Intelligent control method of exhibition automatic lifting display platform based on digital driving

Abstract

The application discloses an intelligent control method for an exhibition automatic lifting display platform based on digital driving, and particularly relates to the technical field of intelligent control of exhibition equipment, which comprises the following steps: synchronously acquiring the body operation state data of a lifting platform body and the environmental state data around the display platform, performing multi-source data fusion through a Kalman filtering algorithm, and constructing a dynamic digital twin model; based on the model, dynamically calculating speed threshold values and height forbidden zones and other constraint conditions, and combining with personnel density information to match interactive response strategies; inputting the model, constraint conditions and strategies into a central decision unit, combining with a safety evaluation and prediction model, generating a first type of instruction including emergency braking and safety locking and a second type of instruction for controlling lifting, rotation and display effects; updating the digital twin model in real time after executing the instructions, and realizing dynamic correction of the control instructions through a closed loop cycle; and the application realizes the unity of the safety, adaptability and interactivity of the display platform operation, and improves the intelligent level of the exhibition and the audience experience.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for exhibition equipment, and more specifically, to an intelligent control method for an automatic lifting exhibition platform based on digital drive. Background Technology

[0002] Currently, automated lifting display stands are widely used in museums, exhibition halls, and commercial showrooms for the dynamic presentation of various exhibits. Traditional control methods for lifting display stands mostly rely on preset programmed instructions or simple sensor feedback, such as controlling the lifting rhythm through timers or using a single position sensor to achieve fixed-point stopping. These methods have achieved automation to a certain extent, but they still have some shortcomings in actual use. For example, because they cannot perceive the operating status of the stand itself and the surrounding environment in real time and comprehensively, the system cannot dynamically assess safety risks and it is difficult to adjust the operating strategy in time when people approach, which poses potential safety hazards.

[0003] On the other hand, the display strategy is singular and rigid, unable to adapt to the number and distribution of the audience, resulting in a lack of interactivity and appeal in the display effect. Especially in situations where the audience flow fluctuates greatly, the display behavior and the audience experience are difficult to match effectively. In addition, the system's decision-making process usually lacks foresight and cannot predict the trend of people gathering in the short term, thus failing to optimize the display strategy in advance to cope with the upcoming high or low traffic conditions. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an intelligent control method for an automatically lifting exhibition platform based on digital drive, which addresses the problems of insufficient security, poor adaptability and interactivity, and lack of predictability mentioned in the background art through the following solutions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent control method for a digitally driven automatic lifting exhibition platform, comprising:

[0006] S1: Synchronously acquire the main body operation status data of the lifting platform and the environmental status data around the display platform, fuse the main body operation status data and the environmental status data, and construct a dynamic digital twin model.

[0007] S2: Based on the dynamic digital twin model, the dynamic constraints of the lifting platform under the current situation are dynamically calculated; based on the analysis of the environmental state data, the interactive response strategy that the system should adopt is determined.

[0008] S3: Input the dynamic digital twin model, dynamic constraints, and interactive response strategies into a central decision-making unit;

[0009] The central decision-making unit comprehensively evaluates the input information based on preset decision-making logic and generates corresponding control instructions; the control instructions include a first type of instructions for responding to risk situations and a second type of instructions for guiding normal operations.

[0010] S4: Execute the control command, update the dynamic digital twin model based on the latest real-time status data of the system, and repeat S2 and S3 to perform real-time correction of the control command.

[0011] Preferably, the operating status data of the lifting platform includes real-time load, platform height, platform tilt angle, and drive mechanism temperature;

[0012] The environmental status data around the display stand includes personnel distance information and personnel density distribution information obtained through visual analysis, as well as environmental audio data collected through sound sensors.

[0013] Preferably, the construction of the dynamic digital twin model employs the Kalman filter algorithm to fuse and estimate the state of multi-source data.

[0014] Preferably, the dynamic constraints include a speed threshold dynamically calculated based on real-time load and platform height, and a height restricted area dynamically defined based on personnel distance information.

[0015] Preferably, the rules for delineating the height restricted area are as follows:

[0016] When the personnel distance information indicates that personnel have entered the circular monitoring area with radius R centered on the lifting platform, the platform will be dynamically prohibited from descending below a safe height; the value of radius R is positively correlated with the real-time load weight of the platform.

[0017] Preferably, the interactive response strategy is as follows:

[0018] Based on the personnel density distribution information in the environmental status data, one of the predefined strategy levels is matched;

[0019] The strategy levels include a dormant level corresponding to an unattended state, a regular display level corresponding to a low-density state, and a high-dynamic display level corresponding to a high-density state.

[0020] Preferably, the pre-set decision logic in the central decision-making unit is configured as follows:

[0021] Prioritize security assessment; if the dynamic digital twin model indicates that the system state touches any forbidden zone or threshold in the dynamic constraints, then generate a first-class instruction.

[0022] After passing the security assessment, the display effect is optimized based on the interactive response strategy to generate the second type of instruction.

[0023] Preferably, the first type of instruction is an instruction for emergency braking, entering safety lock, or issuing an alarm;

[0024] The second type of instruction is to control the lifting platform to move to the target height, move along the planned trajectory, or adjust the display posture.

[0025] Preferably, when the central decision-making unit comprehensively evaluates the input information, it also accesses a prediction model trained on historical data; this prediction model is used to predict the trend of personnel gathering in the future based on the current dynamic digital twin model, and the interactive response strategy will be adjusted in advance according to the predicted trend.

[0026] The technical effects and advantages of this invention are as follows:

[0027] 1. This invention utilizes a dynamic digital twin model and the highest-priority real-time safety assessment. The system can dynamically calculate and strictly adhere to safety boundaries (such as speed thresholds and height restrictions). Once these boundaries are breached, it immediately executes first-class instructions such as emergency braking and safety locking, eliminating potential safety hazards at their inception. Furthermore, this safety assurance is adaptive and can be completed without interrupting the display process. This overcomes the problem of traditional equipment's fixed and rigid safety settings affecting the display effect, achieving a balance between safety and operational efficiency.

[0028] 2. This invention analyzes crowd density in real time and combines it with predictive models to anticipate crowd trends. It intelligently matches and switches interactive response strategies such as dormancy, normal, and high-dynamic modes in advance. This allows parameters such as the display stand's movement, lighting, and sound effects to accurately match and even guide the audience's emotions, avoiding the problems of rigid display strategies and disconnection from the atmosphere of the event, which are common in traditional methods. Especially in situations with large fluctuations in audience traffic, the system can prepare in advance, significantly enhancing the attractiveness of the exhibition and the quality of interactive response.

[0029] 3. This invention forms a tightly closed loop driven by data; the execution results of control commands are fed back in real time and update the digital twin model, performing dynamic corrections in extremely short cycles to ensure control accuracy. More importantly, the system periodically uses new data to incrementally train the predictive model, enabling it to continuously learn the long-term changing patterns of the exhibition scene, ensuring continuous optimization and long-term effectiveness of the intelligent control effect, and possessing the ability to continuously evolve. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the overall structure of the present invention;

[0031] Figure 2This is a schematic diagram of the dynamic digital twin model construction process of the present invention;

[0032] Figure 3 This is a schematic diagram of the workflow of the central decision-making unit of the present invention;

[0033] Figure 4 This is a schematic diagram of the instruction execution and closed-loop correction process of the present invention. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] As attached Figure 1-4 The intelligent control method for the digitally driven automated lifting exhibition platform shown includes:

[0036] S1: Synchronously acquire the main body operation status data of the lifting platform and the environmental status data around the display platform, fuse the main body operation status data and the environmental status data, and construct a dynamic digital twin model.

[0037] It should be specifically noted that the operating status data of the lifting platform includes real-time load, platform height, platform tilt angle, and drive mechanism temperature;

[0038] The environmental status data around the display stand includes personnel distance information and personnel density distribution information obtained through visual analysis, as well as environmental audio data collected through sound sensors.

[0039] Further explanation is needed regarding the following: Real-time load acquisition utilizes load cells with a measurement range of 0-500kg and an accuracy of 0.1%. Four load cells are evenly installed below the load-bearing support of the lifting platform, tightly fitted to the support, to acquire the platform load weight (m) in kg at a sampling frequency of 10Hz. Platform height acquisition employs a laser displacement sensor with a measurement range of 0-3000mm and an accuracy of ±0.03mm, fixed to the side of the platform with the laser perpendicular to the ground and facing the reflector, to acquire the platform height (h) in mm at a sampling frequency of 10Hz. Platform tilt angle acquisition uses a MEMS gyroscope with a measurement range of ±180° and an accuracy of ±0.5° installed at the horizontal center of the platform. After calibration, the X-axis and Y-axis tilt angles α and β are acquired at a sampling frequency of 10Hz. Drive mechanism temperature acquisition uses digital temperature sensors with a measurement range of -55℃ to 125℃ and an accuracy of ±0.5℃, adhered to the drive mechanism housing with thermally conductive adhesive. Two sensors are installed for each mechanism, and the average value is taken as the temperature (t) at a sampling frequency of 10Hz.

[0040] It should be further explained that the visual analysis equipment is equipped with four 4-megapixel, 25fps high-definition network cameras that support wide dynamic range, covering an area with a radius of 10m around the platform, installed at a height of 3-4m; it is paired with an edge computing device with real-time high-definition image processing capabilities for analyzing personnel distance d and density ρ, with an image sampling frequency of 5Hz; environmental audio data is converted into digital audio data A by four sound sensor modules installed at a height of 1.5-2m around the platform, which can collect audio signals from 20-16000Hz, with a sampling frequency of 5Hz.

[0041] It should be specifically noted that the construction of the dynamic digital twin model is based on the development of dynamic digital twin model construction and visualization software using the Unity3D engine. The software reads the fused ontology running status data and environmental status data from the database, and uses the Kalman filter algorithm to fuse and estimate the status of the multi-source data to construct the dynamic digital twin model of the lifting platform.

[0042] It should be further explained that the Kalman filter algorithm includes two steps: prediction and update.

[0043] Prediction steps: through state transition equations Covariance prediction equation Predict the system state, where F is the predicted state vector at time k. k Here is the state transition matrix. B is the optimal estimated state vector at time k-1. k To control the input matrix, u k The input vector is the control vector at time k. Predict the covariance matrix at time k. Q is the optimal estimated covariance matrix at time k-1. k Let be the process noise covariance matrix.

[0044] Update steps: Combine measurement equations Calculate the Kalman gain. Then update to obtain the optimal estimated state vector at time k. and the optimal estimated covariance matrix , where z k Let H be the measurement vector at time k. k For the measurement matrix, v k To measure the noise vector, R k To measure the noise covariance matrix, I is the identity matrix.

[0045] It should be further explained that, since the sampling frequency of the system's operating status data is different from that of the environmental status data, the system uses 10Hz as the reference clock and employs a linear interpolation algorithm to synchronize the environmental status data, ensuring that all data are aligned in timestamps for subsequent model building and decision-making.

[0046] S2: Based on the dynamic digital twin model, the dynamic constraints of the lifting platform under the current situation are dynamically calculated; based on the analysis of the environmental state data, the interactive response strategy that the system should adopt is determined.

[0047] It should be specifically noted that the dynamic constraints include a speed threshold dynamically calculated based on real-time load and platform height, and a height restricted area dynamically defined based on personnel distance information.

[0048] It should be further explained that the speed threshold calculation is based on the real-time load m and the platform height h, and the safe speed threshold is calculated by fitting a polynomial formula to experimental data. Where a0 is the base speed, a0 = 60 mm / s, a1 is the load linear influence coefficient, a1 = -0.02 mm / (s∙kg), a2 is the height linear influence coefficient, a2 = -0.01 (1 / s), a3 is the load-height interaction coefficient, a3 = 0.00001 (1 / (kg∙s)), and a4 is the load quadratic coefficient, a4 = 0.000005 mm / (s∙kg) 2 a5 is the coefficient of the quadratic term of altitude, a5=0.000002 (1 / (mm∙s)). The specific values ​​of each coefficient are obtained through experimental data fitting and safety verification. The specific steps are as follows:

[0049] First, multiple sets of load and height combination experiments were designed for the target exhibition stand, measuring the maximum safe speed under each condition: m values ​​covered 6 gradients: 0kg, 100kg, 200kg, 300kg, 400kg, and 500kg; h values ​​covered 7 gradients: 0mm, 500mm, 1000mm, 1500mm, 2000mm, 2500mm, and 3000mm. Each experiment was repeated 3 times, and the average of the 3 measurements of the "maximum safe speed" was taken as the valid value of that set of data. Finally, 42 sets of valid experimental data were obtained.

[0050] Then, the 42 sets of experimental data were imported into the data fitting tool, and the least squares method was used to estimate the parameters of the quadratic polynomial model: minimizing the sum of squares of the predicted and measured values, and determining the coefficients a0-a5 that minimize the error through iterative calculation. The fitting results showed that the goodness of fit of the quadratic polynomial model reached 0.96, indicating that the model can effectively reflect the influence of load and height on the speed threshold.

[0051] Finally, to ensure the safety of the formula calculation results, the fitted coefficients were modified for safety: the absolute values ​​of the coefficients of the first and second terms were appropriately amplified so that the calculated speed threshold was slightly lower than the measured maximum safe speed; after modification, the accuracy of the formula was verified through 10 additional verification experiments. The results showed that the deviation between the calculated value and the measured maximum safe speed was ≤5%, which met the safety requirements, and the coefficient values ​​were finally determined.

[0052] In summary, a0-a5 are statistical fitting parameters based on the actual structure of the platform and experimental data. Their values ​​reflect the influence of load and height on speed, and the stability of the platform's operation is ensured through safety corrections. If the structure of the platform or the characteristics of the exhibits change, the coefficient values ​​are readjusted through the above experimental steps.

[0053] It should be further explained that the radius of the circular monitoring area in the high-altitude restricted zone should first be calculated using the formula. Where R0 is the baseline safety radius, taken as 2000 mm, and k is the load sensitivity coefficient, k = 0.001 kg. -1 The value of k was determined through multiple sets of load experiments.

[0054] Different weights of load are applied to the lifting platform, and changes in the personnel safety distance are measured simultaneously; for example, when the load increases from 0 kg to 300 kg, the safety radius increases from 2000 mm to 2600 mm, as shown by the formula. The calculated value is k = 0.001 kg. -1 Then, the least squares method was used to perform linear regression on the experimental data, and the formula was transformed into: Treating k*R0 as the slope, the load m is fitted to the increment of the safety radius. The relationship between the theoretical and experimental values ​​is used to determine the optimal value of k, so that the error between the theoretical and experimental values ​​is minimized.

[0055] It should be further explained that the determination of the interactive response strategy first involves classifying the personnel density, dividing the personnel density of the exhibition area into three categories according to personnel density ρ: unmanned state (ρ<0.1 people / m²) 2 Low density (0.1 people / m²) 2 <ρ<0.5 people / m 2 High-density conditions (ρ>0.5 people / m²) 2 );

[0056] In unattended mode, a sleep strategy is implemented, initializing the sleep parameters of the drive mechanism and lighting parameters, cutting off some power, reducing energy consumption by 50%, and retaining only basic safety lighting. In low-density mode, a conventional display strategy is implemented, initializing motion parameters with a movement range of 800-1500mm, a speed of 30mm / s, completing one up-and-down reciprocating motion every 30 seconds, and using warm white light with 50% brightness. In high-density mode, a high-dynamic display strategy is implemented, initializing motion parameters with a movement range of 500-2000mm, a speed of 40mm / s, ±15° rotation, a rotation speed of 5° / s, and completing one combined up-and-down reciprocating and left-and-right rotation motion every 20 seconds. Lighting parameters alternate between warm and cool white, with brightness adjustable from 70% to 100%. Audio parameters adjust volume according to ambient audio A: 60dB for A < 60dB, 70dB for 60dB ≤ A < 70dB, and 80dB for A ≥ 70dB.

[0057] S3: Input the dynamic digital twin model, dynamic constraints, and interactive response strategies into a central decision-making unit;

[0058] The central decision-making unit comprehensively evaluates the input information based on preset decision-making logic and generates corresponding control instructions; the control instructions include a first type of instructions for responding to risk situations and a second type of instructions for guiding normal operations.

[0059] It should be noted that when the central decision-making unit comprehensively evaluates the input information, it also incorporates a prediction model trained on historical data. This prediction model is used to predict the trend of personnel gathering in the future based on the current dynamic digital twin model, and the interactive response strategy will be adjusted in advance according to the predicted trend.

[0060] It is necessary to further explain the details of the preprocessing of the training data for the predictive model, which specifically includes the following steps:

[0061] Step 1: Determine the source and dimensions of the training data

[0062] The training data is taken from the system's historical operating data over the past year, containing nine core dimensions, organized into two categories: one is the system's operational status dimension, including real-time load m, platform height h, X-axis tilt angle α, Y-axis tilt angle β, and drive mechanism temperature t; the other is the environmental status and control command dimension, including personnel distance d, personnel density ρ, ambient audio A, and historical average speed V. avg Historical height change trend H trend .

[0063] Step 2: Data Acquisition and Storage Format Standardization

[0064] The raw historical data was exported from the MySQL database and uniformly converted to CSV format. Each row of data contained "timestamp + 9 dimension values". The timestamp was uniformly formatted as "YYYY-MM-DD HH:MM:SS.ms" to ensure time alignment of data from different devices. If there was a deviation of ±50ms, the missing timestamp data was supplemented by linear interpolation. Numerical data was retained to two decimal places, and enumerated values ​​were stored as integers. The storage files were split by "month", and the data size of a single file was controlled within 1 million rows to avoid memory overflow during processing.

[0065] Step 3: Data Cleaning

[0066] Missing value handling: If ≤3 consecutive sampling points are missing, linear interpolation is used to fill in the missing points; if >3 consecutive sampling points are missing, the historical average value within the same time period is used instead; if a control command is missing, the command value from the previous moment is used directly.

[0067] Outlier handling: Outliers are identified based on the 3σ criterion. The mean µ and standard deviation σ of historical data are calculated for each dimension. Values ​​exceeding [µ-3σ, µ+3σ] are considered outliers. The outlier replacement rules are as follows: when the load m is abnormal, it is replaced with 90% of the equipment's maximum load; when the temperature t is abnormal, it is replaced with the upper limit of the safe temperature of 80℃; when the personnel density ρ is abnormal, it is replaced with 0 (unmanned state); and outliers in other dimensions are replaced with µ.

[0068] Redundant data processing: Delete duplicate data with the same timestamp, keep the latest record, and remove invalid data from the equipment debugging period.

[0069] Step 4: Data Normalization

[0070] Due to the large differences in the dimensions of the data, Min-Max normalization is used to shrink all data to the [0,1] interval. The formula is as follows: , where x min x max These represent the minimum and maximum values ​​of historical data for each dimension.

[0071] Step 5: Time Series Construction

[0072] The LSTM model requires a fixed-length time window of data as input. The time window length was selected as 30 seconds through experiments, corresponding to 300 sampling points for the ontology data and 150 sampling points for the environment data. The prediction window is for the next 5 minutes, corresponding to the average population density after 300 sampling points. The sliding window method is used to generate training samples, with a window step size of 10 seconds. Each sample contains a 30-second input sequence and 1 prediction label. The data is divided into training set, validation set, and test set in a 7:2:1 ratio, strictly in chronological order.

[0073] Step 6: Data Augmentation

[0074] To address the issue of low sample size during peak periods, a sliding window overlapping sampling enhancement method was adopted: the overlap rate was set to 50%, and the original step size of 10 seconds was changed to 5 seconds, which doubled the sample size during peak periods; at the same time, ±2dB Gaussian noise was added to the environmental audio A to generate 20% perturbation samples, thus avoiding model overfitting.

[0075] It should be further explained that the prediction model is an end-to-end structure of sequence input—feature extraction—overfitting prevention—feature integration—regression output. The specific configuration, dimensions, and functions of each level are as follows:

[0076] Input layer: The input shape is (None, 300, 10), where None indicates that the batch size is not fixed, 64 during training and 1 during inference to adapt to the data volume requirements of different scenarios, 300 corresponds to the number of sampling points in a 30-second time window, and 10 corresponds to 10 input feature dimensions; it receives the preprocessed input data and converts it into a tensor format that the LSTM layer can recognize, providing raw data for subsequent feature extraction;

[0077] LSTM layer: 64 neurons, return_sequences=False, activation function is tanh, recurrent_activation is sigmoid; core feature extraction module, through forget gate discarding irrelevant historical information and input gate retaining key new information, solves the gradient vanishing problem of traditional RNN, and accurately captures the dynamic change pattern of time series;

[0078] Dropout layer: dropout=0.2, which means randomly dropping 20% ​​of the neuron connections; during training, the connections between some LSTM layers and subsequent fully connected layers are randomly disconnected to avoid the model over-relying on certain features, reduce the risk of overfitting, and improve the model's generalization ability in different exhibition scenarios;

[0079] Fully connected layer: 32 neurons, activation function is ReLU; the 64-dimensional feature vector output by the LSTM layer is mapped to a 32-dimensional high-dimensional integrated feature through the weight matrix. At the same time, the non-linear transformation ReLU(x)=max(0,x) is introduced through the ReLU activation function to select the most critical feature combination for people density prediction, in order to prepare for the final regression output.

[0080] Output layer: 1 neuron, activation function is Linear; it maps the 32-dimensional feature vector output by the fully connected layer to a continuous numerical value, which represents the population density ρ for the next 5 minutes. pred The normalized value, ranging from 0 to 1, needs to be denormalized subsequently (using the ρ value saved during the preprocessing stage). min =0 people / m 2 ρ max =2 people / m2 This can be converted into actual physical personnel density to provide a basis for decision-making regarding interactive response strategy switching. For example, a normalized value of 0.5 corresponds to an actual ρ = 1 person / m². 2 .

[0081] It should be further noted that the LSTM prediction model, after being trained, has been deployed in a high-performance edge computing device. In the actual control loop, the model inference time is less than 50 milliseconds, which is far less than the 0.1-second control cycle, thus meeting the real-time requirements.

[0082] It should be specifically noted that the pre-set decision logic in the central decision-making unit is configured as follows:

[0083] Prioritize security assessment; if the dynamic digital twin model indicates that the system state touches any forbidden zone or threshold in the dynamic constraints, then generate a first-class instruction.

[0084] After passing the security assessment, the display effect is optimized based on the interactive response strategy to generate the second type of instruction.

[0085] It should be further explained that the security assessment and Type I instruction generation include:

[0086] Speed ​​assessment: Obtain the current velocity v of the platform. If v > v th It generates an emergency braking command to prevent the platform from running at excessive speed;

[0087] Height assessment: If the current platform height h <h s Furthermore, if the personnel distance d≤R, the safety lock command will only allow the platform to rise and generate an audible and visual alarm signal: a red flashing light at a frequency of 1Hz and a buzzer sound at a frequency of 1kHz to remind on-site personnel to stay away.

[0088] Temperature assessment: If the temperature t of the drive mechanism is greater than 80°C, an over-temperature alarm signal and a load reduction command are generated to control the drive mechanism to reduce its output power; if t is greater than 90°C, an emergency braking command is generated to stop the drive mechanism from operating.

[0089] Tilt assessment: If the tilt angle α>3 or β>3, a safety lock command is generated to lock the movement of the platform and a tilt alarm signal is issued to remind staff to check and adjust.

[0090] If the safety assessment is passed, and the predicted 5-minute crowd gathering trend ρ output by the model is combined with the results... pred The interaction response strategy is optimized based on the current personnel density ρ to generate a second type of instruction. For example, if the current personnel density ρ is at the lower limit of the "normal display level", but the predicted trend ρ... predIf the system indicates that it will enter a high-density state within the next 5 minutes, it may generate instructions in advance to gradually increase the movement speed and light brightness, so as to achieve a smooth and imperceptible strategy transition, rather than suddenly switching when the density threshold is exceeded, thereby improving the audience experience.

[0091] The central decision-making unit converts the generated control commands into a signal format that the actuators can recognize, and transmits them to the actuators such as servo motors, braking devices, and audible and visual alarms through pulse output interfaces and relay output interfaces.

[0092] It should be further explained that the first type of instruction has the highest interrupt priority. When the system is executing the second type of instruction, if the security assessment triggers the generation of the first type of instruction, the system will immediately stop the execution of the second type of instruction and forcibly execute the first type of instruction. After the security risk is eliminated, the system needs to be reassessed before it can resume or regenerate the second type of instruction.

[0093] S4: Execute the control command, update the dynamic digital twin model based on the latest real-time status data of the system, and repeat S2 and S3 to perform real-time correction of the control command.

[0094] It should be further explained that the execution of the first type of instruction specifically includes:

[0095] Emergency braking: After receiving the command, the braking device completes the braking action within 0.5 seconds. The platform stops moving through the friction between the brake pads and the motor shaft. The braking torque is adjusted according to the platform load and movement speed to ensure smooth braking and avoid damage to the platform or exhibits.

[0096] Safety Lock: After receiving the command, the drive motor stops outputting driving force and locks the motor shaft to prevent the platform from moving under gravity or external force; if it is a height restricted area lock, only the motor's descent drive signal is blocked, while the motor is allowed to receive the ascent drive signal.

[0097] Alarm: After receiving the command, the audible and visual alarm will continuously emit red flashes and a buzzer sound until the system returns to normal or the alarm is manually deactivated by staff.

[0098] The execution of the second type of instructions specifically includes:

[0099] Lifting control: After receiving the command, the servo motor adjusts its position according to the target height h. target and velocity v target The PID control algorithm is used to adjust the output power, and the encoder provides real-time feedback on the motor speed and angle to ensure precise lifting and lowering of the platform.

[0100] Rotation control: After receiving the command, the rotation drive motor rotates according to the target rotation angle θ. target and rotational speed ωtarget The stage body is driven to rotate around the vertical axis. The rotation angle sensor has a measurement accuracy of ±0.1° and provides real-time feedback of the rotation angle to ensure precise rotation.

[0101] Posture and lighting adjustment: If the platform is equipped with an item tilt adjustment mechanism, the mechanism will adjust the posture of the exhibited items according to the target tilt angle after receiving the instruction; after receiving the instruction, the display lights will switch brightness and color according to the preset mode to enhance the display effect.

[0102] It should be further explained that during the execution of control commands, the data acquisition program continues to collect data such as m, h, α, β, t, d, ρ, and A at a set frequency, and stores them in the database after preprocessing. The model software reads the latest real-time data from the database, re-executes the Kalman filter algorithm for data fusion and state estimation, obtains the latest estimated state vector, updates the parameters of the digital twin model, and ensures that the model is consistent with the actual operating state of the platform and changes in the surrounding environment. Furthermore, it extracts one month's worth of operating data from the database each month, preprocesses it according to the process of data format standardization—data cleaning—data normalization—time series construction—data augmentation, and then... The LSTM model undergoes incremental training, with the learning rate reduced to 0.0001, and 50 iterations are performed. The model weight file is then updated to adapt the model to long-term changes in the exhibition scenario. After the model update is completed, the system returns to step S2, where dynamic constraints are recalculated based on the updated dynamic digital twin model. The interactive response strategy is determined by combining the latest environmental data and prediction model results. Then, in step S3, the central decision-making unit comprehensively evaluates the updated input information and generates new control commands. Finally, in step S4, the commands are executed and the model is updated. With a cycle of 0.1 seconds, the control commands are corrected in real time to ensure the safe and efficient operation of the platform and the optimal display effect.

[0103] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0104] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A digitally driven intelligent control method for an automatic lifting exhibition platform, characterized in that: include: S1: Synchronously acquire the main body operation status data of the lifting platform and the environmental status data around the display platform, fuse the main body operation status data and the environmental status data, and construct a dynamic digital twin model. S2: Based on the dynamic digital twin model, the dynamic constraints of the lifting platform under the current situation are dynamically calculated; based on the analysis of the environmental state data, the interactive response strategy that the system should adopt is determined. The dynamic constraints include a speed threshold dynamically calculated based on real-time load and platform height, and a height restricted area dynamically defined based on personnel distance information. S3: Input the dynamic digital twin model, dynamic constraints, and interactive response strategies into a central decision-making unit; The central decision-making unit comprehensively evaluates the input information based on preset decision-making logic and generates corresponding control instructions; the control instructions include a first type of instructions for responding to risk situations and a second type of instructions for guiding normal operations. The pre-set decision logic in the central decision-making unit is configured as follows: Prioritize security assessment; if the dynamic digital twin model indicates that the system state touches any forbidden zone or threshold in the dynamic constraints, then generate a first-class instruction. After passing the security assessment, the display effect is optimized based on the interactive response strategy to generate the second type of instruction; When the central decision-making unit comprehensively evaluates the input information, it also incorporates a prediction model trained on historical data. This prediction model is used to predict the trend of people gathering in the future based on the current dynamic digital twin model, and the interactive response strategy will be adjusted in advance according to the predicted trend. S4: Execute the control command, update the dynamic digital twin model based on the latest real-time status data of the system, and repeat S2 and S3 to perform real-time correction of the control command.

2. The intelligent control method for a digitally driven automatic lifting exhibition platform according to claim 1, characterized in that: The operating status data of the lifting platform includes real-time load, platform height, platform tilt angle, and drive mechanism temperature. The environmental status data around the display stand includes personnel distance information and personnel density distribution information obtained through visual analysis, as well as environmental audio data collected through sound sensors.

3. The intelligent control method for a digitally driven automatic lifting exhibition platform according to claim 1, characterized in that: The construction of the dynamic digital twin model employs the Kalman filter algorithm to fuse multi-source data and estimate its state.

4. The intelligent control method for the digitally driven automatic lifting exhibition platform according to claim 1, characterized in that: The rules for delineating the restricted altitude areas are as follows: When the personnel distance information indicates that personnel have entered the circular monitoring area with radius R centered on the lifting platform, the platform will be dynamically prohibited from descending below a safe height; the value of radius R is positively correlated with the real-time load weight of the platform.

5. The intelligent control method for a digitally driven automatic lifting exhibition platform according to claim 1, characterized in that: The interactive response strategy is as follows: Based on the personnel density distribution information in the environmental status data, one of the predefined strategy levels is matched; The strategy levels include a dormant level corresponding to an unattended state, a regular display level corresponding to a low-density state, and a high-dynamic display level corresponding to a high-density state.

6. The intelligent control method for a digitally driven automatic lifting exhibition platform according to claim 1, characterized in that: The first type of instruction is an instruction to brake suddenly, enter safety lock, or issue an alarm. The second type of instruction is to control the lifting platform to move to the target height, move along the planned trajectory, or adjust the display posture.

Images