Bionic robot control method and system based on AI learning

Abstract

The invention provides a bionic robot control method and system based on AI learning, and the method comprises the steps: obtaining dynamic environment data including an obstacle position, a target state and topographic features, and collecting and transmitting real-time data to a data processing module through a sensor network; for the dynamic environment data, noise and redundant information are removed, key features are extracted, and a structured environment feature data set is generated; based on the structured environment feature data set, using a Gaussian mixture model to construct an environment model, and if the environment change exceeds a preset threshold, updating parameters of the environment model to reflect the latest environment state; according to the environment model, predicting a future environment change trend by adopting a time sequence prediction algorithm, and if the deviation between a prediction result and a current environment state exceeds a preset range, triggering a model retraining mechanism; and on the basis of the prediction result, in combination with the current state and the target task of the bionic robot, an optimal decision strategy is generated by using a reinforcement learning algorithm.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a bionic robot control method and system based on AI learning. Background Art

[0002] When the bionic robot control system makes predictions and decisions in a dynamic environment, it faces the contradiction between prediction accuracy and real-time performance. On the one hand, in order to improve the prediction accuracy, it is necessary to collect and process a large amount of environmental data, and use complex machine learning algorithms for modeling and prediction. This often requires a long computing time and a large amount of computing resource overhead, which affects the real-time response capability of the system. On the other hand, the bionic robot needs to react and make decisions quickly in a dynamic environment, which places high demands on the real-time performance of the control system. If the prediction process takes too long, the robot will not be able to make correct decisions and control in time, affecting its adaptability and task completion efficiency in a complex environment. In practical applications, the bionic robot control system also faces challenges brought by the uncertainty and variability of the environment. Obstacles, targets, terrain and other factors in the environment may change at any time, and these changes are often difficult to accurately predict. Even if advanced machine learning algorithms are used, it is difficult to fully cover all possible situations. This requires the control system to have a certain degree of robustness and adaptability, and to be able to adjust the strategy in time when the prediction results deviate or the environment changes suddenly, so as to ensure the stable operation of the robot. Summary of the invention

[0003] The present invention provides a bionic robot control method based on AI learning, which mainly includes:

[0004] Acquire dynamic environmental data including obstacle location, target status and terrain features, and collect and transmit real-time data to the data processing module through the sensor network;

[0005] For dynamic environmental data, remove noise and redundant information, extract key features, and generate a structured environmental feature dataset;

[0006] Based on the structured environmental feature data set, a Gaussian mixture model is used to build an environmental model. If the environmental change exceeds the preset threshold, the parameters of the environmental model are updated to reflect the latest environmental status.

[0007] According to the environmental model, a time series prediction algorithm is used to predict future environmental change trends. If the deviation between the prediction result and the current environmental state exceeds the preset range, the model retraining mechanism is triggered;

[0008] Based on the prediction results, combined with the current state and target task of the bionic robot, the reinforcement learning algorithm is used to generate the optimal decision strategy;

[0009] Control instructions are generated through decision-making strategies. If the control instructions conflict with the current execution status, the robustness control module is called to adjust the instruction priority. The robustness control module adapts to environmental changes by dynamically adjusting control parameters to ensure the stability and adaptability of instruction execution.

[0010] The present invention provides a bionic robot control system based on AI learning, which mainly includes:

[0011] Environmental data acquisition module, used to obtain dynamic environmental data including obstacle positions, target status and terrain features, and collect and transmit real-time data to the data processing module through the sensor network;

[0012] The data processing module is used to remove noise and redundant information, extract key features, and generate a structured environmental feature data set for dynamic environmental data;

[0013] The environment modeling module is used to build an environment model based on a structured environment feature data set using a Gaussian mixture model. If the environment change exceeds a preset threshold, the parameters of the environment model are updated to reflect the latest environment status.

[0014] The prediction analysis module is used to predict future environmental change trends based on the environmental model using a time series prediction algorithm. If the deviation between the prediction result and the current environmental state exceeds a preset range, the model retraining mechanism is triggered;

[0015] The decision generation module is used to generate the optimal decision strategy based on the prediction results, combined with the current state and target task of the bionic robot, using the reinforcement learning algorithm;

[0016] The robustness control module is used to generate control instructions through decision-making strategies. If the control instructions conflict with the current execution status, the robustness control module is called to adjust the instruction priority. The robustness control module adapts to environmental changes by dynamically adjusting control parameters to ensure the stability and adaptability of instruction execution.

[0017] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0018] The present invention discloses a bionic robot adaptive control method based on AI learning. The method collects environmental data in real time through a sensor network, denoises and extracts features from the data, and constructs a Gaussian mixture environment model. The present invention uses a time series prediction algorithm to predict environmental change trends, and dynamically updates the model according to the prediction results. Combined with the current state and task objectives of the bionic robot, the present invention uses a reinforcement learning algorithm to generate an optimal decision-making strategy, and adjusts the control instruction priority through a robustness control module to ensure the stability and adaptability of execution. This method can effectively cope with complex and changeable environments and improve the environmental adaptability and task execution efficiency of the bionic robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of a bionic robot control method based on AI learning of the present invention.

[0020] Figure 2 Schematic diagram of a bionic robot control method and system based on AI learning of the present invention.

[0021] Figure 3 This is another schematic diagram of a bionic robot control method and system based on AI learning of the present invention.

[0022] Figure 4 It is a structural schematic diagram of a bionic robot control method and system based on AI learning of the present invention. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0024] like Figure 1-4 In this embodiment, a bionic robot control method based on AI learning may specifically include:

[0025] Step S101, acquiring dynamic environment data including obstacle positions, target states and terrain features, and collecting and transmitting real-time data to a data processing module through a sensor network.

[0026] By deploying sensors such as laser radar and camera on the bionic robot, dynamic environmental data including obstacle position, target state and terrain features are collected in real time. The robot transmits the raw data collected by the sensor to the onboard data processing unit in real time through its own wireless communication module. In the onboard data processing unit, the received laser radar point cloud data and camera image data are first preprocessed. For the laser radar point cloud data, the statistical filtering method is used to remove noise and outliers, and downsampling is performed to reduce the data volume. For the camera image data, the median filter is used to remove noise, and histogram equalization is performed to enhance the image contrast. Then the preprocessed laser radar point cloud data is input into the SLAM module, and the ICP (Iterative Closest Point) algorithm is used to locate the robot position in real time and build an environmental map. The preprocessed camera image data is input into the target detection module, and the YOLO (YouOnly LookOnce) algorithm is used to detect and track the target object in real time. According to the environmental map constructed by the SLAM module, the laser radar point cloud data is rasterized to generate a two-dimensional occupancy grid map, which represents the passable area and obstacle area in the environment. At the same time, the target object bounding box information output by the target detection module is projected into the occupancy grid map to update the position information of the target object. Then, the SVM (support vector machine) algorithm is used to classify each grid in the occupancy grid map to obtain the passable probability of the grid and extract key terrain features such as slope and roughness. According to the passable probability and terrain features, the occupancy grid map is further converted into a cost map to represent the pass cost of each grid. Finally, with the current position of the robot as the starting point, the target position as the end point, the occupancy grid map as the environmental model, and the cost map as the heuristic function, the A* search algorithm is used to search for the optimal path on the grid map. Among them, the node representation of the A* algorithm is grid coordinates, and the cost of the node consists of two parts: the actual cost from the current node to the starting point, and the estimated cost from the current node to the end point, which is calculated by the cost map. During the search process, the node with the smallest cost is expanded first until the target node is found. Through the above steps, the robot can obtain an optimal navigation path from the current position to the target position, and navigate autonomously according to the path guidance, while updating the environmental map and avoiding obstacles in real time.

[0027] For example, the bionic robot collects dynamic environmental data in real time by deploying sensors such as laser radar and camera. Take a city rescue robot as an example. When it searches for survivors in the ruins, the laser radar can scan the surrounding environment, generate point cloud data, and accurately locate the position of obstacles; the camera captures high-definition images for identifying possible signs of life. The robot transmits the collected raw data to the onboard data processing unit through the wireless communication module. In practical applications, this may be an embedded system such as Raspberry Pi or an industrial-grade computer. The data processing unit first preprocesses the received data. For the laser radar point cloud data, statistical filtering is used to remove noise, such as removing abnormal points that are obviously deviated from the main point group. For the camera image, median filtering is used to eliminate noise and improve image quality. The preprocessed laser radar data is input into the SLAM module, and the ICP algorithm is used to locate the robot position in real time and build an environmental map. In the rescue scenario, this enables the robot to accurately locate itself in a complex ruin environment and draw a three-dimensional map of the surrounding environment. At the same time, the preprocessed image data is input into the target detection module, and the YOLO algorithm is used to detect and track the target object. This enables the rescue robot to quickly identify possible survivors or important items. Based on the environmental map constructed by SLAM, the lidar point cloud data is rasterized to generate a two-dimensional occupancy grid map. In the rescue mission, this map clearly marks the passable areas and obstacle areas to help the robot plan a safe path. The target object information output by the target detection module is projected into the occupancy grid map to update the target position. Next, the SVM algorithm is used to classify each grid in the occupancy grid map to obtain the passability probability of the grid. According to the degree of deformation of the building, the integrity and damage of the building are evaluated to determine whether it is at risk of collapse. If the deformation of the building is greater than the preset threshold, it is marked as a potential danger area; in the rescue environment, this step can identify potential danger areas, such as unstable structures or areas that may collapse. At the same time, key terrain features such as slope and roughness are extracted, which are crucial to determine whether the robot can pass through certain areas. Based on the passability probability and terrain features, the occupancy grid map is converted into a cost map. In the rescue scenario, this map not only takes into account physical obstacles, but also factors such as terrain difficulty and potential risks. For example, an area with a large slope may be physically accessible but may be assigned a higher cost value in the cost map. Finally, the A* search algorithm is used to search for the optimal path on the grid map, starting from the robot's current position and ending at the target position. In a rescue mission, this may be the path from the current position to the detected survivor's position. The A* algorithm considers both the actual cost and the estimated cost, and can quickly find a path that balances safety and efficiency in a complex environment.Through this series of steps, the rescue robot can autonomously navigate in complex ruins environments, update environmental information in real time, effectively avoid obstacles, and quickly reach the target location, greatly improving the search and rescue efficiency and success rate.

[0028] Step S102, for the dynamic environment data, remove noise and redundant information, extract key features, and generate a structured environment feature data set.

[0029] According to the dynamic environmental data obtained by the bionic robot, the Kalman filter algorithm is used to filter the data to remove noise and redundant information and obtain clean environmental data. For the filtered environmental data, the SI FT algorithm is used to extract key feature points to obtain feature descriptors with constant scale, rotation and brightness. According to the extracted SI FT feature points, the K-means clustering algorithm is used to cluster the feature points to obtain a set of feature points of different categories, each category representing an environmental feature. For each feature point category, statistical indicators such as mean and variance are used to calculate the statistical characteristics of the category to form a structured feature description. The structured feature descriptions of each category are combined to construct a structured environmental feature data set, and each data sample contains feature descriptions of multiple categories. The CART decision tree algorithm is used to train the environmental feature data set to establish an environmental perception model for judging the type and state of the current environment.

[0030] For example, a patrol robot in a city collects distance data through a distance sensor. The raw data is affected by the robot's posture and equipment errors, and there is noise. Kalman filtering effectively removes these interferences through two stages of prediction and update. Assuming that the robot measures a distance of 10.2 meters from an obstacle at a certain moment, the Kalman filter may output a more accurate distance estimate of 10.18 meters by combining the robot's previous posture state and measurement uncertainty. The SI FT algorithm is particularly important for extracting stable feature points in complex environments. Taking a rescue robot in a city as an example, it needs to identify obstacles under different lighting conditions and viewing angles. The SI FT algorithm can extract feature descriptors from the texture of the obstacle that do not change with scale, rotation, and brightness. For example, the bark of an oak tree may produce hundreds of feature points, each of which is described by a 128-dimensional vector describing its local gradient information. These feature points can be stably identified in different seasons and times, providing a reliable reference for robot navigation. The K-means clustering algorithm helps robots understand the structure of the environment. Taking a patrol robot in a city as an example, it needs to distinguish between buildings, vehicles, and pedestrians. By clustering SI FT feature points, the following categories may be formed: tall vertical structures (buildings), horizontal extension structures (roads), small moving objects (pedestrians), etc. Assuming that the number of clusters K is set to 5, the algorithm may obtain these 5 types of environmental features after iteration, each of which contains hundreds of similar feature points. Statistical feature description provides quantitative indicators for each type of environment. For the "building" feature points, its average height, density and texture complexity can be calculated. For example, the average height is 20 meters, the density is 0.5 feature points per square meter, and the texture complexity (based on gradient variance) is 0.8. These values ​​constitute a structured feature description, which enables the robot to quickly judge the nature of the surrounding environment. The CART decision tree algorithm establishes an environmental perception model based on these structured features. For example, the model may learn the following rules: if the average height of the feature points is >15 meters and the density is >0.3 / ㎡, it is judged as a high-rise area; if the average height is <2 meters and the moving speed is >1m / s, it is judged as a pedestrian-dense area. This enables the robot to analyze the surrounding environment in real time and make corresponding behavioral decisions, such as slowing down to avoid collisions in high-rise areas and being more vigilant in pedestrian areas. Through the synergy of this series of algorithms, the bionic robot is able to extract a high-level understanding of the environment from the raw sensor data. This not only improves the robot's adaptability in complex and dynamic environments, but also lays the foundation for subsequent path planning and task execution. For example, urban patrol robots can respond quickly to abnormal situations and improve the level of urban safety. This improvement in environmental perception allows bionic robots to play a greater role in various application scenarios and promotes the advancement of artificial intelligence and robotics.

[0031] Step S103, based on the structured environmental feature data set, use the Gaussian mixture model to build an environmental model. If the environmental change exceeds a preset threshold, the parameters of the environmental model are updated to reflect the latest environmental status.

[0032] According to the structured environmental feature data set of the bionic robot, the principal component analysis method is used to reduce the dimension of the environmental features to obtain the environmental feature vector after dimension reduction. According to the environmental feature vector after dimension reduction, the K-means clustering algorithm is used to cluster the environmental features to obtain the environmental feature clustering result. According to the environmental feature clustering result, the Gaussian mixture model is used to model each cluster to obtain the initial environmental model. According to the initial environmental model, the recursive least squares method is used to optimize the model parameters to obtain the optimized environmental model parameters. The current environmental data is obtained in real time, and the environmental feature vector is extracted from it. The Euclidean distance is calculated with the feature vector in the historical environmental feature data set to obtain the minimum distance value as the environmental change degree. If the environmental change degree exceeds the preset threshold, the environmental model update mechanism is triggered, and the model parameters are dynamically adjusted according to the latest environmental feature data using the recursive least squares method. The current environmental feature vector is matched with the adjusted environmental model parameters to obtain the cosine similarity as the environmental similarity. If the similarity is higher than the preset threshold, it is judged that the model has adapted to the latest environmental state.

[0033] For example, taking a city patrol robot as an example, the original environmental features may contain hundreds of dimensions, such as building height, road width, and pedestrian density. Through principal component analysis, these features can be compressed to 10 principal components, each of which is a linear combination of the original features. For example, the first principal component may reflect the overall size of the city, and the second principal component may represent commercial activity. This dimensionality reduction not only reduces data storage and computation, but also highlights the essential characteristics of the environment. The K-means clustering algorithm classifies the environment in the reduced feature space. Assuming that the urban environment is divided into 5 categories, the algorithm may obtain the following clustering results: high-density commercial areas, low-density residential areas, industrial parks, parks and green spaces, and transportation hubs. Each category has its characteristic center, such as the characteristic center of a high-density commercial area may be: high building density, large pedestrian flow, and complex road network. This clustering result provides the robot with a macroscopic understanding of the urban structure. The Gaussian mixture model finely models each cluster to capture the probability distribution of environmental features. Taking a high-density commercial area as an example, the model may contain multiple Gaussian components, corresponding to different commercial subtypes, such as shopping malls, office buildings, and catering blocks. Each Gaussian component has its mean and covariance matrix, which describes the characteristic distribution of the subtype. This modeling method can reflect the diversity and complexity of the environment. Recursive least squares is used to dynamically optimize the parameters of the environmental model so that the model can adapt to the gradual changes of the environment. For example, as the city develops, a certain area may gradually change from a low-density residential area to a medium-density mixed area. Recursive least squares can continuously adjust the parameters of the Gaussian mixture model, such as the mean vector and covariance matrix, based on new observations, so that the model always reflects the current state of the environment. The calculation of environmental change is achieved by comparing the Euclidean distance between the current feature vector and the historical data. Suppose the robot finds that the density of pedestrians in an area suddenly increases and the building height changes significantly during patrolling, which may cause the Euclidean distance with the historical data to exceed a preset threshold, such as 0.5. At this time, the environmental model update mechanism is triggered, and the system will re-evaluate the classification of the area, possibly adjusting it from a low-density residential area to an emerging commercial area under development. The calculation of environmental similarity uses cosine similarity to measure the degree of consistency between the current environmental feature vector and the model prediction. For example, if the cosine similarity between the observed features of an area and the high-density commercial district model reaches 0.9 or above, the robot can be sure that it is currently in a typical commercial district environment. This real-time environmental recognition capability enables the robot to quickly adjust its behavioral strategy, such as reducing the speed of travel and increasing vigilance in crowded areas. Through this series of environmental modeling and dynamic update mechanisms, the bionic robot is able to build an environmental cognition system that is both stable and flexible. This not only improves the robot's adaptability in complex urban environments, but also provides a solid foundation for intelligent decision-making and task planning.For example, patrol robots can adjust patrol frequency and focus according to the characteristics of different areas, focusing on preventing theft in commercial areas and paying attention to potential safety hazards in residential areas. This intelligent environmental perception and response mechanism has greatly improved the efficiency and accuracy of urban management.

[0034] Step S104, based on the environmental model, a time series prediction algorithm is used to predict future environmental change trends. If the deviation between the prediction result and the current environmental state exceeds a preset range, a model retraining mechanism is triggered.

[0035] Obtain historical environmental data from the environmental model and construct a time series data set, which includes environmental state parameters (such as obstacle location, target state, and terrain features) and corresponding timestamps. Preprocess the time series data, including missing value filling, outlier processing, data normalization, etc. Use the sliding window method to construct training samples, each sample contains the environmental state sequence of the past period as input, and the environmental state of the future period as output. Use the long short-term memory (LSTM) neural network algorithm to train the training samples and establish an environmental state prediction model. Optimize the prediction performance of the model by adjusting the hyperparameters of the LSTM, such as the number of layers, the number of hidden units, and the learning rate. Obtain the actual environmental state data at the current moment online, and collect parameter values ​​such as temperature, humidity, and air pressure in real time through environmental sensors. Input the current environmental state data into the prediction model to obtain the predicted value of the environmental state in the future period. Calculate the deviation between the current actual environmental state and the predicted value. The deviation calculation can use common error measurement indicators such as mean square error (MSE) or mean absolute error (MAE). Compare the deviation value with the preset threshold. The threshold can be set according to the fluctuation of historical data, such as when the deviation exceeds 2 times the fluctuation range of historical data. If the deviation exceeds the threshold, the model retraining mechanism is triggered. The latest environmental status data is added to the training set, and the environmental status prediction model is incrementally trained using online learning to adapt the model to the latest environmental change trends.

[0036] Exemplarily, the construction of the environmental prediction model of the bionic robot begins with the collection and organization of historical environmental data. These data contain rich environmental state information, such as obstacle locations, target states, and terrain features. For example, in an urban environment, obstacles may be moving vehicles or pedestrians, target states may be changes in traffic lights, and terrain features may include road slopes and road surface materials. These data are arranged in chronological order to form a multidimensional time series data set. Data preprocessing is a key step to ensure the quality of model training. In practical applications, sensors may cause data loss due to failure or signal interference. For example, some optical sensors may not work properly under severe weather conditions. At this time, interpolation methods can be used to fill in missing values, such as using linear interpolation or more complex spline interpolation methods. Outlier processing is equally important. For example, sudden extreme temperature readings may be caused by sensor failures, which need to be identified and eliminated or corrected by statistical methods. Data normalization helps to balance the scale differences of different features, so that data of different dimensions such as temperature (which may be in the range of -10 to 40°C) and humidity (0-100%) can be compared and calculated on the same scale. The sliding window method is an effective way to construct training samples. Assume that 24 hours is selected as the input window and 6 hours is selected as the prediction window. Then, each training sample contains the sequence of environmental states in the past 24 hours as input and the environmental states in the next 6 hours as output. This method allows the model to learn short-term and medium-term environmental change patterns. The choice of long short-term memory (LSTM) network is based on its ability to capture long-term dependencies. This feature is particularly important in environmental prediction because some environmental changes may be affected by events that happened hours or even days ago. For example, changes in urban traffic flow may be related to large-scale events that occurred a few days ago. The prediction performance can be optimized by adjusting the structure of LSTM, such as increasing the number of layers to capture more complex patterns, or increasing the number of hidden units to increase the model capacity. The adjustment of the learning rate affects the convergence speed and stability of the model. The acquisition of real-time environmental status data provides the latest input to the model. For example, a temperature sensor may record readings once a minute, while a humidity sensor may record once every 5 minutes. After preprocessing, these data are input into the prediction model to generate a prediction of the environmental state in the next 6 hours. The calculation and evaluation of the prediction deviation is the trigger mechanism for the dynamic update of the model. Assuming that the predicted temperature value is 25°C and the actual observed value is 28°C, the mean square error (MSE) is 9. If the temperature fluctuation range of historical data is usually within ±2°C, the threshold can be set to 4°C. When the prediction deviation exceeds this threshold, it means that the current model may no longer accurately reflect the trend of environmental changes and needs to be updated. The model's online learning mechanism ensures its continuous adaptability to environmental changes. For example, if a city starts a large-scale greening project, the temperature and humidity characteristics of the environment may gradually change.By continuously incorporating new observation data into the training set and performing incremental training on the model, the prediction results can always maintain a high level of accuracy. This dynamic update mechanism enables the bionic robot to maintain efficient decision-making capabilities in a constantly changing environment, laying the foundation for its application in complex scenarios.

[0037] Step S105, based on the prediction results, combined with the current state and target task of the bionic robot, a reinforcement learning algorithm is used to generate an optimal decision strategy.

[0038] A1. Obtain the current state information and target task requirements of the bionic robot, extract features and vectorize them as the state input of the reinforcement learning algorithm. A2. Based on the state input and the pre-defined action space, use the Q earning algorithm to evaluate and select the action. In the Q earning algorithm, the state space is the set of all possible states of the bionic robot, the action space is the set of all executable actions of the robot, and the reward function is designed according to the degree of proximity between the state after each action is executed and the target state. A3. By continuously trying actions and observing state changes and reward feedback, the Bellman equation is used to update the Q value table so that the Q value converges to obtain the optimal action value function. A4. At each decision moment, query the Q value table according to the current state, select the action with the largest Q value as the optimal decision, and form the optimal decision sequence. A5. Convert the optimal decision sequence into control instructions, drive the actuator of the bionic robot to perform the corresponding action, and obtain the state feedback after execution through the sensor. A6. Determine whether the current state meets the target task requirements. If so, output the optimal decision strategy and end; if not, use the current state as the new initial state, return to step 2 and iterate the decision again until the goal is achieved.

[0039] For example, the core of the decision-making system of bionic robots lies in the reinforcement learning algorithm, among which Ql earning is a widely used method. First, obtain the current state information and target task requirements of the robot. For example, for an indoor service robot, its state may include position coordinates, battery power, current task progress, etc. This information needs to be converted into vector representation through feature extraction for algorithm processing. Assuming that the robot's position is represented by (x, y) coordinates, the battery power is represented by percentage, and the task progress is represented by completion, then a possible state vector may be [2.5, 3.7, 80%, 60%]. The definition of the action space directly affects the robot's behavioral capabilities. For indoor service robots, actions may include moving forward, backward, turning, grabbing objects, etc. Each action needs to be quantified, for example, moving forward can be defined as [1, 0], moving backward as [-1, 0], turning left as [0, -1], and turning right as [0, 1]. This representation method enables the robot to move accurately on a two-dimensional plane. The design of the reward function is the key to the Ql earning algorithm. It determines how the robot evaluates its own behavior. For service robots, completing tasks can earn positive rewards, while colliding with obstacles or running out of power can earn negative rewards. For example, successfully delivering an item to a destination may earn a reward of +100, while each move consumes a reward of -1, which encourages the robot to find the shortest path. The update of the Q-value table is the core of the learning process. Initially, the Q-value table may be randomly initialized. As the robot continues to try different actions and observe the results, the Q-value will gradually converge. For example, if the robot finds that turning right can get closer to the target faster, then the Q-value of turning right in a certain state will gradually increase. The generation of the optimal decision sequence is based on the converged Q-value table. Assuming that the robot needs to move from room A to room B, querying the Q-value table may get a decision sequence like this: [forward, forward, turn right, forward, turn left, forward]. This sequence will be converted into specific control instructions, such as motor speed and direction. After executing the action, the robot obtains new state information through various sensors. For example, the lidar can update the surrounding environment map, the odometer can accurately calculate the moving distance, and the camera can identify whether the target position has been reached. This feedback information is used to evaluate the effectiveness of the current decision and serves as input for the next round of decision-making. The entire process is iterative, and the robot will continuously adjust its decisions to adapt to environmental changes and task requirements. This adaptive ability enables bionic robots to efficiently complete tasks in complex and dynamic environments and exhibit intelligent behaviors similar to those of living things.

[0040] Step S106, generating a control instruction through a decision strategy. If the control instruction conflicts with the current execution state, the robustness control module is called to adjust the instruction priority. The robustness control module adapts to environmental changes by dynamically adjusting control parameters to ensure the stability and adaptability of instruction execution.

[0041] Generate initial control instructions based on decision-making strategies, and obtain current system execution status information such as environmental parameters and equipment operation parameters through sensors. Perform preprocessing such as filtering and normalization on the collected status information to improve data quality. Compare and analyze the processed status information with the initial control instructions to determine whether there are logical conflicts or physical conflicts, such as instructions exceeding the equipment operation range and timing conflicts between instructions. If there is a conflict, trigger the robustness control module for processing. The robustness control module obtains relevant information such as control instructions, execution status, and environmental changes. With system stability as the control target, the controller output is dynamically adjusted through the PID control algorithm so that the system state variable can quickly track the set value under interference. The input of the PID controller is the state variable deviation, and the output is the actuator drive signal. On the basis of PID control, an adaptive control algorithm such as the MIT rule is introduced to estimate the parameters of the controlled object online, correct the PID parameters in real time, and improve the system adaptability. The adaptive rate can be dynamically adjusted according to the trend of system state changes. The control instructions are sorted according to the optimized control parameters output by the robustness control module. The evaluation indicators of instruction priority include: task urgency, execution timeliness, impact on system stability, etc. The weighted summation method is used to calculate the comprehensive priority. The instruction timestamp information is used to solve the execution order when the priority is the same. The execution module receives the optimized control instruction sequence and determines the execution order according to the instruction priority and timestamp. For key instructions with high priority and strong timeliness, the interrupt mechanism is used to preempt execution. The resource scheduling strategy adopts a priority preemption algorithm to give priority to the resources required by high-priority instructions.

[0042] Exemplarily, first, the initial control instructions are generated according to the decision strategy. These instructions may include the robot's moving direction, speed, force, etc. For example, for a handling robot, the initial instruction may be "move forward 5 meters at a speed of 0.5 meters per second". Next, the environment and equipment operating parameters are obtained through various sensors. These sensors may include cameras, lidars, accelerometers, etc. Taking the handling robot as an example, it may detect an obstacle 2 meters ahead through lidar, and sense the current load of 50kg through the force sensor. The acquired raw data needs to be preprocessed to improve the quality. For example, the accelerometer data is low-pass filtered to remove high-frequency noise, and the camera image is normalized to adapt to different lighting conditions. These processed data are more reliable and help subsequent decision-making and control. The system will compare and analyze the processed state information with the initial control instruction. In the example of the handling robot, the system may find that the instruction of "moving forward 5 meters" conflicts with the state information of "there is an obstacle 2 meters ahead". This conflict may cause a collision, so it is necessary to trigger the robustness control module for processing. The core of the robustness control module is the PID control algorithm. For a handling robot, the PID controller may use the "distance to the obstacle" as input and output the robot's speed adjustment signal. For example, when the distance to the obstacle is far, the PID controller may output a large positive speed; when approaching the obstacle, the output decreases or even becomes negative, achieving deceleration or stopping. In order to improve the adaptability of the system, adaptive control algorithms such as the MIT rule are introduced. This enables the PID parameters to be automatically adjusted according to environmental changes. For example, when the robot enters a rugged terrain from a flat ground, the MIT rule may increase the integral term weight of the PID controller to better overcome the disturbance caused by terrain changes. The optimized control instructions need to be prioritized. According to the preset evaluation index system, multi-dimensional evaluation data such as the task urgency, execution timeliness, and impact on system stability of each initial control instruction are obtained. The principal component analysis method is used to reduce the dimensionality of the multi-dimensional evaluation data, and the comprehensive evaluation score is extracted as the optimization control parameter of each instruction. The weight coefficient of each evaluation index is set according to the business rules, and the weighted sum method is used to calculate the comprehensive priority score of each instruction. The initial control instructions are sorted in descending order according to the comprehensive priority score to obtain the optimized instruction sequence. Determine whether there are instructions with the same priority score in the instruction sequence. If so, perform secondary sorting based on the time sequence of the timestamps to obtain the instruction priority. For a handling robot, the instruction "avoid collision" may be given the highest priority, followed by "maintain balance" and then "complete the task on time." This sorting ensures that the robot prioritizes safety while completing the task. Finally, the execution module executes the instructions based on the priority and timestamp. High-priority instructions may interrupt the current task.For example, if the handling robot suddenly detects a person approaching during its movement, the "avoid collision" command will be executed immediately, suspending the current handling task. This mechanism ensures that the robot can flexibly respond to various emergencies, improving overall safety and efficiency. Through this series of steps, the bionic robot can operate stably and reliably in a complex and changing environment, showing adaptability and intelligence similar to that of living things. The design of this control system not only improves the performance of the robot, but also enhances its practicality and reliability in various application scenarios.

[0043] The present invention provides a bionic robot control system based on AI learning, which mainly includes:

[0044] Environmental data acquisition module, used to obtain dynamic environmental data including obstacle positions, target status and terrain features, and collect and transmit real-time data to the data processing module through the sensor network;

[0045] The data processing module is used to remove noise and redundant information, extract key features, and generate a structured environmental feature data set for dynamic environmental data;

[0046] The environment modeling module is used to build an environment model based on a structured environment feature data set using a Gaussian mixture model. If the environment change exceeds a preset threshold, the parameters of the environment model are updated to reflect the latest environment status.

[0047] The prediction analysis module is used to predict future environmental change trends based on the environmental model using a time series prediction algorithm. If the deviation between the prediction result and the current environmental state exceeds a preset range, the model retraining mechanism is triggered;

[0048] The decision generation module is used to generate the optimal decision strategy based on the prediction results, combined with the current state and target task of the bionic robot, using the reinforcement learning algorithm;

[0049] The robustness control module is used to generate control instructions through decision-making strategies. If the control instructions conflict with the current execution status, the robustness control module is called to adjust the instruction priority. The robustness control module adapts to environmental changes by dynamically adjusting control parameters to ensure the stability and adaptability of instruction execution.

[0050] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A bionic robot control method based on AI learning, characterized in that: The method comprises: Acquire dynamic environmental data including obstacle location, target status and terrain features, and collect and transmit real-time data to the data processing module through the sensor network; For dynamic environmental data, remove noise and redundant information, extract key features, and generate a structured environmental feature dataset; Based on the structured environmental feature data set, a Gaussian mixture model is used to build an environmental model. If the environmental change exceeds the preset threshold, the parameters of the environmental model are updated to reflect the latest environmental status. According to the environmental model, a time series prediction algorithm is used to predict future environmental change trends. If the deviation between the prediction result and the current environmental state exceeds the preset range, the model retraining mechanism is triggered; Based on the prediction results, combined with the current state and target task of the bionic robot, the reinforcement learning algorithm is used to generate the optimal decision strategy; Control instructions are generated through decision-making strategies. If the control instructions conflict with the current execution status, the robustness control module is called to adjust the instruction priority. The robustness control module adapts to environmental changes by dynamically adjusting control parameters to ensure the stability and adaptability of instruction execution.

2. The method according to claim 1, characterized in that The acquisition of dynamic environment data including obstacle positions, target states and terrain features, and the collection and transmission of real-time data to a data processing module through a sensor network, includes: Obtain the laser radar point cloud data and camera image data collected by the bionic robot; Perform filtering, denoising and downsampling on the laser radar point cloud data to obtain preprocessed laser radar point cloud data; Perform filtering, denoising and histogram equalization processing on the camera image data to obtain pre-processed camera image data; The pre-processed LiDAR point cloud data is input into the SLAM module, the ICP algorithm is used to locate the robot in real time, and an environmental map is constructed; The preprocessed camera image data is input into the target detection module, and the YOLO algorithm is used to detect and track the target object in real time; According to the environment map constructed by the SLAM module, the lidar point cloud data is rasterized to generate a two-dimensional occupancy grid map; Project the target object bounding box information output by the target detection module onto the two-dimensional occupancy grid map and update the target object position information; The SVM algorithm is used to classify the grids in the two-dimensional occupancy grid map, obtain the grid passability probability, and extract the terrain characteristics of the grid; Convert the two-dimensional occupancy grid map into a cost map based on the grid traversability probability and terrain characteristics; Taking the robot's current position as the starting point, the target position as the end point, the two-dimensional occupancy grid map as the environmental model, and the cost map as the heuristic function, the A* search algorithm is used to search for the optimal navigation path; Guide the robot to navigate autonomously according to the optimal navigation path, and update the environmental map and obstacle avoidance information in real time.

3. The method according to claim 1, characterized in that The method for removing noise and redundant information, extracting key features, and generating a structured environmental feature data set for dynamic environmental data includes: Acquire dynamic environmental data, and use Kalman filter algorithm to filter the dynamic environmental data to obtain filtered environmental data; For the filtered environmental data, SIFT algorithm is used to extract key feature points to obtain feature descriptors that are invariant in scale, rotation and brightness. According to the extracted SIFT feature points, the K-means clustering algorithm is used to cluster the SIFT feature points to obtain feature point sets of different categories, where each category represents an environmental feature; For each feature point category, the statistical characteristics of the category are calculated using mean and variance statistical indicators to form a structured feature description; The structured feature descriptions of each category are combined to construct a structured environmental feature dataset, where each data sample contains feature descriptions of multiple categories; The CART decision tree algorithm is used to train the structured environmental feature data set to obtain the environmental perception model; During the operation of the bionic robot, the current environmental data is obtained, and the environmental perception model is used to analyze the current environmental data to determine the type and state of the current environment.

4. The method according to claim 1, characterized in that The method uses a Gaussian mixture model to construct an environmental model based on a structured environmental feature data set. If the environmental change exceeds a preset threshold, the parameters of the environmental model are updated to reflect the latest environmental status, including: Obtain structured environmental feature datasets; The principal component analysis method is used to reduce the dimension of the environmental feature data set to obtain the reduced-dimensional environmental feature vector; According to the environmental feature vector after dimension reduction, the K-means clustering algorithm is used to cluster and obtain the environmental feature clustering result; For each environmental feature clustering result, a Gaussian mixture model is used to build a model to obtain an initial environmental model; The recursive least square method is used to optimize the parameters of the initial environmental model to obtain the optimized environmental model parameters; Obtain current environment data and extract current environment feature vector; The Euclidean distance between the current environmental feature vector and the historical environmental feature vector in the environmental feature data set is calculated, and the minimum distance value is determined as the degree of environmental change; If the degree of environmental change exceeds the preset threshold, the environmental model update mechanism is triggered, and the environmental model parameters are dynamically adjusted using the recursive least squares method based on the latest environmental feature data; The adjusted environment model parameters are matched with the current environment feature vector to obtain the cosine similarity; If the cosine similarity is higher than a preset threshold, it is determined that the environmental model has adapted to the latest environmental state.

5. The method according to claim 1, characterized in that The time series prediction algorithm is used to predict the future environmental change trend based on the environmental model. If the deviation between the prediction result and the current environmental state exceeds a preset range, the model retraining mechanism is triggered, including: Obtain historical environmental status data in the environmental model, which includes obstacle locations, target states, terrain feature parameters, and corresponding timestamps, and construct a time series data set; Preprocessing of time series datasets, including missing value filling, outlier handling, and data normalization; The sliding window method is used to construct training samples. Each training sample contains a sequence of environmental states in the past as input and an environmental state in the future as output. The long short-term memory neural network algorithm is used to train the training samples and establish an environmental state prediction model; Obtain the actual environmental status data at the current moment, input the actual environmental status data into the environmental status prediction model, and obtain the predicted environmental status data for a period of time in the future; Calculate the deviation between the actual environment state data and the predicted environment state data. If the deviation exceeds the preset threshold, the model retraining mechanism is triggered; The latest environmental status data is added to the training samples, and the environmental status prediction model is incrementally trained using online learning to make the environmental status prediction model adapt to the latest environmental change trends.

6. The method according to claim 1, characterized in that Based on the prediction results, combined with the current state and target task of the bionic robot, the reinforcement learning algorithm is used to generate the optimal decision strategy, including: A1. Obtain the current state information and target task requirements of the bionic robot, extract features and vectorize the state information and target task requirements to obtain a state feature vector; A2. Use the state feature vector as the state input of the reinforcement learning algorithm, combine it with the preset action space, use the Q-learning algorithm to evaluate the action, and obtain the Q value of each action; Among them, in the Q-learning algorithm, the state space is the set of all states of the bionic robot, the action space is the set of all executable actions of the bionic robot, and the reward function is designed according to the degree of the state after each action is executed and the target state; A3. By continuously trying actions and observing state changes and reward feedback, the Bellman equation is used to update the Q value table so that the Q value converges to obtain the optimal action value function. A4. At each decision moment, query the Q value table according to the current state, select the action with the largest Q value as the optimal decision, and form the optimal decision sequence; A5. Convert the optimal decision sequence into control instructions, drive the actuator of the bionic robot to perform corresponding actions, and obtain the state feedback after execution through sensors; A6. Determine whether the state feedback meets the target task requirements. If so, output the optimal decision strategy and end; If not, the current state is used as the new initial state and the process returns to step A2 to iterate the decision again until the goal is achieved.

7. The method according to claim 1, characterized in that The control instruction is generated by the decision strategy. If the control instruction conflicts with the current execution state, the robustness control module is called to adjust the instruction priority. The robustness control module adapts to environmental changes by dynamically adjusting control parameters to ensure the stability and adaptability of instruction execution, including: Obtaining initial control instructions and current system execution status information, the current system execution status information includes environmental parameters and equipment operation parameters; Preprocessing the current system execution status information to obtain processed status information; Determine whether the processed state information conflicts with the initial control instruction, and if so, trigger the robustness control module; The robust control module obtains the initial control instructions and the processed state information and environmental change information, and outputs the optimized control parameters through the PID control algorithm, where the input of the PID controller is the state variable deviation and the output is the actuator drive signal; Based on the PID control algorithm, an adaptive control algorithm is used to estimate the parameters of the controlled object online, correct the PID parameters in real time, and dynamically adjust the adaptive rate according to the trend of system state changes; The initial control instructions are sorted according to the optimized control parameters to obtain the control instruction sequence. The instruction priority evaluation indicators include task urgency, execution timeliness and impact on system stability. The weighted sum method is used to calculate the comprehensive priority. The execution module receives a sequence of control instructions and determines the execution order based on the instruction priority and timestamp. For critical instructions with high priority and strong timeliness, the interrupt mechanism is used to preempt execution. The resource scheduling strategy adopts a priority preemption algorithm to give priority to the resources required by high-priority instructions.

8. A bionic robot control system based on AI learning, characterized in that: The system comprises: Environmental data acquisition module, used to obtain dynamic environmental data including obstacle positions, target status and terrain features, and collect and transmit real-time data to the data processing module through the sensor network; The data processing module is used to remove noise and redundant information, extract key features, and generate a structured environmental feature data set for dynamic environmental data; The environment modeling module is used to build an environment model based on a structured environment feature data set using a Gaussian mixture model. If the environment changes beyond a preset threshold, the parameters of the environment model are updated to reflect the latest environment status. The prediction analysis module is used to predict future environmental change trends based on the environmental model using a time series prediction algorithm. If the deviation between the prediction result and the current environmental state exceeds a preset range, the model retraining mechanism is triggered; The decision generation module is used to generate the optimal decision strategy based on the prediction results, combined with the current state and target task of the bionic robot, using the reinforcement learning algorithm; The robustness control module is used to generate control instructions through decision-making strategies. If the control instructions conflict with the current execution status, the robustness control module is called to adjust the instruction priority. The robustness control module adapts to environmental changes by dynamically adjusting control parameters to ensure the stability and adaptability of instruction execution.

Images