Wheeled patrol robot remote control system

Abstract

The invention discloses a remote control system for a wheeled patrol robot. The remote control system comprises the following steps: S1, establishing local area network communication with a mini industrial personal computer through a main console, and initializing the system; s2, generating an advancing path through a laser radar and GPS data in combination with a dynamic path planning algorithm; s3, analyzing video data collected by a camera based on a convolutional neural network model, and identifying a dynamic object and an abnormal environment; s4, position control and posture adjustment of the robot are achieved through an FPGA control panel and a driver; s5, performing dynamic obstacle avoidance by using an ultrasonic obstacle avoidance device and a sparse clustering path optimization algorithm; s6, performing path calibration through an infrared calibration device and a nonlinear filtering algorithm; and S7, when the voltage is lower than the threshold value, automatically navigating to a charging station for charging. According to the invention, the obstacle avoidance capability, the path planning precision and the system operation stability of the robot in the patrol process are improved, and the method is suitable for patrol tasks in complex environments.

Description

Technical Field

[0001] The present invention relates to the technical field of patrol robot remote control systems, in particular to a wheeled patrol robot remote control system. Background Art

[0002] With the rapid advancement of modernization, the number and scale of various facilities, such as supermarkets, large docks, international airports, convention and exhibition centers, logistics warehouses, luxury hotels, and hospitals, are constantly expanding. At the same time, the proliferation of large-scale residential communities and event venues has exacerbated the complexity of human activities and logistics, requiring greater investment in human, material, and security resources to ensure the safe operation of systems. However, existing security technologies, which primarily rely on human patrols and camera surveillance, are no longer sufficient to meet today's large-scale and complex security needs. In this context, the use of patrol robots to replace traditional human patrols has become a viable solution. The introduction of this technology transforms the patrol management model from a traditional, human-dependent model to a technology-driven, intelligent one, providing new ideas and methods for security management in complex modern environments.

[0003] LiDAR (Light Detection and Ranging) technology has significant application value in indoor positioning. Its operating principle is primarily based on the reflection of light. By emitting laser pulses and calculating the time it takes for them to reflect back, it determines distance and thus obtains highly accurate location information. This technology is particularly effective in indoor environments, as it can provide millimeter-level accuracy, which is critical for complex indoor navigation and positioning needs. In contrast, the Global Positioning System (GPS) is primarily used for outdoor positioning, determining the receiver's geographic location by receiving signals from satellites in geosynchronous orbit. However, GPS's accuracy in outdoor environments is typically only meter-level, which may not be sufficient for applications requiring high-precision positioning. Furthermore, GPS signals are susceptible to interference from obstacles such as buildings and trees, further reducing positioning accuracy. While LiDAR offers significant advantages for indoor positioning, it can face challenges when operating over long distances or in symmetrical architectural environments. For example, LiDAR's limited scanning range can lead to position loss during long-distance movement. Furthermore, LiDAR can be unable to distinguish similar features in symmetrical architectural structures, leading to misjudgments.

[0004] The patrol robot's environmental detection capabilities rely heavily on the performance of its onboard PC. As the core computing unit, the PC processes data from various sensors, such as lidar, cameras, and ultrasonic sensors. These sensors work together to collect real-time information about the surrounding environment. When the PC's performance is insufficient to meet these data processing requirements, several issues can arise: data processing delays, data storage bottlenecks, impacted decision-making accuracy, and decreased system stability.

[0005] The ARM controller has limitations in fast servo adjustments, which impacts the patrol robot's operational efficiency. Due to its limited processing power, the ARM controller cannot achieve the high-speed and high-precision servo adjustments required to drive the actuators. This increases the time delay during the patrol robot's posture adjustments, affecting the robot's overall speed and navigation accuracy.

[0006] In the current technological landscape, patrol robots are designed to cover wide areas and perform surveillance missions. The vastness of these areas makes them difficult to effectively manage using traditional close-range control methods. Patrol robots are often deployed in security-sensitive or harsh environments, such as large industrial facilities, disaster sites, or restricted areas. These locations often present various potential hazards and extreme working conditions, such as chemical contamination, radiation, or other harmful factors, posing a threat to personnel safety. Therefore, adopting direct close-range control is not only impractical but also potentially puts operators at risk. Therefore, how to provide a remote control system for wheeled patrol robots is an urgent problem that those skilled in the art need to solve. Summary of the Invention

[0007] One object of the present invention is to propose a remote control system for a wheeled patrol robot. The present invention integrates components such as a remote console, a mini industrial computer, an FPGA control board, a lidar, a GPS positioning system, a camera, an ultrasonic obstacle avoidance device, and an infrared calibration device, and combines dynamic path planning, visual inertial navigation, and a nonlinear filtering algorithm to achieve path planning, dynamic obstacle avoidance, and high-precision path calibration for the robot in complex environments. The system has the advantages of stable operation, accurate obstacle avoidance, efficient path planning, and automatic charging.

[0008] A remote control system for a wheeled patrol robot according to an embodiment of the present invention includes the following steps:

[0009] S1. Establish a LAN communication connection with the mini industrial computer on the patrol robot through the remote console. The remote console sends task instructions to the mini industrial computer. After receiving the instructions, the mini industrial computer initializes the system and loads control parameters.

[0010] S2, the mini industrial computer, parses motion instructions to the FPGA control board through data exchange, and combines the laser radar and GPS positioning system to obtain real-time location information and environmental data, and combines the dynamic path planning algorithm to generate a preliminary travel path;

[0011] S3 and the mini industrial computer are connected to the camera hardware to collect video information of the external environment. The convolutional neural network model is used to analyze the video data in real time to identify dynamic objects and abnormal environments, and the recognition results are sent back to the main console.

[0012] S4, FPGA control board controls the motion actuator through the driver according to the instructions issued by the mini industrial computer. The motion actuator includes a servo motor, encoder and reducer to achieve position control and posture adjustment;

[0013] S5. Use the ultrasonic obstacle avoidance device to detect obstacles around the robot in real time, and dynamically adjust the obstacle avoidance path through multi-sensor data fusion and sparse clustering path optimization algorithm;

[0014] S6. Detect road markers in front and behind the robot using an infrared calibration device, and calibrate the robot's path using a combination of visual inertial navigation and a nonlinear filtering algorithm.

[0015] S7. When the voltage level of the robot is lower than the preset threshold during the execution of a task, the robot automatically saves the task status and uses infrared assisted positioning to navigate to the charging station for automatic charging, while transmitting the power status to the remote console in real time.

[0016] Optionally, the S1 includes the following steps:

[0017] S11. The remote console and the mini industrial computer on the patrol robot perform network initialization. The console and the mini industrial computer establish a physical connection through the LAN communication interface and start the data exchange channel.

[0018] S12. The remote console assigns a unique network identifier ID to the mini industrial computer, and the console determines a communication address mapping relationship with the mini industrial computer based on the identifier;

[0019] S13, the remote console sends a system startup command to the mini industrial computer through an initialization instruction, wherein the system startup command includes a control parameter set P, where P = {p1, p2, ..., p n}, p i is the initial value of the control parameter, n is the number of parameters;

[0020] S14, the mini industrial computer receives the system startup command, initializes the hardware interface of the FPGA control board, laser radar, GPS positioning system, and camera, and loads the system task schedule;

[0021] S15, the mini industrial computer performs a hardware self-test operation to generate initialization state data S, the initialization state data S={H i ,L i ,C i}, where H i Indicates the initialization status of each hardware module, L i Indicates the status of the LAN communication link, C i Indicates the current control parameter loading status;

[0022] S16. The mini industrial computer sends the initialization status data S to the remote console. The console updates the system communication status based on the initialization status data and prepares to issue task instructions.

[0023] Optionally, the S2 includes the following steps:

[0024] S21, the mini industrial computer sends a motion control instruction M to the FPGA control board, and the motion control instruction M includes the target position coordinate X t ,Y t And the path planning initialization parameter set P init ;

[0025] S22, FPGA control board activates the laser radar and GPS positioning system to obtain the current location information in real time c ,Y c and environmental data E, wherein the environmental data E includes an obstacle position set O and a path feasibility weight set W;

[0026] S23. The mini industrial computer builds a dynamic path planning optimization model. The path cost function C is defined as:

[0027]

[0028] Where k is the total number of path nodes, W i is the feasibility weight of the i-th path node; d i,i+1 N is the adjacent path node i and N i+1 The Euclidean distance between them is expressed as: T exec is the time constraint of path execution; β is the adjustment coefficient of the time constraint;

[0029] S24. Based on the path cost function C, the mini industrial computer uses a dynamic window path planning algorithm to search for the optimal path within the current time window while satisfying the obstacle avoidance constraint, and dynamically updates the path node set P. opt ;

[0030] S25, mini industrial computer will optimize the path data P optThe data is sent to the FPGA control board, which analyzes the path data and provides real-time feedback on the path execution status S. p , the path execution state S p Including the current path node execution status and dynamic path correction identification;

[0031] S26, the mini industrial computer uses the dynamic iteration mechanism of the current environment data E and the path planning model to calculate the path node set P opt Update and adjust in real time.

[0032] Optionally, S3 includes the following steps:

[0033] S31. The mini industrial computer is connected to the camera through the hardware interface, and the camera is started to collect video data frames F of the external environment. t , the video data frame F t ={f1,f2,...,f n} includes a plurality of time-continuous image frame sets; wherein, f n represents the nth frame of image acquired at time t, where n is the total number of image frames;

[0034] S32, mini industrial computer to video data frame F t Perform preprocessing, including resolution adjustment and noise filtering

[0035] and image compression to generate a pre-processed data frame set F pre ={f1′,f2′,...,f n ′}; where f n ' is the nth frame image after preprocessing;

[0036] S33, the mini industrial computer will pre-process the video data frame F pre Input the convolutional neural network model CNN, and generate the dynamic object recognition result set R through feature extraction and classification operations. d ={r1,r2,...,r k}; where r k is the feature information of the kth dynamic object detected, including the object position coordinates (X k ,Y k ) and object classification label L k ;

[0037] S34, dynamic object recognition result set R d The object position coordinates (X k ,Y k ) and classification label L k Perform data screening to eliminate redundant or misidentified object features and generate an optimized object feature set R opt={(X1,Y1,L1),(X2,Y2,L2),...,(X m ,Y m ,L m )}; where m≤k represents the number of valid object detection results;

[0038] S35, mini industrial computer to optimize the object feature set R opt Perform time series analysis to generate abnormal environment identification data E a , the abnormal environment identification data E a , including the location of the abnormal object (X e ,Y e ) and time tag T e ;

[0039] S36, mini industrial computer will identify abnormal environment data E a and the optimized object feature set R opt The data is transmitted back to the remote console through the network communication module for further data analysis and task instruction adjustment.

[0040] Optionally, the S4 includes the following steps:

[0041] S41, the FPGA control board receives the motion control instruction M issued by the mini industrial computer, and the motion control instruction M includes the target position coordinates (X t ,Y t ) and initial control parameters P init ;

[0042] S42, FPGA control board sets the target position coordinates (X t ,Y t ) and current location information (X c ,Y c ) to compare and calculate the displacement vector V and the required motion angle θ;

[0043] S43, FPGA control board generates speed control instruction S according to displacement vector V and motion angle θ v and angle control command S θ , sending control instructions to the driver;

[0044] S44, the driver receives the speed control instruction S v and angle control command S θ , the servo motor is speed-regulated and direction-controlled, the speed v of the servo motor is proportional to the displacement vector V;

[0045] S45, the driver collects the position information fed back by the encoder in real time (X f ,Y f ) and speed feedback value vf , calculate the deviation between the feedback information and the target position information to generate the position error e p and the speed error e v ;

[0046] S46, FPGA control board based on position error e p and the speed error e v , using proportional-integral-differential control algorithm to generate the corrected control instruction S c :

[0047]

[0048] Among them, K p is the proportionality coefficient, K i is the integral coefficient, K d is the differential coefficient;

[0049] S47, the driver receives the correction control instruction S c , adjust the speed and posture of the servo motor, and output the final motion command through the reducer to achieve robot position control and posture adjustment;

[0050] S48, FPGA control board will execute the motion state S e Including current position information, speed feedback information and control correction status, which are fed back to the mini industrial computer through data interaction.

[0051] Optionally, the S5 includes the following steps:

[0052] S51, the ultrasonic obstacle avoidance device transmits ultrasonic signals through the ultrasonic transmitter, receives the reflected signals, and calculates the distance to the obstacle. Among them, d i represents the distance between the obstacle and the ultrasonic sensor, v is the ultrasonic propagation speed, t i is the time interval between ultrasonic emission and reception;

[0053] S52, the ultrasonic obstacle avoidance device sends the calculated obstacle distance set D to the FPGA control board, wherein the obstacle distance set D = {d1, d2, ..., d n}; where n is the number of ultrasonic sensors;

[0054] S53, the FPGA control board receives the obstacle distance set D, combines the environmental data E collected by the laser radar and GPS system and the current location information (X c ,Y c ), generate a multi-sensor fusion dataset F = {D, E, (X c ,Y c )};

[0055] S54, based on the multi-sensor fusion data set F, the FPGA control board detects the obstacle distribution and determines the feasible path through the sparse clustering path optimization algorithm, and generates the optimized obstacle avoidance path node set P opt ;

[0056] S55, FPGA control board sets the optimized obstacle avoidance path node set P opt Send to mini industrial computer;

[0057] S56, mini industrial computer receives optimized path node set P opt , dynamically adjust the current path and re-issue the motion control instruction M' to the FPGA control board;

[0058] S57, FPGA control board updates the servo motor execution path based on the motion control instruction M', detects obstacle information in real time, and performs iterative path correction in combination with the sparse clustering path optimization algorithm.

[0059] Optionally, the S6 includes the following steps:

[0060] S61, the infrared calibration device transmits infrared signals to the front and back, receives reflected signals, and calculates the reflected signals based on the reflection time t r Calculate the landmark distance d r ;

[0061] S62, infrared calibration device based on the marker distance d r and reflection angle θ r , converted to the marker position in the coordinate system (X r ,Y r );

[0062] S63, the mini industrial computer receives the landmark coordinate set R = {(X r1 ,Y r1 ),(X r2 ,Y r2 ),...,(X rn ,Y rn )}; where n is the number of markers detected;

[0063] S64, mini industrial computer calculates path calibration error e c , the path calibration error is defined as:

[0064]

[0065] S65, mini industrial computer combined with visual inertial navigation data, based on the state transfer function f(S k ,u k ) and the observation function h(S k), use the extended Kalman filter algorithm to estimate and correct the path deviation, specifically including:

[0066] S651, Status Prediction: Path Status Calculated by the following state transfer function:

[0067]

[0068] Among them, S k is the current path status, u k is the control input, w k is the process noise; the state transfer function f(S k ,u k ) is defined as:

[0069]

[0070] Among them, v k is the linear velocity, ω k is the angular velocity; Δt is the time step; θ k is the current posture angle of the robot;

[0071] S652, Observation Update: Observation Function Among them, X r ,Y r Coordinates of markers provided for infrared calibration devices;

[0072] S653, calibration correction: using Kalman gain K k Correct the status:

[0073]

[0074] Among them, z k is the observed value, K k is the Kalman gain, defined as: Among them, P k is the prediction error covariance, H k is the observation matrix, R k is the observation noise covariance matrix;

[0075] S66, mini industrial computer will calibrate the path status (X c ',Y c ',θ') is sent to the FPGA control board, and the path status includes the calibrated position information and attitude angle;

[0076] S67, FPGA control board regenerates motion control instruction M' based on the calibrated path status and sends it to the motion actuator to achieve path correction and dynamic calibration.

[0077] Optionally, the S7 includes the following steps:

[0078] S71, the robot power management module detects the battery voltage V in real time b , the battery voltage is compared with the preset threshold V th By comparison, the voltage state is judged to be ΔV=V th -V b If ΔV≥0, the low voltage state is triggered and the automatic charging process begins.

[0079] S72. After receiving the low voltage status signal, the mini industrial computer saves the current task status T s , task status T s Including the current position (X c ,Y c ), Task Progress P t and path node set P opt :

[0080] S73, the infrared calibration device is started to detect the location information of the charging station marker (X r ,Y r );

[0081] S74, mini industrial computer based on charging station location information (X r ,Y r ) and the current robot position information (X c ,Y c ), calculate the displacement vector V to the charging station ch ;

[0082] S75, the mini industrial computer generates the motion control instruction M to go to the charging station c , the motion control instruction M c Including path planning results P ch ;

[0083] S76, FPGA control board receives motion control instruction M c , control the driver to drive the motion actuator, along the path planning result P ch Move to the charging station while monitoring the robot's position in real time (X c ',Y c ') and path offset ΔP;

[0084] S77. After the robot arrives at the charging station, it completes the docking through the infrared docking system and starts the automatic charging process; the power management module monitors the battery voltage status V in real time. b ', when the voltage reaches the safety threshold V safe When , charging is finished;

[0085] S78, during and after charging, the mini industrial computer will show the power status V b ' and task status T s The data is transmitted to the remote console via the local area network, and the remote console records the current battery status and task recovery point.

[0086] The beneficial effects of the present invention are:

[0087] The present invention proposes a remote control system for a wheeled patrol robot. Through network communication between a remote main console and a mini industrial computer, combined with a dynamic path planning algorithm and multi-sensor data fusion, it achieves efficient planning and dynamic correction of the robot's path, effectively improving the robot's path execution accuracy and real-time obstacle avoidance capabilities in complex environments, reducing errors in path planning, and is suitable for large-scale security patrol scenarios.

[0088] This invention introduces visual inertial navigation and nonlinear filtering algorithms, combined with an infrared calibration device, to perform high-precision path calibration on the robot. Through state estimation and path error correction, it effectively reduces the path drift problem and improves the navigation stability of the robot in a dynamic environment, ensuring that the robot can still achieve precise path control and calibration even when the positioning error is large.

[0089] The present invention integrates laser radar, GPS positioning system and ultrasonic obstacle avoidance device, and combines with sparse clustering path optimization algorithm to dynamically detect and avoid surrounding obstacles, generate the optimal obstacle avoidance path, and significantly improve the robot's autonomous obstacle avoidance capability and path correction efficiency in complex environments, ensuring the continuity and safety of the robot's patrol mission.

[0090] The present invention proposes an automatic charging function. When the voltage level is lower than the preset threshold, the robot uses infrared assisted positioning to navigate to the charging station for automatic charging, and transmits the power status and mission recovery point in real time, realizing the preservation of mission status and automatic endurance, thereby improving the system's operational reliability and the continuity of patrol missions.

[0091] The present invention uses a convolutional neural network model to analyze external video information collected by the camera, realizes dynamic object recognition and abnormal environment detection, combines the time series analysis mechanism to generate abnormal recognition data, optimizes the robot's perception of the external environment, and effectively improves the monitoring effect and data analysis capabilities during the patrol process. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0093] Figure 1This is a flow chart of a remote control system for a wheeled patrol robot proposed by the present invention;

[0094] Figure 2 This is a control system block diagram of a remote control system for a wheeled patrol robot proposed in the present invention;

[0095] Figure 3 This is a video processing block diagram of a remote control system for a wheeled patrol robot proposed in the present invention. DETAILED DESCRIPTION

[0096] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0097] refer to Figure 1-3 , a wheeled patrol robot remote control system, comprising the following steps:

[0098] S1. Establish a LAN communication connection with the mini industrial computer on the patrol robot through the remote console. The remote console sends task instructions to the mini industrial computer. After receiving the instructions, the mini industrial computer initializes the system and loads control parameters.

[0099] S2, the mini industrial computer, parses motion instructions to the FPGA control board through data exchange, and combines the laser radar and GPS positioning system to obtain real-time location information and environmental data, and combines the dynamic path planning algorithm to generate a preliminary travel path;

[0100] S3 and the mini industrial computer are connected to the camera hardware to collect video information of the external environment. The convolutional neural network model is used to analyze the video data in real time to identify dynamic objects and abnormal environments, and the recognition results are sent back to the main console.

[0101] S4, FPGA control board controls the motion actuator through the driver according to the instructions issued by the mini industrial computer. The motion actuator includes a servo motor, encoder and reducer to achieve position control and posture adjustment;

[0102] S5. Use the ultrasonic obstacle avoidance device to detect obstacles around the robot in real time, and dynamically adjust the obstacle avoidance path through multi-sensor data fusion and sparse clustering path optimization algorithm;

[0103] S6. Detect road markers in front and behind the robot using an infrared calibration device, and calibrate the robot's path using a combination of visual inertial navigation and a nonlinear filtering algorithm.

[0104] S7. When the voltage level of the robot is lower than the preset threshold during the execution of a task, the robot automatically saves the task status and uses infrared assisted positioning to navigate to the charging station for automatic charging, while transmitting the power status to the remote console in real time.

[0105] In this embodiment, S1 includes the following steps:

[0106] S11. The remote console and the mini industrial computer on the patrol robot perform network initialization. The console and the mini industrial computer establish a physical connection through the LAN communication interface and start the data exchange channel.

[0107] S12. The remote console assigns a unique network identifier ID to the mini industrial computer, and the console determines a communication address mapping relationship with the mini industrial computer based on the identifier;

[0108] S13, the remote console sends a system startup command to the mini industrial computer through an initialization instruction, wherein the system startup command includes a control parameter set P, where P = {p1, p2, ..., p n}, p i is the initial value of the control parameter, n is the number of parameters;

[0109] S14, the mini industrial computer receives the system startup command, initializes the hardware interface of the FPGA control board, laser radar, GPS positioning system, and camera, and loads the system task schedule;

[0110] S15, the mini industrial computer performs a hardware self-test operation to generate initialization state data S, the initialization state data S={H i ,L i ,C i}, where H i Indicates the initialization status of each hardware module, L i Indicates the status of the LAN communication link, C i Indicates the current control parameter loading status;

[0111] S16. The mini industrial computer sends the initialization status data S to the remote console. The console updates the system communication status based on the initialization status data and prepares to issue task instructions.

[0112] In this embodiment, S2 includes the following steps:

[0113] S21, the mini industrial computer sends a motion control instruction M to the FPGA control board, and the motion control instruction M includes the target position coordinate X t ,Y t And the path planning initialization parameter set P init ;

[0114] S22, FPGA control board activates the laser radar and GPS positioning system to obtain the current location information in real time c ,Y c and environmental data E, wherein the environmental data E includes an obstacle position set O and a path feasibility weight set W;

[0115] S23. The mini industrial computer builds a dynamic path planning optimization model. The path cost function C is defined as:

[0116]

[0117] Where k is the total number of path nodes, W i is the feasibility weight of the i-th path node; d i,i+1 N is the adjacent path node i and N i+1 The Euclidean distance between them is expressed as: T exec is the time constraint of path execution; β is the adjustment coefficient of the time constraint;

[0118] S24. Based on the path cost function C, the mini industrial computer uses a dynamic window path planning algorithm to search for the optimal path within the current time window while satisfying the obstacle avoidance constraint, and dynamically updates the path node set P. opt ;

[0119] S25, mini industrial computer will optimize the path data P opt The data is sent to the FPGA control board, which analyzes the path data and provides real-time feedback on the path execution status S. p , the path execution state S p Including the current path node execution status and dynamic path correction identification;

[0120] S26, the mini industrial computer uses the dynamic iteration mechanism of the current environment data E and the path planning model to calculate the path node set P opt Update and adjust in real time.

[0121] In this embodiment, S3 includes the following steps:

[0122] S31. The mini industrial computer is connected to the camera through the hardware interface, and the camera is started to collect video data frames F of the external environment. t , the video data frame F t ={f1,f2,...,f n} includes a plurality of time-continuous image frame sets; wherein, f n represents the nth frame of image acquired at time t, where n is the total number of image frames;

[0123] S32, mini industrial computer to video data frame F t Perform preprocessing, including resolution adjustment and noise filtering

[0124] and image compression to generate a pre-processed data frame set F pre ={f1′,f2′,...,f n ′}; where f n ' is the nth frame image after preprocessing;

[0125] S33, the mini industrial computer will pre-process the video data frame F pre Input the convolutional neural network model CNN, and generate the dynamic object recognition result set R through feature extraction and classification operations. d ={r1,r2,...,r k}; where r k is the feature information of the kth dynamic object detected, including the object position coordinates (X k ,Y k ) and object classification label L k ;

[0126] S34, dynamic object recognition result set R d The object position coordinates (X k ,Y k ) and classification label L k Perform data screening to eliminate redundant or misidentified object features and generate an optimized object feature set R opt ={(X1,Y1,L1),(X2,Y2,L2),...,(X m ,Y m ,L m )}; where m≤k represents the number of valid object detection results;

[0127] S35, mini industrial computer to optimize the object feature set R opt Perform time series analysis to generate abnormal environment identification data E a , the abnormal environment identification data E a , including the location of the abnormal object (X e ,Y e ) and time tag T e ;

[0128] S36, mini industrial computer will identify abnormal environment data E a and the optimized object feature set R opt The data is transmitted back to the remote console through the network communication module for further data analysis and task instruction adjustment.

[0129] In this embodiment, S4 includes the following steps:

[0130] S41, the FPGA control board receives the motion control instruction M issued by the mini industrial computer, and the motion control instruction M includes the target position coordinates (X t ,Y t ) and initial control parameters P init ;

[0131] S42, FPGA control board sets the target position coordinates (X t ,Y t ) and current location information (X c ,Y c ) to compare and calculate the displacement vector V and the required motion angle θ;

[0132] S43, FPGA control board generates speed control instruction S according to displacement vector V and motion angle θ v and angle control command S θ , sending control instructions to the driver;

[0133] S44, the driver receives the speed control instruction S v and angle control command S θ , the servo motor is speed-regulated and direction-controlled, the speed v of the servo motor is proportional to the displacement vector V;

[0134] S45, the driver collects the position information fed back by the encoder in real time (X f ,Y f ) and speed feedback value v f , calculate the deviation between the feedback information and the target position information to generate the position error e p and the speed error e v ;

[0135] S46, FPGA control board based on position error e p and the speed error e v , using proportional-integral-differential control algorithm to generate the corrected control instruction S c :

[0136]

[0137] Among them, K p is the proportionality coefficient, K i is the integral coefficient, K d is the differential coefficient;

[0138] S47, the driver receives the correction control instruction S c , adjust the speed and posture of the servo motor, and output the final motion command through the reducer to achieve robot position control and posture adjustment;

[0139] S48, FPGA control board will execute the motion state S e Including current position information, speed feedback information and control correction status, which are fed back to the mini industrial computer through data interaction.

[0140] In this embodiment, S5 includes the following steps:

[0141] S51, the ultrasonic obstacle avoidance device transmits ultrasonic signals through the ultrasonic transmitter, receives the reflected signals, and calculates the distance to the obstacle. Among them, d i represents the distance between the obstacle and the ultrasonic sensor, v is the ultrasonic propagation speed, t i is the time interval between ultrasonic emission and reception;

[0142] S52, the ultrasonic obstacle avoidance device sends the calculated obstacle distance set D to the FPGA control board, wherein the obstacle distance set D = {d1, d2, ..., d n}; where n is the number of ultrasonic sensors;

[0143] S53, the FPGA control board receives the obstacle distance set D, combines the environmental data E collected by the laser radar and GPS system and the current location information (X c ,Y c ), generate a multi-sensor fusion dataset F = {D, E, (X c ,Y c )};

[0144] S54, based on the multi-sensor fusion data set F, the FPGA control board detects the obstacle distribution and determines the feasible path through the sparse clustering path optimization algorithm, and generates the optimized obstacle avoidance path node set P opt ;

[0145] S55, FPGA control board sets the optimized obstacle avoidance path node set P opt Send to mini industrial computer;

[0146] S56, mini industrial computer receives optimized path node set P opt , dynamically adjust the current path and re-issue the motion control instruction M' to the FPGA control board;

[0147] S57, FPGA control board updates the servo motor execution path based on the motion control instruction M', detects obstacle information in real time, and performs iterative path correction in combination with the sparse clustering path optimization algorithm.

[0148] In this embodiment, S6 includes the following steps:

[0149] S61, the infrared calibration device transmits infrared signals to the front and back, receives reflected signals, and calculates the reflected signals based on the reflection time t r Calculate the landmark distance d r ;

[0150] S62, infrared calibration device based on the marker distance d r and reflection angle θ r , converted to the marker position in the coordinate system (X r ,Y r );

[0151] S63, the mini industrial computer receives the landmark coordinate set R = {(X r1 ,Y r1 ),(X r2 ,Y r2 ),...,(X rn ,Y rn )}; where n is the number of markers detected;

[0152] S64, mini industrial computer calculates path calibration error e c , the path calibration error is defined as:

[0153]

[0154] S65, mini industrial computer combined with visual inertial navigation data, based on the state transfer function f(S k ,u k ) and the observation function h(S k ), use the extended Kalman filter algorithm to estimate and correct the path deviation, specifically including:

[0155] S651, Status Prediction: Path Status Calculated by the following state transfer function:

[0156]

[0157] Among them, S k is the current path status, u k is the control input, w k is the process noise; the state transfer function f(S k ,u k ) is defined as:

[0158]

[0159] Among them, v k is the linear velocity, ω k is the angular velocity; Δt is the time step; θ k is the current posture angle of the robot;

[0160] S652, Observation Update: Observation Function Among them, X r ,Y r Coordinates of markers provided for infrared calibration devices;

[0161] S653, calibration correction: using Kalman gain K k Correct the status:

[0162]

[0163] Among them, z k is the observed value, K k is the Kalman gain, defined as: Among them, P k is the prediction error covariance, H k is the observation matrix, R k is the observation noise covariance matrix;

[0164] S66, mini industrial computer will calibrate the path status (X c ',Y c ',θ') is sent to the FPGA control board, and the path status includes the calibrated position information and attitude angle;

[0165] S67, FPGA control board regenerates motion control instruction M' based on the calibrated path status and sends it to the motion actuator to achieve path correction and dynamic calibration.

[0166] In this embodiment, S7 includes the following steps:

[0167] S71, the robot power management module detects the battery voltage V in real time b , the battery voltage is compared with the preset threshold V th By comparison, the voltage state is judged to be ΔV=V th -V b If ΔV≥0, the low voltage state is triggered and the automatic charging process begins.

[0168] S72. After receiving the low voltage status signal, the mini industrial computer saves the current task status T s , task status T s Including the current position (X c ,Y c ), Task Progress P t and path node set P opt :

[0169] S73, the infrared calibration device is started to detect the location information of the charging station marker (X r ,Y r );

[0170] S74, mini industrial computer based on charging station location information (X r ,Y r ) and the current robot position information (X c ,Y c ), calculate the displacement vector V to the charging station ch ;

[0171] S75, the mini industrial computer generates the motion control instruction M to go to the charging station c , the motion control instruction M c Including path planning results P ch ;

[0172] S76, FPGA control board receives motion control instruction M c , control the driver to drive the motion actuator, along the path planning result P ch Move to the charging station while monitoring the robot's position in real time (X c ',Y c ') and path offset ΔP;

[0173] S77. After the robot arrives at the charging station, it completes the docking through the infrared docking system and starts the automatic charging process; the power management module monitors the battery voltage status V in real time. b ', when the voltage reaches the safety threshold V safe When , charging is finished;

[0174] S78, during and after charging, the mini industrial computer will show the power status V b ' and task status T s The data is transmitted to the remote console via the local area network, and the remote console records the current battery status and task recovery point.

[0175] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A remote control system for a wheeled patrol robot, characterized in that: The steps include: S1. Establish a LAN communication connection with the mini industrial computer on the patrol robot through the remote console. The remote console sends a task instruction to the mini industrial computer. After receiving the instruction, the mini industrial computer initializes the system and loads the control parameters. S2, the mini industrial computer parses motion instructions to the FPGA control board through data interaction, and combines the laser radar and GPS positioning system to obtain real-time location information and environmental data, and combines the dynamic path planning algorithm to generate a preliminary travel path; S3, the mini industrial computer is connected to the camera hardware to collect video information of the external environment, and conducts real-time analysis of the video data through the convolutional neural network model to identify dynamic objects and abnormal environments, and transmit the recognition results back to the main console; S4, FPGA control board controls the motion actuator through the driver according to the instructions issued by the mini industrial computer, and the motion actuator includes a servo motor, an encoder and a reducer to achieve position control and posture adjustment; S5. Use the ultrasonic obstacle avoidance device to detect the obstacle information around the robot in real time, and dynamically adjust the obstacle avoidance path through multi-sensor data fusion and sparse clustering path optimization algorithm; S6. Detect the road markers in front and behind the robot through the infrared calibration device, and calibrate the robot's path based on visual inertial navigation and nonlinear filtering algorithm; S7. When the voltage level of the robot is lower than the preset threshold during the execution of a task, the robot automatically saves the task status and uses infrared assisted positioning to navigate to the charging station for automatic charging. At the same time, the power status is transmitted to the remote console in real time.

2. A wheeled patrol robot remote control system according to claim 1, characterized in that: The S1 comprises the following steps: S11, the remote console and the mini industrial computer on the patrol robot perform network initialization, the console and the mini industrial computer establish a physical connection through the LAN communication interface, and start the data interaction channel; S12, the remote console assigns a unique network identifier ID to the mini industrial computer, and the console determines a communication address mapping relationship with the mini industrial computer based on the identifier; S13, the remote console sends a system startup command to the mini industrial computer through an initialization instruction, wherein the system startup command includes a control parameter set P, where P= { p1,p2,...,p n} , p i is the initial value of the control parameter, n is the number of parameters; S14, the mini industrial computer receives the system startup command, initializes the hardware interface of the FPGA control board, laser radar, GPS positioning system, and camera, and loads the system task scheduling table; S15, the mini industrial computer performs a hardware self-check operation to generate initialization state data S, wherein the initialization state data S= { H i ,L i ,C i} , where H i Indicates the initialization status of each hardware module, L i Indicates the LAN communication link status, C i Indicates the current control parameter loading status; S16. The mini industrial computer sends the initialization status data S to the remote console. The console updates the system communication status based on the initialization status data and prepares to issue task instructions.

3. A wheeled patrol robot remote control system according to claim 1, characterized in that: The S2 comprises the following steps: S21, the mini industrial computer sends a motion control instruction M to the FPGA control board, and the motion control instruction M includes the target position coordinate X t ,Y t and path planning initialization parameter set P init ; S22, FPGA control board activates laser radar and GPS positioning system to obtain current location information in real time c ,Y c and environmental data E, wherein the environmental data E includes an obstacle position set O and a path feasibility weight set W; S23. Mini industrial computer builds a dynamic path planning optimization model, and the path cost function C is defined as: Where k is the total number of path nodes, W i is the feasibility weight of the i-th path node; d i,i+1 N is the adjacent path node i and N i+1 The Euclidean distance between them is expressed as: T exec is the time constraint of path execution; β is the adjustment coefficient of the time constraint; S24. Based on the path cost function C, the mini industrial computer uses a dynamic window path planning algorithm to search for the optimal path within the current time window under the condition of satisfying the obstacle avoidance constraint, and dynamically updates the path node set P. opt ; S25, mini industrial computer will optimize path data P opt The data is sent to the FPGA control board, which analyzes the path data and provides real-time feedback on the path execution status S p , the path execution state S p Including the current path node execution status and dynamic path correction identification; S26, the mini industrial computer uses the dynamic iteration mechanism of the current environment data E and the path planning model to calculate the path node set P opt Update and adjust in real time.

4. A wheeled patrol robot remote control system according to claim 1, characterized in that: The S3 comprises the following steps: S31, the mini industrial computer is connected to the camera through the hardware interface, and the camera is started to collect video data frames F of the external environment t , the video data frame F t ={f1,f2,...,f n } includes a plurality of time-continuous image frame sets; wherein, f n represents the nth frame of image acquired at time t, where n is the total number of image frames; S32, mini industrial computer to video data frame F t Perform preprocessing, including resolution adjustment, noise filtering and image compression, to generate a preprocessed data frame set F pre ={f1',f2',...,f n '}; where f n ' is the nth frame image after preprocessing; S33, the mini industrial computer sends the pre-processed video data frame F pre Input the convolutional neural network model CNN, and generate the dynamic object recognition result set R through feature extraction and classification operations. d ={r1,r2,...,r k }; where r k is the feature information of the kth dynamic object detected, including the object position coordinates (X k ,Y k ) and object classification label L k ; S34, dynamic object recognition result set R d The object position coordinates (X k ,Y k ) and classification label L k Perform data screening to eliminate redundant or misidentified object features and generate an optimized object feature set R opt ={(X1,Y1,L1),(X2,Y2,L2),...,(X m ,Y m ,L m )}; where m≤k represents the number of valid object detection results; S35, mini industrial computer to optimize the object feature set R opt Perform time series analysis to generate abnormal environment identification data E a , the abnormal environment identification data E a , including the location of the abnormal object (X e ,Y e ) and time label T e ; S36, mini industrial computer will identify abnormal environment data E a and the optimized object feature set R opt The data is transmitted back to the remote main console through the network communication module for further data analysis and task instruction adjustment.

5. A wheeled patrol robot remote control system according to claim 1, characterized in that: The S4 comprises the following steps: S41, the FPGA control board receives the motion control instruction M issued by the mini industrial computer, and the motion control instruction M includes the target position coordinates (X t ,Y t ) and the initial control parameter P init ; S42, FPGA control board sets the target position coordinates (X t ,Y t ) and current location information (X c ,Y c ) to compare and calculate the displacement vector V and the required motion angle θ; S43, FPGA control board generates speed control instruction S according to displacement vector V and motion angle θ v and angle control command S θ , sending control instructions to the driver; S44, the driver receives the speed control instruction S v and angle control command S θ , the speed and direction of the servo motor are adjusted, and the speed v of the servo motor is proportional to the displacement vector V; S45, the driver collects the position information fed back by the encoder in real time (X f ,Y f ) and speed feedback value v f , calculate the deviation between the feedback information and the target position information to generate the position error e p and speed error e v ; S46, FPGA control board based on position error e p and speed error e v , using the proportional-integral-differential control algorithm to generate the corrected control command S c : Among them, K p is the proportionality coefficient, K i is the integration coefficient, K d is the differential coefficient; S47, the driver receives the correction control instruction S c , adjust the speed and posture of the servo motor, and output the final motion command through the reducer to achieve robot position control and posture adjustment; S48, FPGA control board will move the execution state S e Including current position information, speed feedback information and control correction status, which are fed back to the mini industrial computer through data interaction.

6. A wheeled patrol robot remote control system according to claim 1, characterized in that: The S5 comprises the following steps: S51, the ultrasonic obstacle avoidance device transmits ultrasonic signals through the ultrasonic transmitter, receives reflected signals, and calculates the distance to the obstacle. Among them, d i represents the distance between the obstacle and the ultrasonic sensor, v is the ultrasonic propagation speed, t i is the time interval between ultrasonic emission and reception; S52, the ultrasonic obstacle avoidance device sends the calculated obstacle distance set D to the FPGA control board, wherein the obstacle distance set D = {d1, d2, ..., d n }; where n is the number of ultrasonic sensors; S53, the FPGA control board receives the obstacle distance set D, combines the environmental data E collected by the laser radar and GPS system and the current location information (X c ,Y c ), generate a multi-sensor fusion data set F = {D, E, (X c ,Y c )}; S54, based on the multi-sensor fusion data set F, the FPGA control board detects the obstacle distribution and determines the feasible path through the sparse clustering path optimization algorithm, and generates the optimized obstacle avoidance path node set P opt ; S55, the FPGA control board sets the optimized obstacle avoidance path node set P opt Send to mini industrial computer; S56, mini industrial computer receives optimized path node set P opt , dynamically adjust the current path, and re-issue the motion control instruction M' to the FPGA control board; S57, the FPGA control board updates the servo motor execution path based on the motion control instruction M', detects obstacle information in real time, and performs path iterative correction in combination with the sparse clustering path optimization algorithm.

7. A wheeled patrol robot remote control system according to claim 1, characterized in that: The S6 comprises the following steps: S61, the infrared calibration device transmits infrared signals to the front and rear, receives reflected signals, and based on the reflection time t r Calculate the distance d to the landmark r ; S62, infrared calibration device based on the marker distance d r and the reflection angle θ r , converted to the marker position in the coordinate system (X r ,Y r ); S63, the mini industrial computer receives the coordinate set of the landmark R = {(X r1 ,Y r1 ),(X r2 ,Y r2 ),...,(X rn ,Y rn )}; where n is the number of markers detected; S64, mini industrial computer calculates path calibration error e c , the path calibration error is defined as: S65, mini industrial computer combined with visual inertial navigation data, based on the state transfer function f(S k ,u k ) and the observation function h(S k ), use the extended Kalman filter algorithm to estimate and correct the path deviation, specifically including: S651, Status Prediction: Path Status Calculated by the following state transfer function: Among them, S k is the current path status, u k is the control input, w k is the process noise; the state transfer function f(S k ,u k ) is defined as: Among them, v k is the linear velocity, ω k is the angular velocity; Δt is the time step; θ k is the current posture angle of the robot; S652, Observation Update: Observation Function Among them, X r ,Y r Coordinates of markers provided for infrared calibration devices; S653, calibration correction: using Kalman gain K k To modify the status: Among them, z k is the observed value, K k is the Kalman gain, defined as: Among them, P k is the prediction error covariance, H k is the observation matrix, R k is the observation noise covariance matrix; S66, mini industrial computer will calibrate the path status ( X c ',Y c ',θ' ) Sent to the FPGA control board, the path status includes the calibrated position information and attitude angle; S67, the FPGA control board regenerates the motion control instruction M' based on the calibrated path state, and sends it to the motion actuator to achieve path correction and dynamic calibration.

8. A wheeled patrol robot remote control system according to claim 1, characterized in that: The S7 comprises the following steps: S71, the robot power management module detects the battery voltage V in real time b , the battery voltage is compared with the preset threshold V th By comparison, the voltage state is judged to be ΔV = V th -V b ; Among them, if ΔV≥0, the low voltage state is triggered and the automatic charging process begins; S72, after receiving the low voltage status signal, the mini industrial computer saves the current task status T s , task status T s Include current location ( X c ,Y c ) 、Task Progress P t and path node set P opt : S73: Infrared calibration device starts to detect the location information of charging station markers ( X r ,Y r) ; S74, mini industrial computer based on charging station location information ( X r ,Y r) and current robot position information ( X c ,Y c) , calculate the displacement vector V to the charging station ch ; S75, mini industrial computer generates motion control instruction M to go to the charging station c , the motion control instruction M c Including path planning results P ch ; S76, FPGA control board receives motion control instruction M c , control the driver to drive the motion actuator along the path planning result P ch Move to the charging station while monitoring the robot's position in real time ( X c ',Y c ' ) and path offset ΔP; S77. After the robot arrives at the charging station, it completes the docking through the infrared docking system and starts the automatic charging process; the power management module monitors the battery voltage status V in real time. b ', when the voltage reaches the safety threshold V safe When , the charging is finished; S78, during and after charging, the mini industrial computer will display the power status V b ' and task status T s The data is transmitted to the remote console via the LAN, and the remote console records the current battery status and task recovery point.

9. A remote control system for a wheeled patrol robot, characterized in that: include: The remote console is used to exchange information with the patrol robot, send task instructions and receive data sent back by the patrol robot; Mini industrial computer, used to receive task instructions from the remote console, parse instructions and execute data interaction, and generate motion control instructions; The FPGA control board interacts with the mini industrial computer to control the operating status of the robot actuator according to the motion control instructions; The driver is used to receive the control instructions sent by the FPGA control board and drive the motion actuator to perform the corresponding motion; Motion actuators, including servo motors, encoders and reducers, are used to achieve robot position control and posture adjustment; Laser radar and GPS positioning system, used to collect real-time data and location information of the robot's surrounding environment, and send the data to the mini industrial computer; Ultrasonic obstacle avoidance device, used to detect obstacle information around the robot and transmit obstacle distance data to the FPGA control board; Infrared calibration device, used to detect road markers in front and behind the robot and provide position information; A camera, used to collect video data of the robot’s surroundings; The power module is used to provide power to each component in the system.

Images