Multi-sensor fusion indoor robot positioning system and method

Abstract

The invention is suitable for the technical field of indoor robot positioning, and provides a multi-sensor fusion indoor robot positioning method, which comprises the following steps: data acquisition: providing environment data and self state information for an indoor robot; preprocessing sensor data, and providing input for data processing and analysis; constructing a sensor data fusion model, and integrating data of a plurality of sensors; and robot position information is updated and output in real time. The invention also provides a multi-sensor fusion indoor robot positioning system which is used for realizing the multi-sensor fusion indoor robot positioning method. According to the invention, by integrating information collected by different sensors, the positioning accuracy and stability of the indoor robot are significantly improved; the method has the characteristics of high positioning precision, strong anti-interference capability, good environmental adaptability and the like, and is suitable for multiple scenes such as indoor robot navigation systems and intelligent residences.

Description

Technical Field

[0001] The present invention belongs to the technical field of indoor robot positioning, and in particular relates to a multi-sensor fusion indoor robot positioning system and method. Background Art

[0002] With the continuous advancement and innovation of science and technology, indoor robot technology has achieved remarkable results and has been widely used in many industries, especially in smart homes, medical services, elderly care, cleaning and maintenance, etc. The popularization and application of indoor robots have greatly improved people's quality of life and work efficiency. However, to achieve autonomous navigation and intelligent operation of indoor robots, a core issue is the accuracy and reliability of positioning technology.

[0003] Indoor robot positioning technology is the basis for realizing autonomous navigation and executing complex tasks, and it is directly related to the robot's operating efficiency and intelligence level. At present, the indoor robot positioning methods on the market mostly rely on a single sensor for environmental perception and position determination. Common sensors include lidar, camera, ultrasonic sensor, infrared sensor, etc. Although these sensors can meet the positioning needs of indoor robots to a certain extent, they have some limitations in practical applications.

[0004] First, the accuracy of single sensor positioning is generally low. Due to the complex and changeable indoor environment, a single sensor often finds it difficult to cope with various situations, resulting in large errors in the positioning results. Second, the anti-interference ability of a single sensor positioning is poor. In the indoor environment, factors such as light changes, temperature fluctuations, and obstacle movement may interfere with the sensor and affect the accuracy of positioning. In addition, a single sensor may also have shortcomings such as high cost and complex installation and maintenance.

[0005] In summary, the existing indoor robot positioning methods still need to be improved in terms of accuracy, anti-interference ability and cost-effectiveness. Summary of the invention

[0006] The purpose of the embodiments of the present invention is to provide a multi-sensor fusion indoor robot positioning method, aiming to solve the problems raised in the above background technology.

[0007] The embodiment of the present invention is implemented as follows: a multi-sensor fusion indoor robot positioning method comprises the following steps:

[0008] Data acquisition, providing indoor robots with environmental data and their own status information;

[0009] Preprocess sensor data to provide input for data processing and analysis;

[0010] Build a sensor data fusion model to integrate data from multiple sensors;

[0011] Update the robot position information in real time and output it.

[0012] Preferably, the data acquisition step of providing the indoor robot with environmental data and its own status information specifically includes:

[0013] Initialize the sensor, start the sensor, and set its initial state;

[0014] Configuration parameters: adjust and set relevant parameters according to the working requirements of the sensor;

[0015] Start the collection cycle, set the start time and frequency of data collection, so that the sensor collects data according to the established cycle;

[0016] Optimize the path and method of data transmission, monitor the transmission process, and ensure that data is delivered stably and accurately within the specified time.

[0017] Preferably, the step of pre-processing sensor data to provide input for data processing and analysis specifically includes:

[0018] Data denoising operation, which uses statistical analysis methods to identify random noise and system noise in sensor data, and eliminates the identified noise types;

[0019] Data filtering operations, smoothing to reduce high-frequency noise, sharpening to enhance edge information, and specific frequency filtering to deal with interference;

[0020] The data alignment operation uses time synchronization technology to make the sensor data have the same time reference, and uses coordinate transformation methods to convert different sensor data into a unified coordinate system.

[0021] Preferably, the step of constructing a sensor data fusion model and integrating data from multiple sensors specifically includes: using Kalman filtering to optimally estimate the robot state through a state space model, including state equations and observation equations, as well as prediction and update steps, and optimizing uncertainty through a covariance matrix.

[0022] Preferably, the step of constructing a sensor data fusion model and integrating data from multiple sensors specifically includes: using particle filtering to represent state distribution through a set of weighted particles, updating the particle set through importance sampling, and resampling to concentrate particles in high probability areas, and finally estimating the state through weighted averaging.

[0023] Preferably, the step of constructing a sensor data fusion model and integrating data from multiple sensors specifically includes: using a neural network algorithm, selecting a suitable network architecture, learning the mapping relationship between sensor data and state through training, and performing state prediction at runtime.

[0024] Preferably, the step of updating the robot position information in real time and outputting it specifically includes:

[0025] Obtain sensor data, collect environmental characteristics, distance, image, and speed data in real time from the robot's various sensors;

[0026] Calculate the robot position and use the sensor data fusion model to process the collected data and perform preprocessing to calculate the robot's position and posture in three-dimensional space;

[0027] Output position information and send it to the control system for path planning and motion control.

[0028] Another object of an embodiment of the present invention is to provide a multi-sensor fusion indoor robot positioning system, which is used to implement the multi-sensor fusion indoor robot positioning method, including:

[0029] Data acquisition module, used to provide the indoor robot with environmental data and its own status information;

[0030] Preprocessing module, used to provide input for data processing and analysis;

[0031] A fusion module to integrate data from multiple sensors;

[0032] The output module is used to update and output the robot's position information in real time.

[0033] A multi-sensor fusion indoor robot positioning method provided in an embodiment of the present invention is a method for fusing multiple sensor data to achieve precise positioning. By integrating information collected by different sensors, the positioning accuracy and stability of the indoor robot are significantly improved. The embodiment of the present invention has the characteristics of high positioning accuracy, strong anti-interference ability and good environmental adaptability, and is suitable for multiple scenarios such as indoor robot navigation systems and smart homes. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flowchart of a multi-sensor fusion indoor robot positioning method provided by an embodiment of the present invention;

[0035] Figure 2 A flow chart of data acquisition steps provided by an embodiment of the present invention;

[0036] Figure 3A flowchart of steps for preprocessing sensor data provided by an embodiment of the present invention;

[0037] Figure 4 A flowchart of steps for constructing a sensor data fusion model provided by an embodiment of the present invention;

[0038] Figure 5 A flowchart of the steps of updating robot position information in real time and outputting it according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0040] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0041] like Figure 1 FIG. 1 is a flowchart of a multi-sensor fusion indoor robot positioning method provided by an embodiment of the present invention, including:

[0042] S1, data acquisition, providing indoor robots with real-time, accurate information about the environment and their own status as the basis for positioning and navigation decisions;

[0043] S2, preprocess sensor data to ensure that the information collected by the sensor is accurate and reliable, and provide high-quality input for subsequent data processing and analysis;

[0044] S3, build a sensor data fusion model to integrate data from multiple sensors to improve the accuracy and robustness of indoor robot positioning;

[0045] S4. Update the robot's position information in real time to ensure that the robot can identify its exact location in the indoor environment, which is crucial for the robot's navigation and path planning.

[0046] like Figure 2 As shown, as a preferred embodiment of the present invention, the step of data acquisition specifically includes:

[0047] S11. Initialize the sensor: Start the sensor and set its initial state to prepare it for data collection. Specifically, set various sensors on the robot to work. The initialization process may include power-on test, hardware self-test, driver loading, etc. to ensure that the hardware part of the sensor can work normally. In addition, initialization may also involve software configuration of the sensor, such as setting the communication protocol, determining the data format, etc., so that the sensor can effectively exchange data with the robot's control system.

[0048] S12. Configuration parameters: Adjust and set relevant parameters according to the working requirements of the sensor to ensure that the sensor operates as expected. Specifically, the working parameters of the sensor need to be configured in detail to adapt to specific application scenarios and environmental conditions. Parameter configuration may include setting the sensor's resolution, measurement range, sampling frequency, range, etc. For example, for a lidar, you may need to configure the scanning range and scanning density; for a camera, you may need to set the exposure time, white balance, frame rate, etc. Correct parameter configuration is essential for obtaining high-quality, highly relevant sensor data.

[0049] S13, start the collection cycle: set the start time and frequency of data collection, so that the sensor collects data according to the established cycle. Specifically, the sensor data is collected according to the predetermined time interval or trigger condition. This process needs to determine the frequency of data collection to ensure the continuity and timeliness of the data. The setting of the collection cycle needs to balance the real-time nature of the data and the processing capacity of the system to avoid data loss or processing delay. After starting the collection cycle, the sensor will automatically or manually collect data according to the set parameters to provide raw materials for subsequent data preprocessing and analysis;

[0050] S14. Ensure the stability and real-time performance of data transmission, optimize the data transmission path and method, monitor the transmission process, and ensure that the sensor data can be transmitted to the robot's data processing unit in a stable and real-time manner. Specifically, stability requires the integrity and accuracy of data during the transmission process to avoid data loss or damage; real-time performance requires that the speed of data transmission can meet the needs of real-time processing so that the robot can respond quickly to environmental changes. This may require the use of high-speed data transmission interfaces such as USB, Ethernet or wireless communication technology, and may involve data compression, buffering and synchronization technologies to improve the efficiency and reliability of data transmission.

[0051] like Figure 3 As shown, as a preferred embodiment of the present invention, the step of preprocessing sensor data, for data collected from a variety of sensors (such as laser radar, camera, inertial navigation system, etc.), specifically includes:

[0052] S21, denoising operation: Identify random noise and system noise in sensor data through statistical analysis methods, and use corresponding algorithms to eliminate the identified noise types. Specifically, for outliers in sensor data such as lidar and cameras, these outliers may be caused by environmental interference or the noise of the sensor itself. Use median filtering, bilateral filtering or wavelet transform to perform denoising to eliminate the negative impact of outliers on positioning accuracy; for inertial navigation system (INS) data, use digital filters such as low-pass filters to remove high-frequency noise while retaining low-frequency signals to ensure data accuracy;

[0053] S22, filtering operation: Filtering the sensor data to smooth the data and reduce random errors, improve the smoothness and reliability of the data, including smoothing to reduce high-frequency noise, sharpening to enhance edge information, and specific frequency filtering to deal with interference. Specifically, for time series data such as INS speed and acceleration, use a sliding average filter or a Kalman filter to reduce short-term fluctuations in the data and improve data stability; for spatial data such as laser radar point clouds, use spatial filtering techniques such as voxel filtering or radius filtering to remove isolated points, enhance the structural characteristics of the data, and improve the quality of the data;

[0054] S23. Data alignment operation: It involves time synchronization, space calibration and pre-processing before data fusion, such as normalization and feature extraction. Specifically, for the problems of inconsistent timestamps or mismatched coordinate systems that may exist in different sensor data, time synchronization technology, such as the use of high-precision clocks or synchronization signals, is used to ensure that all sensor data have the same time reference and solve the problem of inconsistent timestamps. Coordinate transformation methods, such as rotation and translation matrices, are used to convert different sensor data into a unified coordinate system, laying the foundation for subsequent data fusion.

[0055] like Figure 4As shown, as a preferred embodiment of the present invention, the step of constructing a sensor data fusion model can effectively fuse the preprocessed sensor data, thereby improving the accuracy and reliability of positioning. First, a suitable fusion algorithm is selected. According to the characteristics of the indoor environment and the sensor type, a suitable fusion algorithm is selected. For example, for a linear Gaussian system, Kalman filtering is an efficient choice (optimally estimating the robot state through a state space model, including state equations and observation equations, as well as prediction and update steps. It also handles system and observation noise and optimizes uncertainty through a covariance matrix); for a nonlinear or non-Gaussian system, a particle filter (representing the state distribution through a set of weighted particles, updating the particle set through importance sampling, and resampling to concentrate the particles in a high probability area, and finally estimating the state through weighted average) or a neural network (selecting a suitable network architecture, learning the mapping relationship between sensor data and state through training, and performing state prediction at runtime) may be more applicable, as follows:

[0056] S31. Specific steps of Kalman filter algorithm:

[0057] (1) Initialization: Setting the initial state estimate and error covariance matrix of the filter to lay the foundation for the filtering process;

[0058] (2) Prediction: Using the system’s dynamic model, the state and error covariance at the next moment are predicted to form a state prediction;

[0059] (3) Update: When new sensor data arrives, the residual between the predicted sensor reading and the actual reading is calculated, and the state estimate and error covariance are updated based on the residual and the sensor noise model, thereby refining the state estimate;

[0060] (4) Iterative optimization: Through continuous prediction and update steps, the Kalman filter can optimize the state estimate in real time and provide continuous and accurate positioning information;

[0061] S32. Specific steps of particle filtering algorithm:

[0062] (1) Initialize particles: Generate a set of particles representing possible state estimates to provide diversity for the filtering process;

[0063] (2) Importance sampling: Generate a new set of particles based on the system dynamics model and previous particle distribution to reflect the latest dynamics of the system;

[0064] (3) Weight update: Using sensor data, the weight of each particle is calculated to evaluate the importance of each particle;

[0065] Resampling: Resampling is performed according to particle weights, eliminating particles with low weights and retaining particles with high weights, thereby concentrating particles in high-probability areas;

[0066] (4) State estimation: The state estimation of the system is obtained by weighted averaging of particles, providing a robust state estimation method;

[0067] S33. Specific steps of the neural network algorithm:

[0068] (1) Design and training: Design a neural network model and learn the complex mapping relationship between sensor data and robot position through training, so that the network can capture nonlinear relationships;

[0069] (2) Forward propagation: Using the preprocessed sensor data as input, the position estimate is obtained through the forward propagation of the network to quickly obtain the positioning result;

[0070] (3) Network optimization: Optimize network weights through back propagation algorithm and gradient descent method to improve positioning accuracy and robustness;

[0071] Through the construction and implementation of the above-mentioned data fusion model, the positioning method of the embodiment of the present invention not only improves the accuracy of positioning, but also enhances the system's adaptability to environmental changes, so that indoor robots can achieve efficient and reliable positioning in complex environments. The fusion of these algorithms fully utilizes the advantages of each sensor, overcomes the limitations of a single sensor, and provides strong support for the autonomous navigation and intelligent decision-making of indoor robots.

[0072] like Figure 5 As shown, as a preferred embodiment of the present invention, the step of updating the robot position information in real time specifically includes:

[0073] S41. Sensor data collection: At the beginning of each positioning cycle, data is collected in real time from various sensors equipped on the robot. These sensors may include lidar, cameras, ultrasonic sensors, inertial measurement units (IMUs), etc. The data includes environmental characteristics, distance, images, speed, etc.; ensure the synchronization of sensor data, and align the data collected by different sensors through timestamps or synchronization signals;

[0074] S42, calculation of the robot position, inputs the collected sensor data into a pre-built sensor data fusion model, which may be a Kalman filter, particle filter or neural network:

[0075] For the Kalman filter, prediction and update steps are performed to correct the estimate of the robot's position. For the particle filter, a weighted estimate of the robot's position is obtained through importance sampling, weight update and resampling steps. For the neural network, the output is calculated through forward propagation to obtain a direct estimate of the robot's position.

[0076] In the process of model calculation, the robot's motion model and sensor model are considered to adapt to the robot's dynamic behavior and environmental changes;

[0077] S43, output robot position information: output the position information to the control system or other systems for path planning and motion control. The information output format may be a digital signal, an analog signal or a network data packet according to the requirements, and can be displayed, stored or remotely monitored in real time. Specifically, once the sensor data fusion model calculates the robot's position estimate, the position information is immediately output. The output position information includes but is not limited to the robot's coordinates (x, y, z), direction angle, speed, etc. in two-dimensional or three-dimensional space;

[0078] Position information can be output in the following ways: displayed in real time on the display screen on the robot for operator monitoring; sent to the robot's control system for navigation and path planning; transmitted to a remote monitoring system via a wireless network for use by remote operators or automation systems;

[0079] To ensure real-time performance, the output process needs to be efficient and low-latency, which is usually achieved through optimized data structures and fast communication protocols. The ability to update position information in real time is the key to the robot's intelligent behavior. It provides an accurate position basis for the robot's decision-making and improves the robot's operating efficiency and safety.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-sensor fusion indoor robot positioning method, characterized in that: The following steps are involved: Data acquisition, providing indoor robots with environmental data and their own status information; Preprocess sensor data to provide input for data processing and analysis; Build a sensor data fusion model to integrate data from multiple sensors; Update the robot position information in real time and output it.

2. The multi-sensor fusion indoor robot positioning method according to claim 1, characterized in that: The data acquisition step of providing the indoor robot with environmental data and its own status information specifically includes: Initialize the sensor, start the sensor, and set its initial state; Configuration parameters: adjust and set relevant parameters according to the working requirements of the sensor; Start the collection cycle, set the start time and frequency of data collection, so that the sensor collects data according to the established cycle; Optimize the path and method of data transmission, monitor the transmission process, and ensure that data is delivered stably and accurately within the specified time.

3. The multi-sensor fusion indoor robot positioning method according to claim 1, characterized in that: The step of preprocessing sensor data to provide input for data processing and analysis specifically includes: Data denoising operation, which uses statistical analysis methods to identify random noise and system noise in sensor data, and eliminates the identified noise types; Data filtering operations, smoothing to reduce high-frequency noise, sharpening to enhance edge information, and specific frequency filtering to deal with interference; The data alignment operation uses time synchronization technology to make the sensor data have the same time reference, and uses coordinate transformation methods to convert different sensor data into a unified coordinate system.

4. The multi-sensor fusion indoor robot positioning method according to claim 1, characterized in that: The steps of constructing a sensor data fusion model and integrating data from multiple sensors specifically include: using Kalman filtering to optimally estimate the robot state through a state space model, including state equations and observation equations, as well as prediction and update steps, and optimizing uncertainty through a covariance matrix.

5. The indoor robot positioning method based on multi-sensor fusion according to claim 1, characterized in that: The steps of constructing a sensor data fusion model and integrating data from multiple sensors specifically include: using particle filtering to represent state distribution through a set of weighted particles, updating the particle set through importance sampling, and resampling to concentrate particles in high probability areas, and finally estimating the state through weighted averaging.

6. The multi-sensor fusion indoor robot positioning method according to claim 1, characterized in that: The steps of constructing a sensor data fusion model and integrating data from multiple sensors specifically include: using a neural network algorithm, selecting a suitable network architecture, learning the mapping relationship between sensor data and state through training, and performing state prediction at runtime.

7. The indoor robot positioning method based on multi-sensor fusion according to claim 1, characterized in that: The step of updating the robot position information in real time and outputting it specifically includes: Obtain sensor data, collect environmental characteristics, distance, image, and speed data in real time from the robot's various sensors; Calculate the robot position and use the sensor data fusion model to process the collected data and perform preprocessing to calculate the robot's position and posture in three-dimensional space; Output position information and send it to the control system for path planning and motion control.

8. A multi-sensor fusion indoor robot positioning system, used to implement the multi-sensor fusion indoor robot positioning method according to any one of claims 1 to 7, characterized in that: include: Data acquisition module, used to provide the indoor robot with environmental data and its own status information; Preprocessing module, used to provide input for data processing and analysis; A fusion module to integrate data from multiple sensors; The output module is used to update and output the robot's position information in real time.

Images