we will explain the learning process of the learning device 200 using Figure 15. Figure 15 is a flowchart showing the operation of the learning device 200.

[0085] In step b1, the data acquisition unit 21 acquires airflow data from the transport blowers 90-1 to 90-4 and image data of the parking lot as training data. In step b2, the model generation unit 22 calculates a reward based on the training data. Specifically, the reward calculation unit 22a acquires the training data and determines whether to increase or decrease the reward based on the CO concentration and NOx concentration measured by the multiple environmental sensors 40.

[0086] If the exhaust gas in the parking lot decreases as a result of operating the conveyor blowers 90-1 to 90-4 at the airflow rate indicated by the airflow data, in step b3, the reward calculation unit 22a determines to increase the reward and increases the reward. On the other hand, if the exhaust gas in the parking lot increases as a result of operating the conveyor blowers 90-1 to 90-4 at the airflow rate indicated by the airflow data, in step b4, the reward calculation unit 22a determines to decrease the reward and decreases the reward.

[0087] In step b5, the function update unit 22b updates the action-value function Q(st,at) represented by Equation 1, which is stored in the trained model storage unit 23, based on the reward calculated by the reward calculation unit 22a. The learning device 200 repeatedly executes the steps from step b1 to step b5 and stores the generated action-value function Q(st,at) as a trained model in the trained model storage unit 23.

[0088] The hardware configuration of the learning device 200 is similar to that shown in Figure 4, consisting of a processor, memory, and bus 72, or a configuration with a dedicated processing circuit.

[0089] In this way, the learning device 200 performs training based on training data including airflow data from the transport blowers 90-1 to 90-4 and image data of the parking lot. Through this training, a trained model can be obtained that infers airflow control information suitable for the parking lot where the ventilation system 130 is installed.

[0090] <Utilization Phase>

[0091] Figure 16 is a block diagram showing the configuration of the inference device 300. The inference device 300 includes a data acquisition unit 41 and an inference unit 42.

[0092] The data acquisition unit 41 acquires input data from the ventilation system 130 after the learning period. The input data consists of image data of images taken inside the parking lot. The image data is data from images taken by cameras 20-1 and 20-2, respectively. The input data acquired by the data acquisition unit 41 is output to the inference unit 42.

[0093] The inference unit 42 reads the trained model stored in the trained model storage unit 23 and uses this trained model to infer airflow control information. In other words, by inputting the input data acquired by the data acquisition unit 41 into this trained model, it is possible to output airflow control information suitable for the image data.

[0094] The inference device 300 is connected to the ventilation system 130. The inference device 300 is connected to, for example, a network 140 (see Figure 2) and acquires input data via the network 140. In this case, the inference device 300 may reside on a cloud server. Alternatively, the inference device 300 may be built into the ventilation system 130. When the inference device 300 is used after the training period, the multiple environmental sensors 40 are removed from the ventilation system 130 as described above.

[0095] Although the inference device 300 was described as inferring airflow control information using a trained model learned based on the ventilation system 130, it may also be configured to obtain a trained model learned using another ventilation system and output airflow control information based on this trained model.

[0096] Next, Figure 17 will be used to explain the process for obtaining airflow control information using the inference device 300 and the operation of the ventilation system 130. Figure 17 is a flowchart showing the operation of the inference device 300 and the ventilation system 100.

[0097] In step c1, the data acquisition unit 41 acquires input data. In step c2, the inference unit 42 reads the trained model stored in the trained model storage unit 23 and obtains airflow control information by inputting the input data acquired by the data acquisition unit 151 into this trained model. The airflow control information includes information to identify at least one of the transport blowers 90-1 to 90-4 whose airflow should be changed, and information to identify the airflow after the change for each identified transport blower.

[0098] In step c3, the inference unit 42 outputs the obtained airflow control information to the control unit 30 of the ventilation system 130. In step c4, the control unit 30 of the ventilation system 130 controls the airflow of at least one of the identified conveying blowers 90-1 to 90-4 based on the airflow control information obtained from the inference unit 42. Specifically, the blower control unit 33 generates an airflow control command based on the airflow control information obtained from the inference unit 42 and transmits it to the conveying blower whose airflow should be increased.

[0099] The hardware configuration of the inference device 300 is similar to that shown in Figure 4, consisting of a processor, memory, and bus 72, or a configuration with a dedicated processing circuit.

[0100] Thus, the inference device 300 infers airflow control information for controlling the airflow of the transport blower based on the trained model learned by the learning device 200, and can provide the ventilation system with control of the transport blower that is suitable for the parking lot where the ventilation system 130 is installed. The ventilation system 130, whose transport blower is controlled by this inferred airflow control information, can perform efficient ventilation in response to the movement of vehicles, thereby improving the air quality in the parking lot.

[0101] In the learning device 200 according to this embodiment, airflow data indicating the airflow of the supply fan 5 and the exhaust fan 6 may also be added to the learning data. In this case, the inference device 300 further acquires airflow data indicating the airflow of the supply fan 5 and the exhaust fan 6 as input data. Therefore, in the ventilation system 130 operated after the learning period, the airflow of the supply fan 5 and the exhaust fan 6 is also changed to an appropriate level to perform efficient ventilation.

[0102] Furthermore, in the learning device 200, the learning data of the learning device 200 may include vehicle position data indicating the vehicle's location, either in place of or in addition to image data. The vehicle position data is vehicle position information detected from the image data by the control unit 30 (specifically, the vehicle movement detection unit 31). The vehicle position data is used in reinforcement learning as a parameter indicating action.

[0103] In this embodiment, we have described the case where supervised learning is applied to the learning algorithm used by the model generation unit 22, but this is not the only possible case. In addition to supervised learning, reinforcement learning, unsupervised learning, or semi-supervised learning can also be applied to the learning algorithm.

[0104] Furthermore, the learning algorithm used in the model generation unit 22 can be deep learning, which learns to extract the features themselves, or machine learning can be performed according to other known methods, such as neural networks, genetic programming, functional logic programming, or support vector machines.

[0105] Furthermore, the model generation unit 22 may learn airflow control information according to training data created for multiple ventilation systems. The model generation unit 22 may acquire training data from multiple ventilation systems used in the same parking lot, or it may learn airflow control information using training data collected from multiple ventilation systems operating independently in different parking lots. It is also possible to add ventilation systems from which training data is collected to the target system midway through the process, and to remove those ventilation systems from the target system. Moreover, a learning device that has learned airflow control information for one ventilation system may be applied to another ventilation system, and the airflow control information for that other ventilation system may be relearned and updated.

[0106] The ventilation system for learning may be, instead of the ventilation system 130, the ventilation system 110 of Embodiment 2 with multiple environmental sensors 40 installed, or the ventilation system 110 of Embodiment 3 with multiple environmental sensors 40 installed.

[0107] The embodiments disclosed herein are illustrative, and components can be modified, omitted, or added in each embodiment without departing from the scope of the claims. [Explanation of Symbols]

[0108] 1 Air intake, 2 Exhaust port, 5 Air intake fan, 6 Exhaust fan, 10, 10-1, 10-2 Vehicle movement path, 11 Vehicle, 12-1~12-4 Airflow, 20 Imaging unit, 20-1, 20-2 Camera, 21, 41 Data acquisition unit, 22 Model generation unit, 30 Control unit, 31 Vehicle movement detection unit, 32 Vehicle shape detection unit, 33 Fan control unit, 40 Environmental sensor, 42 Inference unit, 90, 90-1~90-4 Conveyor fan, 91A~91D Air passage, 100, 110, 120, 130, 140 Ventilation system, 200 Learning device, 300 Inference device

Claims

1. An air intake fan supplies outside air to the parking lot through the air intake, An exhaust fan that discharges the air from the parking lot to the outside through an exhaust port, A conveying blower that transports airflow from the supply blower to the exhaust blower, The imaging unit captures images of the interior of the aforementioned parking lot and outputs image data, A control unit detects moving vehicles in the parking lot based on image data output from the imaging unit, further detects the size of the detected moving vehicles, and changes the airflow of the transport blower according to the size of the vehicles. A ventilation system characterized by comprising the following features.

2. There are multiple conveying blowers, The imaging unit includes multiple cameras, The ventilation system according to claim 1, characterized in that the control unit compares the size of the vehicle as captured by the plurality of cameras, identifies at least one of the plurality of transport blowers according to the comparison result, determines the airflow rate of the identified transport blower and changes it for a predetermined time.

3. A data acquisition unit that acquires learning data including image data showing an image of a parking lot where a transport blower is installed and vehicle position data showing the position of a vehicle in the parking lot, and airflow data showing the airflow of the transport blower in the one of the data, A model generation unit generates a trained model for inferring airflow control information to control the airflow of the conveyor blower from at least one of the images of the parking lot and the positions of vehicles in the parking lot, using the aforementioned training data. A learning device characterized by comprising:

4. The data acquisition unit acquires concentration data indicating the concentration of CO and the concentration of NOx in the parking lot, respectively. The learning device according to claim 3, characterized in that the model generation unit generates a trained model for inferring the airflow control command using the concentration data as a reward criterion.

5. A data acquisition unit that acquires at least one of the following data: image data showing an image of a parking lot where a transport blower is installed, and vehicle position data showing the position of a vehicle in the parking lot, An inference device characterized by comprising an inference unit that outputs airflow control information for controlling the airflow of the conveying blower from at least one of the data acquired by the data acquisition unit, using a trained model for inferring the airflow of the conveying blower from at least one of the images of the parking lot and the positions of vehicles in the parking lot.

Images