Robot remote adaptive data acquisition control system and method based on dynamic value evaluation

Abstract

The invention discloses a robot remote adaptive data acquisition control system and method based on dynamic value evaluation. According to the system, dynamic information value factors of data points are calculated in real time through a local processing unit of the robot, and the factors integrate the data change rate, the uncertainty, the task context and the abnormal probability. Based on the factor, the system dynamically adjusts the sampling frequency, the compression strategy and the transmission priority. The remote center is responsible for data reconstruction and fusion and utilizes machine learning closed-loop optimization to acquire parameters. According to the invention, the problems of resource waste and key information acquisition omission caused by fixed frequency sampling are solved, and the optimal balance of data acquisition efficiency and quality under a limited bandwidth is realized.

Description

Technical Field

[0001] This invention relates to the field of data acquisition technology, and specifically to a robot remote adaptive data acquisition control system and method based on dynamic value assessment. Background Technology

[0002] With the deepening of Industry 4.0 and intelligent manufacturing, and the widespread application of special-purpose robots (such as inspection, rescue, and space operation robots), remote, real-time, and precise monitoring of robot operating status has become crucial. This relies on an efficient and reliable data acquisition and control system that can transmit massive amounts of sensor data from the robot body (such as attitude, force, temperature, and video) to a remote monitoring center for status monitoring, fault warning, performance analysis, and remote intervention.

[0003] Existing remote data acquisition solutions for robots mainly fall into the following categories: First, the fixed-frequency sampling and transmission mode. This mode collects and transmits data at fixed time intervals, which is simple to implement but has obvious drawbacks. When the robot is running smoothly and data changes slowly, a large amount of redundant data is generated, consuming valuable wireless communication bandwidth and increasing energy consumption and storage costs. Furthermore, when the robot encounters sudden situations and data changes rapidly, the fixed sampling frequency may fail to capture key transient features, leading to the loss of important information. Second, the trigger sampling mode based on simple thresholds. This mode only increases the sampling rate or initiates transmission when the data exceeds a preset static threshold. While this method offers some improvement, its judgment criteria are singular, failing to distinguish between slight fluctuations caused by noise and true anomalies, and even less able to perceive the different requirements of data importance at different stages of the robot's current task (for example, minute vibrations during precise positioning have higher analytical value than those during uniform movement). Therefore, it is prone to false triggers or missed triggers, limiting its level of intelligence.

[0004] Furthermore, existing data compression and transmission strategies are typically separated from the acquisition process, employing a passive "acquire first, compress later" approach. This lack of assessment of the data's inherent "information value" can lead to high-value data being distorted due to excessive compression, while low-value data consumes resources intended for lossless compression. At the remote data processing end, received data is usually only simply decoded and displayed, lacking a hierarchical reconstruction and fusion mechanism based on data value, and also lacking a closed-loop feedback loop to optimize the front-end acquisition strategy based on long-term operational results.

[0005] Therefore, there is an urgent need in this field for a solution that can overcome the above-mentioned shortcomings, namely, a robot remote data acquisition control system and method that can adapt to data value and intelligently allocate resources, so that it can always ensure that high-value information is collected and transmitted with priority and high quality under complex actual working conditions and limited network conditions. Summary of the Invention

[0006] The purpose of this invention is to provide a robot remote adaptive data acquisition control system and method based on dynamic value assessment, so as to solve the above-mentioned shortcomings in the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A robot remote adaptive data acquisition control system and method based on dynamic value assessment includes a data acquisition terminal deployed locally on the robot and a central processing terminal deployed in a remote monitoring center, wherein the data acquisition terminal and the central processing terminal are connected through a wireless communication network. The data acquisition terminal includes: The sensor module is used to collect multi-dimensional state data during robot operation. The multi-dimensional state data includes at least position information, posture information, joint torque, motor current, temperature and visual image data. A local processing unit, configured to perform the following steps: Receive raw data stream from the sensor module and preprocess the raw data stream, including filtering, noise reduction and timestamp alignment; The dynamic information value factor IV(t) for each data point is calculated in real time and is determined by the following formula:

[0008] Where ΔS(t) represents the absolute value of the first-order difference between the current data point and the previous sampling point, used to quantify the rate of change of the data; S_max is the maximum possible range of this type of data, used for normalization; E(t) is the entropy factor calculated based on the variance of recent data, used to characterize the uncertainty of the data, and the value of E(t) increases when the data fluctuates drastically; C(t) is the context association factor, which is assigned different weights according to the current task stage of the robot (such as startup, high-speed operation, precise positioning, emergency stop), with higher weights in critical task stages; P_anomaly(t) is the instantaneous anomaly probability calculated in real time based on a pre-trained lightweight anomaly detection model; For the corresponding configurable weight coefficients, and satisfying ; An adaptive sampling control module, which is connected to the local processing unit, is used to compare the calculated dynamic information value factor IV(t) with a high threshold Th_high and a low threshold Th_low, and dynamically adjust the sampling frequency F_sampling according to the comparison results: when IV(t)>Th_high, switch the sampling frequency to the highest gear F_max to ensure capturing all details of high-value or abnormal events; when Th_low≤IV(t)≤Th_high, maintain the current standard sampling frequency F_normal; when IV(t)<Th_low, reduce the sampling frequency to the lowest gear F_min to reduce redundant data collection and energy consumption; A data compression and selective transmission module, which receives the data after adaptive sampling and implements a selective data transmission strategy: for data segments where IV(t) exceeds Th_high, perform lossless compression and add high-priority tags; for data with lower IV(t), use a lossy compression algorithm and record its dynamic information value factor in the data packet for reference by the remote end; The central processing end includes: A communication interface for receiving data packets from the data acquisition end; A data fusion and reconstruction engine, which is configured to parse the received data packets, and according to the embedded dynamic information value factor and priority tags, precisely reconstruct high-value data segments, and perform smooth interpolation or statistical analysis reconstruction on low-value data segments to generate a continuous and complete robot state time series curve; A global policy optimizer, which based on historical data and analysis results, uses a machine learning model to optimize and update the weight coefficients to in the dynamic information value factor calculation formula, as well as the thresholds Th_high and Th_low periodically, and send the optimized parameters to the local processing unit of the data acquisition end to achieve closed-loop feedback and continuous improvement of the acquisition strategy; A human-machine interaction interface for visually displaying the reconstructed robot state, system performance indicators, and optimization suggestions generated by the global policy optimizer.

[0009] In a preferred embodiment of the present invention, the preprocessing process of the local processing unit for the original data stream specifically includes: using a Kalman filter to smooth the position and attitude information to reduce random errors caused by sensor noise; for analog signals such as motor current and temperature, using a median filter combined with a moving average window to effectively suppress pulse interference; and using an interpolation method based on a high-precision hardware clock for timestamp alignment to ensure that data from different sensors are strictly synchronized in the time domain, with a synchronization accuracy better than 1 millisecond, laying a solid foundation for the subsequent accurate calculation of the data change rate ΔS(t) and entropy factor E(t), and avoiding value assessment distortion caused by time asynchrony.

[0010] In a preferred embodiment of the present invention, the calculation of the entropy factor E(t) is specifically implemented as follows: within a sliding time window of length N, the sample variance σ² of the data sequence within the window is calculated, and then calculated using the formula... Mapping it to entropy, this logarithmic function form can amplify the contribution of significant variance and avoid the situation where the entropy calculation is negative infinity when the data is absolutely stable (σ² is close to 0). This makes the entropy factor E(t) always positive and has nonlinear response characteristics to data fluctuations, which is more in line with the measurement principle of "uncertainty" in information theory.

[0011] In a preferred embodiment of the present invention, the dynamic frequency adjustment strategy of the adaptive sampling control module is further refined: when IV(t) exceeds Th_high for the first time, the system not only immediately switches to F_max, but also starts a "high-value data retention period" timer. During this retention period, even if IV(t) briefly drops, the sampling frequency will remain at F_max until the retention period ends or IV(t) remains below Th_low for a period of time. This mechanism effectively prevents frequent fluctuations in the sampling frequency near the fluctuation boundary, ensures complete capture of important event processes, and improves system stability and data continuity.

[0012] In a preferred embodiment of the present invention, the selective transmission strategy of the data compression and selective transmission module further includes a bandwidth awareness function: the module monitors the available bandwidth B_available and transmission delay L_delay of the current wireless link in real time. When the bandwidth is tight or the delay increases, it automatically increases the data transmission threshold of IV(t), and only sends the data with the highest dynamic information value factor. At the same time, the data below the threshold is temporarily stored in the local cache. After the network conditions improve, or according to the instructions of the remote end, the statistical summary (such as mean, maximum, minimum value) of the cached data is selectively uploaded in batch processing instead of all the original data, thereby achieving the optimal allocation of communication resources on the most critical data.

[0013] In a preferred embodiment of the present invention, the workflow of the data fusion and reconstruction engine includes: for data packets marked as high priority and losslessly compressed, the engine executes the corresponding lossless decompression algorithm to accurately restore the original data; for regular data packets using lossy compression and marked with IV(t) values, the engine, after decompression, will refer to its IV(t) value to determine the reconstruction granularity. For data segments with extremely low IV(t), it may only record its time interval and statistical characteristics, and display them on the human-computer interaction interface in the form of a shaded band or statistical value, rather than drawing each data point, thereby significantly reducing the data storage and display pressure on the remote end while ensuring the accuracy of the overall trend.

[0014] In a preferred embodiment of the present invention, the global policy optimizer uses a reinforcement learning model as its machine learning model. It treats the entire data acquisition and control system as an intelligent agent, considers different combinations of weight coefficients and thresholds as different action strategies, and uses the overall system objective (such as "minimizing the amount of data transmitted per unit time while meeting specific data analysis accuracy" or "maximizing the integrity of abnormal event capture") as the reward function. Through continuous interaction with the environment (i.e., actual robot operation data and network conditions), it learns autonomously and converges to a set of optimal or near-optimal acquisition and control parameters, enabling the system to have long-term self-optimization capabilities.

[0015] In a preferred embodiment of the present invention, the human-computer interaction interface also integrates a decision support function. This function compares the original adaptive sampling data with the simulated data under the "ideal" sampling strategy suggested by the global policy optimizer, and visually demonstrates to the administrator, in a visual manner (such as a side-by-side curve comparison chart or an efficiency improvement percentage dashboard), the benefits that may be brought about by adopting optimized parameters, such as a reduction in data volume and an increase in anomaly detection rate. This provides the administrator with an intuitive and reliable basis for making the final decision on whether to adopt the parameter update recommended by the system, thereby enhancing the effectiveness of human-in-the-loop control.

[0016] In a preferred embodiment of the present invention, the data acquisition and control method includes the following steps: a sensor module continuously acquires multi-dimensional state raw data of the robot body; a local processing unit preprocesses the raw data and calculates the dynamic information value factor IV(t) of each data point in real time; an adaptive sampling control module dynamically adjusts the actual sampling frequency of the sensor module based on the comparison result of IV(t) and a preset threshold; a data compression and selective transmission module selects an appropriate compression algorithm and transmission priority based on the value of the data and the current network status, packages the data, and sends it to a remote center; a central processing unit at the remote center receives the data and performs fusion, reconstruction, and storage; a global policy optimizer periodically analyzes historical data, optimizes the acquisition strategy parameters, and issues updates; the entire process is executed cyclically, forming a closed-loop adaptive control flow of perception-evaluation-decision-execution-optimization.

[0017] In the above technical solutions, the present invention provides a robot remote adaptive data acquisition control system and method based on dynamic value assessment. (1) The core of the present invention is the introduction of dynamic information value factor (IV). This factor is a multi-dimensional comprehensive quantitative indicator, rather than a single judgment standard. It can not only sense the "physical change" (ΔS) of the data, but also assess the "uncertainty" (E) of the data, combine the "task context" (C) of the robot operation, and identify "potential risks" (P_anomaly). Based on the real-time calculation of IV, the system can make judgments like an experienced engineer: for regular data in the stable operation phase, the sampling frequency is automatically reduced to reduce redundancy; once the increase in data uncertainty or the entry into the critical task phase is detected, even if the absolute change is not large, the attention will be maintained or increased; when a suspected anomaly occurs, the "panoramic recording" mode is immediately activated. This "on-demand allocation" acquisition strategy fundamentally avoids the drawbacks of fixed-frequency sampling and simple thresholding methods, allowing limited sampling, computation, and communication resources to be concentrated on the most valuable information points. This results in information gains far exceeding those of traditional methods with the same resource consumption, or significantly reduces system energy consumption and operating costs while acquiring the same amount of information. (2) This invention is not a static system, but rather a complete closed loop designed from terminal acquisition to cloud analysis and feedback optimization. The remote global policy optimizer continuously analyzes historical data and evaluates the effect of different acquisition parameters (such as weight coefficients and thresholds in the IV formula) on the final system objectives (such as data quality and bandwidth savings). Through machine learning (especially reinforcement learning) algorithms, the system can autonomously learn and adjust these key parameters to adapt to the characteristics of different robot models, different tasks, and even different network environments. For example, for robots performing precision assembly tasks, the system may autonomously learn to increase the weights of the context factor (C) and the anomaly factor (P_anomaly); while in a harsh field environment with poor network conditions, the system may learn to increase the transmission threshold to prioritize the smooth transmission of key instructions. This continuous self-optimization capability enables the system to evolve continuously, maintain optimal working condition over a long period, reduce the cost of manual maintenance and parameter tuning, and improve the system's robustness and applicability. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1This is a schematic diagram illustrating the steps of an embodiment of a robot remote adaptive data acquisition control system and method based on dynamic value assessment according to the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0021] like Figure 1 As shown in the figure, the robot remote adaptive data acquisition and control system based on dynamic value assessment provided by the present invention includes a data acquisition terminal deployed locally on the robot and a central processing terminal deployed in a remote monitoring center. The data acquisition terminal and the central processing terminal are connected through a wireless communication network. The data acquisition terminal includes: The sensor module is used to collect multi-dimensional state data during robot operation. The multi-dimensional state data includes at least position information, posture information, joint torque, motor current, temperature and visual image data. A local processing unit, configured to perform the following steps: Receive raw data streams from sensor modules and preprocess the raw data streams, including filtering, noise reduction, and timestamp alignment; The dynamic information value factor IV(t) for each data point is calculated in real time and is determined by the following formula:

[0022] Where ΔS(t) represents the absolute value of the first-order difference between the current data point and the previous sampling point, used to quantify the rate of change of the data; S_max is the maximum possible range of this type of data, used for normalization; E(t) is the entropy factor calculated based on the variance of recent data, used to characterize the uncertainty of the data, and the value of E(t) increases when the data fluctuates drastically; C(t) is the context association factor, which is assigned different weights according to the current task stage of the robot (such as startup, high-speed operation, precise positioning, emergency stop), with higher weights in critical task stages; P_anomaly(t) is the instantaneous anomaly probability calculated in real time based on a pre-trained lightweight anomaly detection model; For the corresponding configurable weight coefficients, and satisfying ; An adaptive sampling control module, which is connected to the local processing unit, is used to compare the calculated dynamic information value factor IV(t) with a high threshold Th_high and a low threshold Th_low, and dynamically adjust the sampling frequency F_sampling according to the comparison results: when IV(t)>Th_high, switch the sampling frequency to the highest gear F_max to ensure capturing all details of high-value or abnormal events; when Th_low≤IV(t)≤Th_high, maintain the current standard sampling frequency F_normal; when IV(t)<Th_low, reduce the sampling frequency to the lowest gear F_min to reduce redundant data collection and energy consumption; A data compression and selective transmission module, which receives the data after adaptive sampling and implements a selective data transmission strategy: for data segments where IV(t) exceeds Th_high, perform lossless compression and add high-priority tags; for data with lower IV(t), use a lossy compression algorithm and record its dynamic information value factor in the data packet for reference by the remote end; The central processing end includes: A communication interface for receiving data packets from the data collection end; A data fusion and reconstruction engine, which is configured to parse the received data packets, and according to the embedded dynamic information value factor and priority tags, precisely reconstruct high-value data segments, and perform smooth interpolation or statistical analysis reconstruction on low-value data segments to generate a continuous and complete robot state time series curve; A global policy optimizer, which based on historical data and analysis results, uses a machine learning model to optimize and update the weight coefficients to in the dynamic information value factor calculation formula, as well as the thresholds Th_high and Th_low periodically, and send the optimized parameters to the local processing unit of the data collection end to achieve closed-loop feedback and continuous improvement of the collection strategy; A human-machine interaction interface for visually displaying the reconstructed robot state, system performance indicators, and optimization suggestions generated by the global policy optimizer.

[0023] In a preferred embodiment of the present invention, the preprocessing process of the local processing unit for the original data stream specifically includes: using a Kalman filter to smooth the position and attitude information to reduce random errors caused by sensor noise; for analog signals such as motor current and temperature, using a median filter combined with a moving average window to effectively suppress pulse interference; and using an interpolation method based on a high-precision hardware clock for timestamp alignment to ensure that data from different sensors are strictly synchronized in the time domain, with a synchronization accuracy better than 1 millisecond, laying a solid foundation for the subsequent accurate calculation of the data change rate ΔS(t) and entropy factor E(t), and avoiding value assessment distortion caused by time asynchrony.

[0024] In a preferred embodiment of the present invention, the calculation of the entropy factor E(t) is specifically implemented as follows: within a sliding time window of length N, the sample variance σ² of the data sequence within the window is calculated, and then it is mapped to the entropy value through the formula E(t) = log(1 + σ²). This logarithmic function form can amplify the contribution of significant variance and avoid the situation where the entropy value calculation is negative infinity when the data is absolutely stable (σ² is close to 0). This makes the entropy factor E(t) always positive and has nonlinear response characteristics to data fluctuations, which is more in line with the measurement principle of "uncertainty" in information theory.

[0025] In a preferred embodiment of the present invention, the dynamic frequency adjustment strategy of the adaptive sampling control module is further refined: when IV(t) exceeds Th_high for the first time, the system not only immediately switches to F_max, but also starts a "high-value data retention period" timer. During this retention period, even if IV(t) drops briefly, the sampling frequency will remain at F_max until the retention period ends or IV(t) remains below Th_low for a period of time. This mechanism effectively prevents frequent fluctuations in the sampling frequency near the fluctuation boundary, ensures complete capture of important event processes, and improves system stability and data continuity.

[0026] In a preferred embodiment of the present invention, the selective transmission strategy of the data compression and selective transmission module further includes a bandwidth awareness function: the module monitors the available bandwidth B_available and transmission delay L_delay of the current wireless link in real time. When the bandwidth is tight or the delay increases, it automatically increases the data transmission threshold of IV(t) and only sends the data with the highest dynamic information value factor. At the same time, the data below the threshold is temporarily stored in the local cache. After the network conditions improve, or according to the instructions of the remote end, the statistical summary (such as mean, maximum, minimum value) of the cached data is selectively uploaded in batch processing mode instead of all the original data, thereby realizing the optimal allocation of communication resources on the most critical data.

[0027] In a preferred embodiment of the present invention, the workflow of the data fusion and reconstruction engine includes: for data packets marked as high priority and losslessly compressed, the engine executes the corresponding lossless decompression algorithm to accurately restore the original data; for regular data packets that use lossy compression and are marked with IV(t) values, the engine will refer to their IV(t) values ​​to determine the reconstruction granularity after decompression. For data segments with extremely low IV(t), only their time interval and statistical characteristics may be recorded, and displayed on the human-computer interaction interface in the form of a shaded band or statistical values, rather than drawing each data point, thereby significantly reducing the data storage and display pressure on the remote end while ensuring the accuracy of the overall trend.

[0028] In a preferred embodiment of the present invention, the machine learning model used by the global policy optimizer is a reinforcement learning model. It regards the entire data acquisition and control system as an intelligent agent, regards different weight coefficients and threshold combinations as different action strategies, and uses the overall system goal (such as "minimizing the amount of data transmitted per unit time while meeting specific data analysis accuracy" or "maximizing the integrity of abnormal event capture") as the reward function. Through continuous interaction with the environment (i.e., actual robot operation data and network conditions), it learns autonomously and converges to a set of optimal or near-optimal acquisition and control parameters, enabling the system to have long-term self-optimization capabilities.

[0029] In a preferred embodiment of the present invention, the human-computer interaction interface also integrates a decision support function. This function compares the original adaptive sampling data with the simulated data under the "ideal" sampling strategy suggested by the global policy optimizer, and visually demonstrates to the administrator, in a visual manner (such as a side-by-side curve comparison chart or an efficiency improvement percentage dashboard), the benefits that may be brought about by adopting optimized parameters, such as a reduction in data volume and an increase in anomaly detection rate. This provides the administrator with an intuitive and reliable basis for making the final decision on whether to adopt the parameter update recommended by the system, thereby enhancing the effectiveness of human-in-the-loop control.

[0030] In a preferred embodiment of the present invention, the data acquisition and control method includes the following steps: a sensor module continuously acquires multi-dimensional state raw data of the robot body; a local processing unit preprocesses the raw data and calculates the dynamic information value factor IV(t) of each data point in real time; an adaptive sampling control module dynamically adjusts the actual sampling frequency of the sensor module based on the comparison result of IV(t) and a preset threshold; a data compression and selective transmission module selects an appropriate compression algorithm and transmission priority based on the value of the data and the current network status, and packages and sends the data to a remote center; a central processing unit at the remote center receives the data and performs fusion, reconstruction, and storage; a global policy optimizer periodically analyzes historical data, optimizes the acquisition strategy parameters, and issues updates; the entire process is executed cyclically, forming a closed-loop adaptive control flow of perception-evaluation-decision-execution-optimization. Example 1

[0031] A robot remote adaptive data acquisition control system based on dynamic value evaluation, including a data acquisition end deployed locally on the robot and a central processing end deployed in a remote monitoring center, where the data acquisition end is connected to the central processing end through a wireless communication network; The data acquisition end includes: A sensor module for collecting multi-dimensional state data during the operation of the robot. The multi-dimensional state data at least includes position information, attitude information, joint torque, motor current, temperature, and visual image data; A local processing unit configured to perform the following steps: Receive the original data stream from the sensor module and preprocess the original data stream, including filtering, denoising, and timestamp alignment; Real-time calculate the dynamic information value factor for each data point. The dynamic information value factor IV(t) is determined by the following formula:

[0032] where, ΔS(t) represents the absolute value of the first-order difference between the current data point and the previous sampling point, used to quantify the data change rate; S_max is the maximum possible range of this type of data, used for normalization; E(t) is an entropy value factor calculated based on the recent data variance, used to characterize the uncertainty of the data. When the data fluctuates violently, the value of E(t) increases; C(t) is a context correlation factor, which is assigned different weights according to the current task stage of the robot (such as startup, high-speed operation, precise positioning, emergency stop), and the key task stage has a higher weight; P_anomaly(t) is the instantaneous anomaly probability calculated in real time based on a pre-trained lightweight anomaly detection model; is the corresponding configurable weight coefficient and satisfies ; An adaptive sampling control module, which is connected to the local processing unit, used to compare the calculated dynamic information value factor IV(t) with a high threshold Th_high and a low threshold Th_low, and dynamically adjust the sampling frequency F_sampling according to the comparison results: when IV(t)>Th_high, switch the sampling frequency to the highest gear F_max to ensure capturing all details of high-value or abnormal events; when Th_low≤IV(t)≤Th_high, maintain the current standard sampling frequency F_normal; when IV(t)<Th_low, reduce the sampling frequency to the lowest gear F_min to reduce redundant data acquisition and energy consumption; The data compression and selective transmission module receives adaptively sampled data and executes a selective data transmission strategy: for data segments with IV(t) exceeding Th_high, lossless compression is performed and a high-priority marker is added; for data with lower IV(t), a lossy compression algorithm is used, and its dynamic information value factor is recorded in the data packet for reference by the remote end. The central processing unit includes: The communication interface is used to receive data packets from the data acquisition terminal. The data fusion and reconstruction engine is configured to parse the received data packets, accurately reconstruct high-value data segments based on embedded dynamic information value factors and priority markers, and perform smooth interpolation or statistical analysis reconstruction on low-value data segments to generate continuous and complete robot state time series curves. The global policy optimizer, based on historical data and analysis results, uses a machine learning model to optimize the weighting coefficients in the formula for calculating the dynamic information value factor. to The thresholds Th_high and Th_low are periodically optimized and updated, and the optimized parameters are sent to the local processing unit at the data acquisition end to realize closed-loop feedback and continuous improvement of the acquisition strategy. The human-computer interaction interface is used to visually display the reconstructed robot status, system performance indicators, and optimization suggestions generated by the global policy optimizer. Example 2

[0033] A remote adaptive data acquisition and control method for robots based on dynamic value assessment. The sensor module continuously collects multidimensional state data of the robot body. The local processing unit preprocesses the raw data and calculates the dynamic information value factor IV(t) for each data point in real time; the adaptive sampling control module dynamically adjusts the actual sampling frequency of the sensor module based on the comparison result between IV(t) and the preset threshold. The data compression and selective transmission module selects an appropriate compression algorithm and transmission priority based on the value of the data and the current network conditions, and packages the data to send to the remote center. Data is received by the central processing unit at the remote center, and then fused, reconstructed, and stored. The global policy optimizer periodically analyzes historical data, optimizes the collected policy parameters, and issues updates. The entire process is executed cyclically, forming a closed-loop adaptive control flow of perception-evaluation-decision-execution-optimization.

[0034] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A robot remote adaptive data acquisition and control system based on dynamic value assessment, characterized in that, It includes a data acquisition terminal deployed locally on the robot and a central processing terminal deployed in a remote monitoring center. The data acquisition terminal is connected to the central processing terminal through a wireless communication network; The data acquisition terminal includes: A sensor module for collecting multi-dimensional state data during the operation of the robot. The multi-dimensional state data at least includes position information, attitude information, joint torque, motor current, temperature, and visual image data; A local processing unit configured to perform the following steps: Receive the original data stream from the sensor module and preprocess the original data stream, including filtering, denoising, and timestamp alignment; Real-time calculate the dynamic information value factor for each data point. The dynamic information value factor IV(t) is determined by the following formula: ； Where ΔS(t) represents the absolute value of the first-order difference between the current data point and the previous sampling point, used to quantify the rate of change of the data; S_max is the maximum possible range of this type of data, used for normalization; E(t) is the entropy factor calculated based on the variance of recent data, used to characterize the uncertainty of the data, and the value of E(t) increases when the data fluctuates drastically; C(t) is the context association factor, which is assigned different weights according to the current task stage of the robot (such as startup, high-speed operation, precise positioning, emergency stop), with higher weights in critical task stages; P_anomaly(t) is the instantaneous anomaly probability calculated in real time based on a pre-trained lightweight anomaly detection model; For the corresponding configurable weight coefficients, and satisfying ; An adaptive sampling control module, connected to the local processing unit, for comparing the calculated dynamic information value factor IV(t) with a high threshold Th_high and a low threshold Th_low, and dynamically adjusting the sampling frequency F_sampling according to the comparison results: when IV(t)>Th_high, switch the sampling frequency to the highest gear F_max to ensure capturing all details of high-value or abnormal events; when Th_low≤IV(t)≤Th_high, maintain the current standard sampling frequency F_normal; when IV(t)<Th_low, reduce the sampling frequency to the lowest gear F_min to reduce redundant data acquisition and energy consumption; A data compression and selective transmission module, which receives the data after adaptive sampling and executes a selective data transmission strategy: for data segments where IV(t) exceeds Th_high, perform lossless compression and add a high-priority mark; for data with a lower IV(t), use a lossy compression algorithm and record its dynamic information value factor in the data packet for reference by the remote end; The central processing terminal includes: A communication interface for receiving data packets from the data acquisition terminal; A data fusion and reconstruction engine configured to parse the received data packets, and perform precise reconstruction on high-value data segments and perform smooth interpolation or statistical analysis reconstruction on low-value data segments according to the embedded dynamic information value factor and priority mark to generate a continuous and complete robot state time series curve; A global policy optimizer, based on historical data and analysis results, uses a machine learning model to optimize the weight coefficients in the formula for calculating the dynamic information value factor. to The thresholds Th_high and Th_low are periodically optimized and updated, and the optimized parameters are sent to the local processing unit of the data acquisition terminal to realize closed-loop feedback and continuous improvement of the acquisition strategy. A human-computer interaction interface for visually displaying the reconstructed robot state, system performance indicators, and optimization suggestions generated by the global policy optimizer.

2. The robot remote adaptive data acquisition and control system based on dynamic value assessment according to claim 1, characterized in that, The preprocessing of the raw data stream by the local processing unit specifically includes: using a Kalman filter to smooth the position and attitude information to reduce random errors caused by sensor noise; for analog signals such as motor current and temperature, using a median filter combined with a moving average window to effectively suppress pulse interference; and using an interpolation method based on a high-precision hardware clock for timestamp alignment to ensure that data from different sensors are strictly synchronized in the time domain, with a synchronization accuracy better than 1 millisecond, laying a solid foundation for the subsequent accurate calculation of the data change rate ΔS(t) and entropy factor E(t), and avoiding value assessment distortion caused by time asynchrony.

3. A robot remote adaptive data acquisition and control system based on dynamic value assessment according to claim 2, characterized in that, The entropy factor E(t) is specifically calculated as follows: within a sliding time window of length N, the sample variance σ² of the data sequence within the window is calculated, and then mapped to the entropy value using the formula E(t) = log(1 + σ²). This logarithmic function form can amplify the contribution of significant variance and avoid the situation where the entropy value calculation is negative infinity when the data is absolutely stable (σ² is close to 0). This ensures that the entropy factor E(t) is always positive and has a nonlinear response characteristic to data fluctuations, which is more in line with the measurement principle of "uncertainty" in information theory.

4. A robot remote adaptive data acquisition and control system based on dynamic value assessment according to claim 3, characterized in that, The dynamic frequency adjustment strategy of the adaptive sampling control module is further refined: when IV(t) exceeds Th_high for the first time, the system not only immediately switches to F_max, but also starts a "high-value data hold period" timer. During this hold period, even if IV(t) drops briefly, the sampling frequency will remain at F_max until the hold period ends or IV(t) remains below Th_low for a period of time. This mechanism effectively prevents frequent fluctuations in the sampling frequency near the fluctuation boundary, ensures complete capture of important event processes, and improves system stability and data continuity.

5. A robot remote adaptive data acquisition and control system based on dynamic value assessment according to claim 1, characterized in that, The selective transmission strategy of the data compression and selective transmission module also includes a bandwidth awareness function: the module monitors the available bandwidth B_available and transmission delay L_delay of the current wireless link in real time. When the bandwidth is tight or the delay increases, it automatically increases the data transmission threshold of IV(t) and only sends the data with the highest dynamic information value factor. At the same time, the data below the threshold is temporarily stored in the local cache. After the network conditions improve, or according to the instructions of the remote end, the statistical summary of the cached data is selectively uploaded in batch processing instead of all the original data, thereby achieving the optimal allocation of communication resources on the most critical data.

6. A robot remote adaptive data acquisition and control system based on dynamic value assessment according to claim 1, characterized in that, The workflow of the data fusion and reconstruction engine includes: for data packets marked as high priority and losslessly compressed, the engine executes the corresponding lossless decompression algorithm to accurately restore the original data; for regular data packets that use lossy compression and are marked with IV(t) values, the engine will refer to their IV(t) values ​​to determine the reconstruction granularity after decompression. For data segments with extremely low IV(t), only their time interval and statistical characteristics may be recorded and displayed on the human-computer interaction interface as a shaded band or statistical value, rather than drawing each data point, thereby significantly reducing the data storage and display pressure on the remote end while ensuring the accuracy of the overall trend.

7. A robot remote adaptive data acquisition and control system based on dynamic value assessment according to claim 5, characterized in that, The global policy optimizer uses a reinforcement learning model as its machine learning model. It treats the entire data acquisition and control system as an intelligent agent, considers different combinations of weight coefficients and thresholds as different action strategies, and uses the overall system goal as the reward function. Through continuous interaction with the environment, it learns autonomously and converges to a set of optimal or near-optimal acquisition and control parameters, enabling the system to have long-term self-optimization capabilities.

8. A robot remote adaptive data acquisition and control system based on dynamic value assessment according to claim 7, characterized in that, The human-computer interface also integrates a decision support function. This function compares the original adaptive sampling data with the simulated data under the "ideal" sampling strategy suggested by the global policy optimizer, and visually shows the administrator the benefits that may be brought about by adopting optimized parameters, such as the reduction of data volume and the improvement of anomaly detection rate. This provides the administrator with an intuitive and reliable basis for the final decision on whether to adopt the system's recommended parameter update, thereby enhancing the effectiveness of human-in-the-loop control.

9. A data acquisition and control method for the system according to any one of claims 1 to 8, characterized in that, The method includes the following steps: The sensor module continuously collects multidimensional state data of the robot body. The local processing unit preprocesses the raw data and calculates the dynamic information value factor IV(t) for each data point in real time. The adaptive sampling control module dynamically adjusts the actual sampling frequency of the sensor module based on the comparison result between IV(t) and the preset threshold. The data compression and selective transmission module selects an appropriate compression algorithm and transmission priority based on the value of the data and the current network conditions, and packages the data to send to the remote center. Data is received by the central processing unit at the remote center, and then fused, reconstructed, and stored. The global policy optimizer periodically analyzes historical data, optimizes the collection policy parameters, and issues updates.

Images