Method and device for determining running state of elevator and robot

Abstract

The invention discloses a method and device for determining the running state of an elevator and a robot. The method comprises the steps that under the condition that it is detected that a robot enters an elevator, acceleration data collected by an acceleration sensor are obtained, the robot moves along with the elevator, and the acceleration sensor is arranged in the robot; a recognition model is adopted to analyze the acceleration data, a recognition result is obtained, and the recognition result is used for determining the motion state of the robot; and the elevator running state of the elevator is determined according to the recognition result, and the elevator running state information is used for controlling actions of the robot. The technical problem that the detection result is inaccurate due to the fact that the robot is assisted in judging the running state of the elevator through an external ultrasonic sensor or a camera arranged on the robot in the prior art and is easily interfered by the external environment in the detection process is solved.

Description

Technical Field

[0001] The present application relates to the field of robot technology, and in particular to a method, device and robot for determining the operating status of an elevator. Background Art

[0002] With the development of artificial intelligence technology, robots are gradually being applied to inspection, remote sensing, rescue and other tasks. In order to enable robots to be used safely in elevators, relevant technologies use external ultrasonic sensors or cameras on the robots to assist robots in judging the operating status of elevators. However, the detection process is easily affected by external environmental interference. For example, on high floors, the ranging capability of ultrasonic sensors is severely attenuated, and installing cameras may block the camera's field of view in the elevator, resulting in inaccurate detection results.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present application provide a method, device and robot for determining the operating status of an elevator, so as to at least solve the technical problem that the related technology uses an external ultrasonic sensor or a camera set on the robot to assist the robot in determining the operating status of the elevator, and the detection process is easily disturbed by the external environment, resulting in inaccurate detection results.

[0005] According to one aspect of an embodiment of the present application, a method for determining the operating status of an elevator is provided, comprising: when a robot is detected to enter the elevator, acquiring acceleration data collected by an acceleration sensor, wherein the robot follows the movement of the elevator and the acceleration sensor is arranged in the robot; using a recognition model to analyze the acceleration data to obtain a recognition result, wherein the recognition result is used to determine the motion state of the robot; determining the elevator operating status of the elevator according to the recognition result, wherein the elevator operating status information is used to control the movement of the robot.

[0006] In some embodiments of the present application, the recognition model includes an input layer, a recognition network and an output layer, and the recognition network includes a first hidden layer and a second hidden layer; the recognition model is used to analyze the acceleration data to obtain a recognition result, including: obtaining a preset number of acceleration sequence data before the collection time of the acceleration data through the input layer, and the historical recognition result corresponding to the previous recognition network of the current recognition network, to obtain the input data; extracting features of the input data through the first node in the first hidden layer of the recognition network to obtain a first feature vector, and the first hidden layer includes multiple first nodes; feature mapping the first feature vector through the second node in the second hidden layer of the recognition network to obtain a second feature vector, and the second hidden layer includes multiple second nodes, and the number of second nodes is greater than the number of first nodes; converting the second feature vector into the motion state of the robot through the output layer, and determining the recognition result according to the motion state of the robot.

[0007] In some embodiments of the present application, features are extracted from input data by identifying the first node in the first hidden layer of the network to obtain a first feature vector, including: obtaining a first weight set corresponding to each first node, the first weight set including weights corresponding to each acceleration sequence data and weights corresponding to historical recognition results; performing weighted summation of the weights in the first weight set and the corresponding data in the input data to obtain a first result; determining a first feature vector based on the first result, the first feature vector being used to represent the outputs of all first nodes.

[0008] In some embodiments of the present application, a second feature vector is obtained by performing feature mapping on the first feature vector by identifying a second node in a second hidden layer of the network, including: obtaining a second weight set corresponding to each second node, the second weight set including weights corresponding to the outputs of all first nodes respectively; performing weighted summation of the weights in the second weight set and the outputs of the first nodes respectively corresponding to the first feature vector to obtain a second result; determining a second feature vector based on the second result, the second feature vector being used to represent the outputs of all second nodes.

[0009] In some embodiments of the present application, the recognition model includes multiple recognition networks, which share the same number and parameters of first nodes and second nodes, and in each recognition network, the multiple first nodes and the multiple second nodes have different parameters.

[0010] In some embodiments of the present application, a recognition model is used to analyze acceleration data to obtain a recognition result, including: for a target recognition network among multiple recognition networks, the target recognition network is used to analyze the acceleration data to obtain a first recognition result and a second recognition result, wherein the first recognition result is used to reflect the motion state of the robot, and the second recognition result is used as the input of the next recognition network of the target recognition network, and the target recognition network is any one of the multiple recognition networks.

[0011] In some embodiments of the present application, after determining the elevator operating status of the elevator based on the recognition result, the method also includes: when the elevator operating status information indicates that the elevator is in a stationary state, determining whether the robot meets the preset conditions; when the robot meets the preset conditions, controlling the robot to move out of the elevator and stop collecting acceleration data.

[0012] According to another aspect of the embodiment of the present application, a device for determining the operating status of an elevator is also provided, including: an acquisition module, used to acquire acceleration data collected by an acceleration sensor when a robot is detected to enter the elevator, wherein the robot follows the movement of the elevator and the acceleration sensor is arranged in the robot; an identification module, used to analyze the acceleration data using a recognition model to obtain a recognition result, and the recognition result is used to determine the movement state of the robot; a determination module, used to determine the elevator operating status of the elevator according to the recognition result, and the elevator operating status information is used to control the movement of the robot.

[0013] According to another aspect of the embodiment of the present application, a robot is also provided, including: an acceleration sensor, a moving mechanism, a memory and a processor, the acceleration sensor is used to collect acceleration data, the moving mechanism is used to control the robot to enter and exit the elevator, and the memory is used to store program instructions; the processor is connected to the memory and is used to execute the method for determining the above-mentioned elevator operating status.

[0014] According to another aspect of the embodiment of the present application, a non-volatile storage medium is also provided, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned method for determining the operating status of the elevator by running the computer program.

[0015] According to another aspect of the embodiment of the present application, a computer program product is also provided, including computer instructions, which implement the above-mentioned method for determining the operating status of an elevator when executed by a processor.

[0016] In the embodiment of the present application, by analyzing the data of the robot's built-in acceleration sensor, the operating status of the elevator can be accurately identified, thereby achieving the purpose of accurately capturing subtle changes in the acceleration data through deep learning of the recognition network, thereby achieving the technical effect of improving the accuracy of elevator operating status identification and improving the safety and autonomy of the robot in the elevator environment, and further solving the technical problem that the related technology uses external ultrasonic sensors or sets a camera on the robot to assist the robot in judging the operating status of the elevator, which is easily affected by external environmental interference during the detection process, resulting in inaccurate detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 It is a hardware structure block diagram of a computer terminal of a method for determining an elevator operating state according to an embodiment of the present application;

[0019] Figure 2 is a flow chart of a method for determining an elevator operating state according to an embodiment of the present application;

[0020] Figure 3 is a schematic diagram of an identification model structure of a method for determining an elevator operating state according to an embodiment of the present application;

[0021] Figure 4 is a schematic diagram of an identification network structure of a method for determining an elevator operating state according to an embodiment of the present application;

[0022] Figure 5 is a schematic diagram of an identification result of a method for determining an elevator operating state according to an embodiment of the present application;

[0023] Figure 6 It is a schematic diagram of the overall process of a method for determining the operating state of an elevator according to an embodiment of the present application;

[0024] Figure 7 It is a structural schematic diagram of a device for determining the operating status of an elevator according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] In order to better understand the embodiments of the present application, the technical terms involved in the embodiments of the present application are explained as follows:

[0028] ARM-based Industrial Computer (AIC) is an industrial control computer based on the ARM architecture. It has the characteristics of low power consumption, high performance and high integration. It is widely used in embedded systems, Internet of Things (IoT) devices and robot control, and can handle complex control logic and data processing tasks.

[0029] Mobility Mechanism: A part of a robot that is responsible for physical movement, including walking, rolling, crawling, flying, etc. Common forms include wheels, feet, tracks, etc. The design of the mobility mechanism directly affects the flexibility, stability, and efficiency of the robot, and is a key component that enables the robot to perform specific tasks.

[0030] In the inspection work of modern IDC computer rooms, in order to improve the inspection efficiency, the robot is given the important task of autonomously riding the elevator to inspect multiple floors. The key to the robot's successful entry and exit of the elevator is that it must accurately determine whether the elevator is in a stopped state. This challenge not only tests the robot's intelligence, but also concerns the safety and stability of the inspection process.

[0031] Traditional elevator status detection methods mainly rely on monitoring and receiving signals from the elevator control system. However, this method of relying on external system notifications has potential safety hazards. Once the external system gives an erroneous signal, the robot may enter and exit the elevator during the operation of the elevator, causing unpredictable dangers. Therefore, the robot needs to have the ability to independently perceive the elevator's operating status as a second layer of insurance when the elevator control system data is disordered.

[0032] In the related technologies, the methods for independently judging the status of the robot during the elevator ride without relying on the elevator control system of the elevator itself mainly include: (1) installing ultrasonic sensors in the elevator shaft, directly measuring the position information of the elevator through the ultrasonic sensors, and thus judging the operating status of the elevator; (2) installing cameras, using deep learning technology to identify the switch status of the elevator door in the visual image, and thus actually judging the current elevator door status and indirectly obtaining the current elevator operating status. In these two technical routes, firstly, installing sensor equipment in the elevator shaft will bring about a large construction cost, which is difficult to construct and inconvenient to use. In addition, the shaft environment is different, and the ranging ability of the ultrasonic sensor is seriously attenuated at high floors; secondly, the method of using deep network technology by installing a visual camera on the robot body does not have the problem of reduced detection accuracy at high levels, but there is a problem that the current detection result will be invalid when the camera field of view is blocked by the elevator.

[0033] In order to solve the above technical problems, the embodiments of the present application provide corresponding solutions, which are described in detail below.

[0034] The method for determining the running status of an elevator provided in the embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computer terminal for implementing a method for determining the running state of an elevator. Figure 1 As shown, the computer terminal 10 may include one or more (102a, 102b, ..., 102n are used to illustrate) processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission module 106 for communication functions connected via a wired and / or wireless network. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. Those skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0035] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10. As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0036] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the method for determining the running state of the elevator in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the above-mentioned method for determining the running state of the elevator is realized. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0037] The transmission module 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission module 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0038] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 .

[0039] It should be noted that, in some optional embodiments, the above Figure 1The computer terminal shown may include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of hardware elements and software elements. It should be noted that Figure 1 This is merely one example of a particular embodiment and is intended to illustrate the types of components that may be present in the computer terminal described above.

[0040] In the above operating environment, an embodiment of the present application provides an embodiment of a method for determining the operating status of an elevator. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0041] Figure 2 is a flow chart of a method for determining the running state of an elevator according to an embodiment of the present application, such as Figure 2 As shown, the method comprises the following steps:

[0042] Step S202, when it is detected that the robot enters the elevator, the acceleration data collected by the acceleration sensor is obtained, wherein the robot follows the movement of the elevator and the acceleration sensor is arranged in the robot.

[0043] In the above step S202, in the IDC room inspection scenario, the main purpose of the robot entering the elevator is to achieve autonomous movement between floors. When the robot follows the elevator, whether the elevator is stationary, accelerating up, decelerating down, or translating between floors, the robot will move synchronously with the elevator, so the robot's acceleration will directly reflect the movement state of the elevator.

[0044] The accelerometer can be installed at the center of the bottom of the robot to capture the acceleration changes of the robot in three dimensions (X, Y, and Z axes). During the operation of the elevator, the accelerometer can record the acceleration data of the robot as it moves with the elevator. For example,

[0045] (1) Elevator static state: When the elevator stops at a certain floor, the robot's acceleration will be close to the earth's gravity acceleration (about 9.8m / s 2 ), the Z-axis shows a relatively stable value, while the accelerations of the X-axis and Y-axis should be close to zero, indicating that there is no acceleration change in the horizontal direction.

[0046] (2) Elevator accelerating upward: When the elevator starts to rise, since the acceleration is proportional to the elevator's rising speed, the robot's acceleration sensor will detect a positive Z-axis acceleration, and the value will be higher than the gravity acceleration in the static state, indicating that the elevator is accelerating upward.

[0047] (3) Elevator deceleration and descent state: When the elevator descends from a high floor and gradually decelerates, the robot's acceleration sensor will measure a negative Z-axis acceleration, indicating that the elevator is decelerating and descending, and the magnitude of the value reflects the degree of deceleration.

[0048] (4) Elevator translation state: If the elevator moves horizontally between floors (such as a slight shake when leveling), the accelerometer will capture the corresponding acceleration changes on the X-axis and Y-axis, while the acceleration on the Z-axis should remain close to the gravity acceleration in the static state.

[0049] When the robot enters the elevator, the acceleration sensor can start to collect acceleration data at a fixed frequency (such as 100Hz), and the acceleration data is input into the robot's internal memory to form a data sequence. In order to accurately determine the operating status of the elevator, the robot can analyze the acceleration data over a period of time, rather than the reading at a single moment. For example, the robot can send the data of the last five sampling moments (i.e., sequence data) into the deep neural network model for analysis. These five data points can provide enough information to determine whether the elevator is in an accelerating, decelerating, or stable state.

[0050] Step S204: using a recognition model to analyze the acceleration data to obtain a recognition result, wherein the recognition result is used to determine the motion state of the robot.

[0051] In the above step S204, the recognition model refers to a deep neural network model that uses acceleration data to parse and classify the elevator operation status. Through training, the model can learn and understand the characteristics of acceleration data, so as to accurately predict whether the robot is in a stationary or moving state.

[0052] In various complex elevator environments, in order to extract deeper features from acceleration data and improve the accuracy and robustness of elevator operation status recognition, in some embodiments of the present application, the recognition model includes an input layer, a recognition network and an output layer, and the recognition network includes a first hidden layer and a second hidden layer; the recognition model is used to analyze the acceleration data to obtain a recognition result, including: obtaining a preset number of acceleration sequence data before the acquisition time of the acceleration data through the input layer, and the historical recognition result corresponding to the previous recognition network of the current recognition network, to obtain the input data; extracting features of the input data through the first node in the first hidden layer of the recognition network to obtain a first feature vector, and the first hidden layer includes multiple first nodes; feature mapping the first feature vector through the second node in the second hidden layer of the recognition network to obtain a second feature vector, and the second hidden layer includes multiple second nodes, and the number of second nodes is greater than the number of first nodes; converting the second feature vector into the motion state of the robot through the output layer, and determining the recognition result according to the motion state of the robot.

[0053] The input layer is the starting part of the recognition model, which is used to receive the raw data of the acceleration sensor and the historical recognition results of the previous recognition network. For example, the input layer obtains the acceleration data at the current moment t, as well as the data of the previous 4 moments, and receives the output results of the previous recognition network of the current recognition network. The input layer in the recognition model receives not only the acceleration data at the current moment, but also the model can process time series data and capture the trend of acceleration data changing over time, so as to more accurately identify the motion state of the robot.

[0054] The first hidden layer is the first processing layer in the model (recognition network), which contains multiple first nodes, which are used to perform preliminary feature extraction on the input data. For example, key patterns and features in acceleration data can be identified through weighted and activation functions (such as ReLU). Each of the multiple first nodes performs weighted summation on the input data through different weight parameters, and then converts it through the activation function to extract different feature representations.

[0055] The second hidden layer is located in the processing layer after the first hidden layer, and contains more second nodes than the first hidden layer, which is used to perform deeper feature mapping on the first feature vector, so as to capture more complex patterns. By setting more second nodes than the first nodes, the model can be distinguished between stable and non-steady states. Specifically, the nodes of the second hidden layer can distinguish between the stable operation of the elevator (relatively stable acceleration signal) and non-steady operation (acceleration signal mutation) by learning the non-steady signal changes of acceleration data when the elevator starts and stops. In the stable state, the acceleration signal is close to the acceleration of gravity, while in the non-steady state, the acceleration signal deviates from the acceleration of gravity and shows obvious fluctuations. In addition, this setting is also helpful for dynamic acceleration trend analysis. Specifically, during the operation of the elevator, the acceleration signal will change with the acceleration, deceleration and stop of the elevator. More nodes allow the model to analyze the trend of the acceleration signal. For example, the recognition model can learn the acceleration change law of the elevator from static acceleration to running speed, and then deceleration to stop. This trend analysis helps to more accurately identify the motion state of the robot.

[0056] The output layer is the final layer of the recognition model, which is responsible for converting the second feature vector output by the second hidden layer into the motion state of the robot. For example, it is converted into a value between 0 and 1 through an activation function (such as sigmoid). If the output value is greater than 0.8, the robot is judged to be in a running state; if it is less than or equal to 0.8, especially when it is close to 0, the robot is judged to be in a stationary state.

[0057] In order to analyze acceleration data from different angles and depths, a multi-layer recognition network can be constructed. Specifically: the recognition model includes multiple recognition networks, and the multiple recognition networks share the same number and parameters of first nodes and second nodes. In each recognition network, there are different parameters between the multiple first nodes and between the multiple second nodes.

[0058] Each recognition network is an independent processing unit in the recognition model, responsible for analyzing the acceleration data and outputting the elevator's operating status. In the design of multi-layer recognition networks, each network can focus on different aspects or levels of acceleration data, thereby improving the overall analysis capability and accuracy of the model.

[0059] The first node and the second node are located in the first hidden layer and the second hidden layer of the recognition network, respectively. Multiple recognition networks share the same number and parameters of the first and second nodes, that is, in the multi-layer recognition network, the nodes of the first hidden layer and the second hidden layer use the same parameter configuration (such as weights and biases) in all networks. For example, a recognition model consisting of 5 recognition networks can be constructed, each of which has a first hidden layer containing 50 nodes and a second hidden layer containing 100 nodes. Although the number of nodes and parameter structure remain consistent in all networks, the node parameters (weights and biases) of each network are individually adjusted through the training process to adapt to the characteristics and change patterns of the acceleration data at different time points.

[0060] The design of multi-layer recognition networks enables the model to analyze acceleration data from different angles and depths. The first layer of recognition networks may focus on the basic features of acceleration data, such as the amplitude and frequency changes of the signal. As the depth of the network increases, the subsequent recognition networks can learn more advanced and abstract features, such as the time series trend of the signal, the details of non-stationary signals, and abnormal patterns in elevator operation. For example, the acceleration data is first input into the first recognition network, and after being processed by the first hidden layer and the second hidden layer, the preliminary robot motion state recognition result is output. Subsequently, the result and the original acceleration data are input into the next recognition network to continue deeper feature analysis and state judgment. This progressive processing method helps the model gradually accumulate information about the robot's motion state, and the final recognition result is obtained through comprehensive analysis of multiple recognition networks.

[0061] When constructing a multi-layer recognition network, the acceleration data can be analyzed through the following steps to obtain the recognition result: for the target recognition network among the multiple recognition networks, the target recognition network is used to analyze the acceleration data to obtain a first recognition result and a second recognition result, wherein the first recognition result is used to reflect the motion state of the robot, and the second recognition result is used as the input of the next recognition network of the target recognition network, and the target recognition network is any one of the multiple recognition networks.

[0062] The first recognition result refers to the main result output by each recognition network, which is used to reflect the robot's motion state at a specific time point, that is, to determine whether the elevator is stationary or running. After all recognition networks have processed the input data sequence, the model will fuse the first recognition results of each network to determine the final recognition result.

[0063] The second recognition result refers to the auxiliary result generated by each recognition network, which is used as the input of the next recognition network. This helps to transfer information between networks, so that the latter network can perform more in-depth processing and judgment based on the analysis results of the previous network.

[0064] In some embodiments of the present application, features of input data are extracted by identifying the first node in the first hidden layer of the network to obtain a first feature vector. Specifically: a first weight set corresponding to each first node is obtained, the first weight set includes weights corresponding to each acceleration sequence data and weights corresponding to historical recognition results; the weights in the first weight set are weighted summed with the corresponding data in the input data to obtain a first result; a first feature vector is determined based on the first result, and the first feature vector is used to represent the output of all first nodes.

[0065] During the model training process, the first node will learn a set of weights (the first weight set), including the weights for different time points in the acceleration sequence data and the weights of the previous recognition network output results. The output of the first node can be calculated using the following formula:

[0066] n output_i =RELU(n i1 *x t-4 +n i2 *x t-3 +n i3 *x t-2 +n i4 *x t-1 +n i5 *x t +

[0067] n i6 *B t-1 )(Formula 1)

[0068] Among them, n output_i Indicates that each n i The output of the first node is the result obtained by converting the first result through the activation function, n i1 , n i2 , n i3 , n i4 , n i5 , n i6 n i The weight of the node, that is, the first weight set, n i1 to n i5 Indicates the weights corresponding to the acceleration sequence data, n i6 Indicates the weight corresponding to the historical recognition results.

[0069] The first hidden layer is composed of multiple first nodes, each of which independently performs weighted summation and feature extraction on the input data, and multiple first nodes can be processed in parallel. The outputs of all first nodes in the first hidden layer are combined to form a first feature vector, which can be regarded as a compact representation of acceleration data and historical recognition results, capturing the key features of time series data and facilitating deeper feature mapping in the second hidden layer.

[0070] In some embodiments of the present application, a second feature vector is obtained by performing feature mapping on the first feature vector by identifying a second node in a second hidden layer of the network. Specifically: a second weight set corresponding to each second node is obtained, the second weight set includes weights corresponding to the outputs of all first nodes respectively; the weights in the second weight set are weighted summed with the outputs of the first nodes respectively corresponding to the first feature vector to obtain a second result; and a second feature vector is determined based on the second result, the second feature vector being used to represent the outputs of all second nodes.

[0071] Each second node has a set of weights (second weight set) for determining the relative importance of different first node outputs in the first feature vector. The output of the second node can be calculated by the following formula:

[0072]

[0073] Among them, n output_k Indicates that each n i (the output of the first node), m output_i Indicates that each m i The output of the node (second node), m ik Indicates that the output of the first node is m i The weight of the node.

[0074] The second hidden layer processes the first feature vector in parallel by multiple second nodes, and each node independently performs weighted summation and feature mapping. This parallel processing mechanism enables the model to extract features from multiple angles and depths, increasing the richness of feature representation. Through the interaction between nodes (for example, by connecting to the same output node), the second hidden layer achieves information fusion, ensuring that the second feature vector can fully reflect the complexity of the elevator state.

[0075] After obtaining the output of each second node, the first recognition result reflecting the motion state of the robot can be calculated by the following formula:

[0076]

[0077] Among them, h output_i Indicates that each h iThe output, i.e. the first recognition result, h ik represents the output weight of each second node, m output_k Represents the output of each second node.

[0078] The second recognition result as the input of the next recognition network of the recognition network can be calculated by the following formula:

[0079]

[0080] b output_i Indicates that each b i The output, i.e. the second recognition result, b ik represents the output weight of each second node, m output_k Represents the output of each second node.

[0081] The activation functions RELU, sigmod, and tanhx in Formulas 1 to 4 above are defined as follows:

[0082]

[0083]

[0084]

[0085] Figure 3 is a schematic diagram of the structure of an identification model of a method for determining an elevator operating state according to an embodiment of the present application, such as Figure 3 As shown, in some embodiments of the present application, the model uses a recursive structure to perform in-depth analysis on the acceleration sensor data to identify the operating status of the elevator. Specifically, the figure describes the process of data starting from the acquisition point xt, passing through multiple recognition networks Bt, and finally generating the elevator operating status information. Figure 3In the above example, the acceleration data at each moment is input into the recognition network B at the next five moments. For example, xi is input into Bt, Bt+1, Bt+2, Bt+3, and Bt+4. Therefore, each B network receives the acceleration data at the five adjacent sampling moments. For example, Bt receives data from xt-4, xt-3, xt-2, xt-1, and xt, where xt represents the acceleration sensor data at a certain moment t. These data reflect the acceleration change of the elevator where the robot is located. These data are then input into a series of recognition networks Bt, each of which contains multiple first nodes and second nodes for feature extraction and mapping to generate a recognition result ht representing the robot's motion state. It is worth noting that the acceleration data is not only input into the recognition network Bt at the current moment, but also into the subsequent recognition networks Bt+1 to Bt+4, which reflects the recursive nature of the model, that is, the acceleration data at the current moment is used to predict the future elevator state.

[0086] Each recognition network Bt outputs two results: ht represents the motion state of the robot at time t (that is, the operating state of the elevator), where the value closer to 1 indicates that the elevator is running, and the value closer to 0 indicates that the elevator is close to stationary; bt serves as the input of the next recognition network Bt+1, reflecting the impact of historical recognition results on the model's prediction of future states. The design of the entire model is based on the principles of deep learning, aiming to capture complex patterns in acceleration data, so as to accurately judge the real-time operating state of the elevator.

[0087] Figure 4 is a schematic diagram of an identification network structure of a method for determining an elevator operating state according to an embodiment of the present application, such as Figure 4As shown in Figure 1, the structure of the recognition network in the elevator running state determination method is further refined, showing how to extract features from acceleration data and map them to the elevator state output. The recognition model mainly consists of an input layer, a recognition network B (composed of the first hidden layer and the second hidden layer), and a final output layer. The input layer contains acceleration data xt and its first four time points xt-4 to xt-1, as well as the output Bt-1 from the previous recognition network. These data together constitute the input data for the analysis of the recognition network. The first hidden layer of the recognition network B includes 100 first nodes (n1-n100) for extracting the features of the input data and generating the first feature vector; the second hidden layer includes 300 second nodes (m1-m300), which perform more complex feature mapping on the first feature vector to generate the second feature vector. The design of the B network adopts a deep learning architecture, in which each node processes data through a complex connection relationship, and each node has a specific set of weights for weighted summation of the input data to determine the feature vector. The output layer finally converts the output of the second hidden layer into the first recognition result ht of the robot's motion state and the second recognition result bt passed to the next recognition network. This multi-layer network architecture allows the model to capture deeper features from the acceleration data, thereby improving the accuracy of elevator state recognition. The setting of the number of nodes in the recognition network (for example, the number of nodes in the second hidden layer is greater than the number of nodes in the first hidden layer) is intended to increase the complexity of the model to improve its processing power and recognition accuracy.

[0088] Step S206, determining the elevator operation status of the elevator according to the recognition result, wherein the elevator operation status information is used to control the action of the robot.

[0089] In the above step S206, based on the elevator operation status information, the robot can make corresponding action decisions, such as waiting for the elevator to stop before entering or exiting the elevator, or keeping a safe distance when the elevator is running. Controlling the robot's actions involves planning and executing the robot's behavior, ensuring that the robot can make reasonable responses according to the environmental status and ensuring the safe implementation of the inspection task.

[0090] In this application, the recognition model is first used to obtain the recognition result to determine the motion state of the robot, and then the operation state of the elevator is determined according to the recognition result. Although the motion state of the robot can be basically consistent with the operation state of the elevator in some cases, in some cases (such as external environmental interference such as the robot falling), there is an error between the obtained motion state of the robot and the actual operation state of the elevator. In order to reduce misjudgment and ensure the safe movement of the robot in the elevator through multi-dimensional analysis and verification in the case of external interference or sensor abnormality, the following steps can also be performed:

[0091] A time window length is set, for example, 1-2 seconds. The time window length can be determined based on the sampling frequency of the acceleration sensor and the average motion cycle of the elevator. The recognition results are collected within the time window, and the distribution and change trend of these results are counted to obtain a first result. For example, a value of 0-1 can be used to quantify the change trend, such as analyzing the fluctuation of the recognition result. If the fluctuation is large or the state is frequently switched (such as frequently jumping from a static state to a running state and then jumping back to a static state), the first result can be determined as 1. The robot maintains a historical state queue inside, which is used to record the elevator running state in the past period of time. The current recognition result is compared with the state in the historical state queue to determine the difference between the current recognition result and the historical state to obtain a second result. For example, a value of 0-1 can be used to quantify the difference. For example, the second result can be determined as 1 if the state suddenly changes from a long-term static state to a running state. The recognition result used to determine the elevator running state is determined based on the first result (time window analysis) and the second result (historical state comparison). For example, the recognition result with errors is filtered out.

[0092] It should be noted that, in order to save computing resources, in some embodiments of the present application, the identification result for determining the elevator operating status may also be determined only based on the first result or the second result.

[0093] In some embodiments of the present application, the elevator operation status of the elevator can be determined according to the recognition results through the following steps, specifically: obtaining multiple recognition results within a preset time length, wherein each recognition result is used to indicate the probability that the robot's motion state belongs to the target in a motion state, and the target motion state includes: a moving state and a stationary state; comparing the multiple recognition results with a preset threshold value respectively to obtain a comparison result, wherein the comparison result includes the number of recognition results that meet the preset threshold value; and determining the elevator operation status information based on the comparison result.

[0094] Figure 5 : is a schematic diagram of the recognition result of a method for determining the running state of an elevator according to an embodiment of the present application. In a specific embodiment, the acceleration sensor collects acceleration data at a specific time point 198012.01s to 198015.06s, for example, 9.8121m / s 2 These data are then input into the recognition network B, which outputs a floating point number between 0 and 1, which reflects the operating status of the elevator. The closer the value is to 1, the elevator is running, and the closer it is to 0, the elevator is stationary. Figure 3Special emphasis is placed on how to judge whether the elevator is currently in a stationary state based on the recognition results. Specifically, when the robot needs to judge the current state of the elevator, it counts the output results of the recognition network in the previous 2 seconds. Since the acquisition frequency of the sensor is 100Hz, this means that there are 200 recognition results in 2 seconds. If more than 20 (i.e. 10%) of the recognition results in these 200 results are greater than 0.8, the robot judges that the current elevator is in operation; otherwise, the elevator is judged to be stationary. This judgment method combines the concepts of time window and threshold, making the judgment of the elevator state more accurate and stable. Through the statistics and analysis of multiple recognition results, the model can more robustly avoid misjudgments caused by short-term data fluctuations, thereby enhancing the reliability and practicality of the elevator state judgment.

[0095] After determining the elevator operating status of the elevator based on the recognition results, the following steps can also be performed: when the elevator operating status information indicates that the elevator is in a stationary state, determine whether the robot meets the preset conditions; when the robot meets the preset conditions, control the robot to move out of the elevator and stop collecting acceleration data.

[0096] Pre-conditions refer to a series of safety and operational requirements that the robot must meet before it can move out of the elevator. These conditions may include, but are not limited to: the elevator is confirmed to be stationary; the elevator door is fully open; the robot is positioned in the appropriate position of the elevator door for safe movement; the robot confirms that the surrounding environment is clear of obstacles; and the robot itself is in a ready state and can move smoothly.

[0097] In some embodiments of the present application, the robot can obtain information about the floor at which the current elevator stops, and compare it with the floor to be reached in the task instructions, wherein the floor information can be obtained by communicating with the elevator control system, or indirectly obtained through display screens inside the elevator, sound prompts, and the like; when the elevator is in a stationary state, the robot compares the current floor with the target floor to confirm whether it is on the predetermined floor; determines whether the elevator door is in an open state; after confirming that the floor is correct and the elevator door is fully open, the robot uses a moving mechanism to exit the elevator.

[0098] In order to enable the robot to use the acceleration sensor data, combined with the analysis of the historical elevator operation status, to predict the elevator stop in advance, and make pre-action preparations to ensure that the elevator can be quickly and safely completed when the elevator stops, the robot's inspection efficiency and safety in scenarios such as IDC rooms can be improved. The following steps can also be performed:

[0099] (1) Determine the change trend of the acceleration data through the collected acceleration data. For example, the acceleration data can be trend analyzed using a sliding time window (e.g., 5 seconds) to observe the slope of the acceleration change and the distribution of the acceleration value. By setting a prediction threshold (e.g., the acceleration change slope is less than a certain threshold, and the acceleration value continues to be within a stable range), the elevator can be predicted to stop (i.e., be in a stationary state) in advance.

[0100] (2) Determine the acceleration change pattern using the historical elevator operation status. For example, the robot stores the acceleration data and recognition results in the most recent period of time to form a historical elevator operation status database, analyzes the historical data, and recognizes the acceleration change pattern when the elevator starts, runs, and stops. In some embodiments of the present application, the prediction threshold in step (1) can also be adjusted according to the determined acceleration change pattern. For example, if the historical data shows that the acceleration reduction slope of the elevator before stopping is usually within a certain target range, the prediction threshold can be set near the target range (such as the prediction threshold belongs to the target range);

[0101] (3) When the acceleration data within a preset time (such as 1-2 seconds) meets the threshold condition for predicting the elevator stop, and the historical data trend analysis also supports the prediction that the elevator is about to stop, the pre-action is triggered, where the pre-action may include adjusting the moving mechanism to move toward the elevator door, communicating with the elevator control system to send a request for floor information, and reminding passengers that they are about to exit the elevator through a speech synthesis device.

[0102] Through the above steps S202 to S206, by analyzing the data of the robot's built-in acceleration sensor, the operating status of the elevator can be accurately identified, and the purpose of accurately capturing subtle changes in the acceleration data through deep learning of the recognition network is achieved, thereby achieving the technical effect of improving the accuracy of elevator operating status identification and improving the safety and autonomy of the robot in the elevator environment, and further solving the technical problem that the related technology uses external ultrasonic sensors or sets a camera on the robot to assist the robot in judging the elevator operating status, which is easily affected by external environmental interference during the detection process, resulting in inaccurate detection results.

[0103] Figure 6 is a schematic diagram of the overall flow of a method for determining an elevator operating state according to an embodiment of the present application, such as Figure 6 As shown, the overall process of the elevator operation status determination method may include business process control, acceleration data collection, recognition network analysis, status judgment, and robot behavior control. Specifically:

[0104] (1) Step 1: The business process controller starts the robot to enter the elevator. For example, when the robot performs a task, it needs to take the elevator to reach different floors.

[0105] (2) Step 2: The acceleration sensor on the robot body starts to collect data, which reflects the motion state of the robot (and the elevator). For example, the sensor can collect data at a frequency of 100 Hz to ensure the real-time and continuity of the data.

[0106] (3) Step 3: The collected acceleration data is input into the recognition network B for analysis. In each recognition network, the data first passes through the input layer, and then undergoes feature extraction and mapping in the first hidden layer and the second hidden layer to generate the recognition result ht of the elevator operation status and the recognition result bt as the input of the next recognition network.

[0107] (4) Step 4: Determine the elevator operating status based on the recognition result. For example, the value range of ht can be set between 0 and 1, and the running or stationary status of the elevator can be obtained through this floating point number.

[0108] (5) Step 5: When judging the elevator status, the system will consider the current business needs and the elevator network identification results. If the current business judgment requires the robot to exit the elevator, and the identification network judges that the elevator is stationary, the robot will perform the exit action. Conversely, if the elevator is judged to be still running or the robot does not need to exit the elevator, the process will return to step 2 to continue collecting acceleration data and making status judgments.

[0109] (6) Step 6 (service termination step): Once the robot successfully exits the elevator, the service process will end and the detection of the identification network will also stop to save computing resources. This closed-loop control process ensures the efficient and safe movement of the robot in the elevator environment while minimizing unnecessary resource consumption.

[0110] In the present application, the determination of the elevator's operating status does not rely on external equipment or additional sensors, but relies entirely on the data provided by the robot's own acceleration sensor. The sensor can detect changes in the elevator's acceleration during vertical movement, and these changes can be used to infer the elevator's various states, such as stationary, rising, falling, or stopping. Therefore, there is no need to detect the elevator's environment through an external optical detection system, and it is not affected by the obstruction of people in the elevator. The determination of the elevator's operating status is less affected by the environment.

[0111] Figure 7 is a structural diagram of a device for determining the running state of an elevator according to an embodiment of the present application, such as Figure 7 As shown, the device comprises:

[0112] An acquisition module 702 is used to acquire acceleration data collected by an acceleration sensor when detecting that the robot enters the elevator, wherein the robot moves along with the elevator and the acceleration sensor is disposed in the robot;

[0113] The recognition module 704 is used to analyze the acceleration data using a recognition model to obtain a recognition result, wherein the recognition result is used to determine the motion state of the robot;

[0114] The determination module 706 is used to determine the elevator operation status of the elevator according to the recognition result, wherein the elevator operation status information is used to control the action of the robot.

[0115] It should be noted that Figure 7 The elevator operation status determination device shown is used to perform Figure 2 The method for determining the elevator operating status shown in FIG. Figure 2 The relevant explanations in the method for determining the elevator operating status in Figure 7 The device for determining the running state of the elevator shown will not be described in detail here.

[0116] The embodiment of the present application also provides a robot, including: an acceleration sensor, a moving mechanism, a memory and a processor, wherein the acceleration sensor is used to collect acceleration data, the moving mechanism is used to control the robot to enter and exit the elevator, and the memory is used to store program instructions; the processor is connected to the memory and is used to execute the method for determining the operation state of the elevator. For example, the processor performs the following functions by executing the program instructions stored in the memory: when the robot is detected to enter the elevator, the acceleration data collected by the acceleration sensor is obtained, wherein the robot follows the movement of the elevator and the acceleration sensor is arranged in the robot; the acceleration data is analyzed by using a recognition model to obtain a recognition result, wherein the recognition result is used to determine the movement state of the robot; the elevator operation state of the elevator is determined according to the recognition result, wherein the elevator operation state information is used to control the action of the robot.

[0117] In some embodiments of the present application, the robot can be an IDC room inspection service robot, and also includes an ARM industrial computer. The acceleration sensor is installed at the center of the bottom of the robot, and the ARM industrial computer is used to run the neural network (i.e., recognition network) for elevator status detection.

[0118] An embodiment of the present application further provides a non-volatile storage medium, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the steps of the method for determining the elevator operating status in each embodiment of the present application by running the computer program.

[0119] The embodiment of the present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the method for determining the elevator operating state in each embodiment of the present application.

[0120] The embodiments of the present application also provide a computer program, which, when executed by a processor, implements the steps of the method for determining the operating status of an elevator in each embodiment of the present application.

[0121] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0122] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0123] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0124] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0125] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0126] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0127] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for determining the running state of an elevator, characterized in that: include: When detecting that a robot enters an elevator, acquiring acceleration data collected by an acceleration sensor, wherein the robot moves following the elevator and the acceleration sensor is disposed in the robot; Using a recognition model to analyze the acceleration data to obtain a recognition result, wherein the recognition result is used to determine the motion state of the robot; The elevator operation status of the elevator is determined according to the recognition result, wherein the elevator operation status information is used to control the action of the robot.

2. The method according to claim 1, characterized in that The recognition model includes an input layer, a recognition network and an output layer, and the recognition network includes a first hidden layer and a second hidden layer; The acceleration data is analyzed using a recognition model to obtain a recognition result, including: Acquire a preset number of acceleration sequence data before the acquisition time of the acceleration data and a historical recognition result corresponding to a previous recognition network of the current recognition network through the input layer to obtain input data; Extracting features from the input data through a first node in the first hidden layer of the recognition network to obtain a first feature vector, wherein the first hidden layer includes a plurality of first nodes; Performing feature mapping on the first feature vector through a second node in the second hidden layer of the recognition network to obtain a second feature vector, wherein the second hidden layer includes a plurality of second nodes, and the number of the second nodes is greater than the number of the first nodes; The second feature vector is converted into the motion state of the robot through the output layer, and the recognition result is determined according to the motion state of the robot.

3. The method according to claim 2, characterized in that Extracting features from the input data through a first node in the first hidden layer of the recognition network to obtain a first feature vector includes: Obtaining a first weight set corresponding to each first node, wherein the first weight set includes a weight corresponding to each acceleration sequence data and a weight corresponding to the historical recognition result; Performing a weighted summation on the weights in the first weight set and the corresponding data in the input data to obtain a first result; The first feature vector is determined according to the first result, wherein the first feature vector is used to represent outputs of all first nodes.

4. The method according to claim 2, characterized in that: Performing feature mapping on the first feature vector through a second node in the second hidden layer of the recognition network to obtain a second feature vector includes: Obtain a second weight set corresponding to each second node, wherein the second weight set includes weights corresponding to outputs of all first nodes respectively; Performing a weighted summation on the weights in the second weight set and the outputs of the first nodes respectively corresponding to the first feature vector to obtain a second result; The second eigenvector is determined according to the second result, wherein the second eigenvector is used to represent outputs of all second nodes.

5. The method according to claim 2, characterized in that: The recognition model includes a plurality of recognition networks, which share the same number and parameters of first nodes and second nodes, and in each recognition network, the first nodes and the second nodes have different parameters.

6. The method according to claim 5, characterized in that The acceleration data is analyzed using a recognition model to obtain a recognition result, including: For a target recognition network among the multiple recognition networks, the acceleration data is analyzed using the target recognition network to obtain a first recognition result and a second recognition result, wherein the first recognition result is used to reflect the motion state of the robot, and the second recognition result is used as an input of a next recognition network of the target recognition network, and the target recognition network is any one of the multiple recognition networks.

7. The method according to claim 1, characterized in that After determining the elevator operation state of the elevator according to the recognition result, the method further includes: When the elevator operation status information indicates that the elevator is in a stationary state, determining whether the robot meets a preset condition; When the robot meets a preset condition, the robot is controlled to move out of the elevator and stop collecting the acceleration data.

8. A device for determining the running state of an elevator, characterized in that: include: an acquisition module, configured to acquire acceleration data collected by an acceleration sensor when detecting that a robot enters an elevator, wherein the robot moves following the elevator and the acceleration sensor is disposed in the robot; An identification module, used to analyze the acceleration data using an identification model to obtain an identification result, wherein the identification result is used to determine the motion state of the robot; A determination module is used to determine the elevator operation status of the elevator according to the recognition result, wherein the elevator operation status information is used to control the action of the robot.

9. A robot, characterized in that: include: An acceleration sensor, a moving mechanism, a memory and a processor, wherein the acceleration sensor is used to collect acceleration data, the moving mechanism is used to control the robot to enter and exit the elevator, and the memory is used to store program instructions; the processor is connected to the memory and is used to execute the method for determining the elevator operating status as described in any one of claims 1 to 7.

10. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the method for determining the elevator operating state according to any one of claims 1 to 7 by running the computer program.

11. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by the processor, the method for determining the elevator operating status according to any one of claims 1 to 7 is implemented.

Images