Omnidirectional wheel walking equipment and anti-collision system and method

Abstract

The invention relates to the technical field of film and television shooting, in particular to omni-directional wheel walking equipment and an anti-collision system and method.The omni-directional wheel walking equipment comprises a base, a supporting table, a moving path setting module, a moving module and a shooting module, the supporting table is fixedly connected with the base and located at the top of the base, and the moving path setting module is fixedly connected with the base and located at the top of the base; the moving path is set according to shooting requirements; the moving module is used for generating a driving instruction based on the moving path and driving the corresponding omnidirectional wheel to move; and the shooting module is used for shooting in the moving process according to the shooting requirement. Therefore, an efficient, flexible and reliable mobile shooting solution can be provided for a user, and compared with a track type shooting mode, the cost can be saved.

Description

Technical Field

[0001] The present invention relates to the technical field of film and television shooting, and particularly to an omnidirectional wheel walking device, an anti-collision system and method. Background Art

[0002] With the continuous development of the film and television industry, the technology and requirements for the motion control of cameras in the film and television industry are also increasing day by day. Industrial six-axis robots have the advantages of fast motion speed, high precision, repeatability, etc., and can be used for film and television shooting to achieve various special effects such as character replication, high-speed shooting, and virtual-real combination. At the same time, ordinary current camera equipment cannot cooperate with industrial six-axis robots to meet the film and television industry with continuously increasing technical requirements.

[0003] Currently, most of the film and television industry uses manual control of jibs and track carts for complex trajectory shooting. However, manual control has the disadvantages of being greatly affected by subjective factors, low efficiency, and easy fatigue, and cannot guarantee the high efficiency and accuracy of shooting. It is difficult to achieve the ideal effect, and often a single shot needs to be taken multiple times. Summary of the Invention

[0004] The purpose of the present invention is to provide an omnidirectional wheel walking device, an anti-collision system and method, aiming to provide a user with an efficient, flexible and reliable mobile shooting solution, thereby saving costs compared with the track shooting method.

[0005] To achieve the above purpose, in the first aspect, the present invention provides an omnidirectional wheel walking device, including a base and a support platform. The support platform is fixedly connected to the base and is located on the top of the base. It further includes a moving path setting module, a moving module and a shooting module;

[0006] The moving path setting module is used to set a moving path according to shooting requirements;

[0007] The moving module is used to generate a driving instruction based on the moving path and drive the corresponding omnidirectional wheel to move;

[0008] The shooting module is used to perform shooting during the movement according to shooting requirements.

[0009] Among them, the moving path setting module includes an initial path setting unit, a control point setting unit and an optimization adjustment unit;

[0010] The initial path setting unit is used to set an initial movement path based on the shooting picture requirements and shooting terrain;

[0011] The control point setting unit is used to set control points on the initial movement path;

[0012] The optimization and adjustment unit is used to optimize the initial motion path based on the parameters of the photographing unit to obtain a shooting path.

[0013] Among them, the moving module includes a path import unit, a path decomposition unit, a driving information calculation unit, a driving unit, and a compensation unit;

[0014] The path import unit is used to import the data of the moving path;

[0015] The path decomposition unit is used to decompose the moving path into a series of discrete trajectory points, and the trajectory points include position coordinates, direction angles, and speed settings;

[0016] The driving information calculation unit is used to calculate the required speed and steering angle according to the information of each trajectory point, and then convert it into a control signal for the motor;

[0017] The driving unit is used to control the speed and position of the omnidirectional wheel based on the control signal;

[0018] The compensation unit is used to compensate for the control error of the control signal based on historical data.

[0019] Among them, the compensation unit includes a data acquisition sub-unit, an abnormal data removal sub-unit, a data classification sub-unit, a judgment model training sub-unit, an error calculation sub-unit, and an adjustment sub-unit;

[0020] The data acquisition sub-unit is used to acquire historical movement data when driving the omnidirectional wheel, and the movement data includes the motor PWM signal, the position and speed feedback by the encoder;

[0021] The abnormal data removal sub-unit is used to remove the abnormal values and missing values in the movement data;

[0022] The data classification sub-unit is used to divide the processed movement data into a training set and a validation set;

[0023] The judgment model training sub-unit is used to train the long short-term memory network with the training set to obtain a judgment model;

[0024] The error calculation sub-unit is used to calculate the adjustment error based on the judgment model and the currently acquired real-time movement data;

[0025] The adjustment sub-unit is used to adjust the control signal in advance based on the adjustment error.

[0026] Among them, the shooting module includes a camera unit, a camera path planning unit, and a monitoring unit;

[0027] The camera unit is used to be fixed on the mobile platform;

[0028] The camera path planning unit is configured to set the shooting path of the camera unit based on the moving path, and the shooting path includes a starting point, an ending point, a speed change point, and key frames;

[0029] The monitoring unit is configured to track the shooting process in real time based on the monitor.

[0030] In a second aspect, the present invention further provides an anti-collision system, including an omnidirectional wheel walking device, an anti-collision detection module, and an obstacle pushing module;

[0031] The anti-collision detection module is configured to obtain environmental data during movement and perform anti-collision detection;

[0032] The obstacle pushing module is configured to calculate the size of the obstacle after detecting the obstacle, and if the size of the obstacle is within a preset range, start the pushing mechanism to push.

[0033] Among them, the anti-collision detection module includes an initialization unit, a data acquisition unit, a data cleaning unit, an obstacle recognition unit, a distance monitoring unit, and a speed adjustment unit;

[0034] The initialization unit is configured to perform initialization calibration on the sensor;

[0035] The data acquisition unit is configured to acquire environmental data;

[0036] The data cleaning unit is configured to clean noise data;

[0037] The obstacle recognition unit is configured to identify fixed obstacles and moving objects based on the cleaned environmental data using the YOLO algorithm;

[0038] The distance monitoring unit is configured to continuously calculate the distances to fixed obstacles and moving objects;

[0039] The speed adjustment unit is configured to decelerate when the distance to a moving object is less than a preset value; and stop when the distance to a fixed obstacle reaches a preset value.

[0040] Among them, the obstacle recognition unit includes an environmental image acquisition subunit, a marking subunit, a training subunit, and an identification subunit; <{

[0041] The environmental image acquisition subunit is configured to acquire environmental images containing different types of obstacles;

[0042] The marking subunit is configured to label the obstacles in the environmental images as fixed obstacles or moving objects;

[0043] The training subunit is configured to train the YOLO model using the marked environmental images to obtain an obstacle recognition model;

[0044] The recognition subunit is configured to recognize a target obstacle based on the obstacle recognition model and the currently captured environmental image.

[0045] In a third aspect, the present invention further provides an anti-collision method, including: setting a moving path according to shooting requirements;

[0046] generating a driving instruction based on the moving path and driving the corresponding omnidirectional wheels to move;

[0047] performing shooting during the movement according to shooting requirements;

[0048] acquiring environmental data during the movement and performing anti-collision detection;

[0049] calculating the size of the obstacle after detecting the obstacle, and if the size of the obstacle is within a preset range, starting the pushing mechanism to push.

[0050] An omnidirectional wheel walking device, an anti-collision system and method of the present invention, the base is the basic structure of the entire device, carrying all internal mechanical and electronic components, and ensuring the stability and firmness of the overall structure. The support platform is fixedly connected to the top of the base, providing a stable platform for installing other functional modules, such as the shooting module, etc. At the same time, it can also enhance the overall rigidity of the device, ensuring that there will be no unnecessary shaking or tilting during the movement, which will affect the shooting quality. The moving path setting module is one of the cores of the device's intelligence. Through preset or real-time input shooting requirements, this module can plan the optimal moving route to ensure that the device can reach the specified position in the shortest time and avoid obstacles and other potential risks. This process involves the support of advanced algorithms, such as path planning algorithms, obstacle avoidance algorithms, etc., to achieve safe and efficient movement. The moving module is responsible for converting the path planned by the moving path setting module into specific actions. It generates precise driving instructions based on the set path, and then controls the actions of the omnidirectional wheels. The unique design of the omnidirectional wheels allows the device to achieve translation in any direction without the need to rotate, greatly improving the flexibility and mobility of the device. In addition, the moving module also includes a feedback control system to monitor and adjust the moving speed and direction to ensure accurate movement along the predetermined path. The shooting module is a key part of performing image capture tasks. According to the previously determined shooting requirements, it can perform high-quality photo or video recording during the movement of the device. To adapt to different shooting scenarios, this module is equipped with lenses with different focal lengths, an autofocus system, optical image stabilization technology, and other professional-level photographic equipment to ensure clear and stable images even under dynamic conditions.

[0051] In summary, this omnidirectional wheel walking device integrates advanced mechanical design, intelligent control technology, and high-quality image acquisition capabilities, aiming to provide users with an efficient, flexible, and reliable mobile shooting solution. Brief Description of the Drawings

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0053] Figure 1 It is a structural diagram of an omnidirectional wheel walking device of the present invention.

[0054] Figure 2 It is a diagram of the mobile path setting module of the present invention.

[0055] Figure 3 It is a diagram of the mobile module of the present invention.

[0056] Figure 4 It is a diagram of the compensation unit of the present invention.

[0057] Figure 5 It is a diagram of the shooting module of the present invention.

[0058] Figure 6 It is a diagram of an anti-collision system of the present invention.

[0059] Figure 7 It is a diagram of the anti-collision detection module of the present invention.

[0060] Figure 8 It is a diagram of the obstacle recognition unit of the present invention.

[0061] Figure 9 It is a flowchart of an anti-collision method of the present invention.

[0062] Base 101, support platform 102, movement path setting module 103, movement module 104, shooting module 105, initial path setting unit 106, control point setting unit 107, optimization and adjustment unit 108, path import unit 109, path decomposition unit 110, drive information calculation unit 111, drive unit 112, compensation unit 113, data acquisition sub-unit 114, abnormal data removal sub-unit 115, data classification sub-unit 116, judgment model training sub-unit 117, error calculation sub-unit 118, adjustment sub-unit 119, camera unit 120, camera path planning unit 121, monitoring unit 122, anti-collision detection module 123, obstacle pushing module 124, initialization unit 125, data acquisition unit 126, data cleaning unit 127, obstacle recognition unit 128, distance monitoring unit 129, speed adjustment unit 130, environmental image acquisition sub-unit 131, marking sub-unit 132, training sub-unit 133, recognition sub-unit 134. Detailed implementation mode

[0063] The embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present invention and should not be construed as a limitation of the present invention.

[0064] First embodiment

[0065] Please refer to Figures 1 to 5 , the present invention provides an omnidirectional wheel walking device, including a base 101 and a support platform 102. The support platform 102 is fixedly connected to the base 101 and is located on the top of the base 101. It also includes a movement path setting module 103, a movement module 104, and a shooting module 105. The movement path setting module 103 is used to set the movement path according to the shooting requirements. The movement module 104 is used to generate a driving instruction based on the movement path and drive the corresponding omnidirectional wheels to move. The shooting module 105 is used to take pictures during the movement according to the shooting requirements.

[0066] In this embodiment, the base 101 serves as the basic structure of the entire device, carrying all internal mechanical and electronic components and ensuring the stability and firmness of the overall structure. The support platform 102 is fixedly connected to the top of the base 101, providing a stable platform for installing other functional modules, such as the shooting module 105, etc. At the same time, it can also enhance the overall rigidity of the device and ensure that there will be no unnecessary shaking or tilting during the movement.

[0067] The movement path setting module 103 can plan a movement route according to preset or real-time input shooting requirements, so as to ensure that the device can reach the specified position in the shortest time and avoid obstacles and other potential risks.

[0068] The movement module 104 then converts the path planned by the movement path setting module 103 into specific actions. It generates precise drive instructions based on the set path, and then controls the actions of the omnidirectional wheels. The unique design of the omnidirectional wheels allows the device to translate in any direction without the need to rotate, greatly enhancing the flexibility and mobility of the device. In addition, the movement module 104 also includes a feedback control system to monitor and adjust the movement speed and direction to ensure accurate movement along the predetermined path.

[0069] The shooting module 105 is a key part for performing image capture tasks. According to the pre-determined shooting requirements, it can take high-quality photos or record videos during the movement of the device. To adapt to different shooting scenarios, this module is equipped with lenses of different focal lengths, an autofocus system, optical image stabilization technology, and other professional-level photographic equipment to ensure clear and stable images even under dynamic conditions.

[0070] In summary, this omnidirectional wheel walking device can automatically set the walking route according to requirements, aiming to provide users with an efficient, flexible and reliable mobile shooting solution, thus saving costs compared to the rail shooting method.

[0071] The movement path setting module 103 includes an initial path setting unit 106, a control point setting unit 107, and an optimization and adjustment unit 108; the initial path setting unit 106 is used to set an initial movement path based on the shooting scene requirements and shooting terrain; the control point setting unit 107 is used to set control points on the initial movement path; the optimization and adjustment unit 108 is used to optimize the initial movement path based on the parameters of the photographing unit to obtain a shooting path.

[0072] The initial path setting unit 106 plans a preliminary movement path based on the requirements of the shooting scene and the on-site terrain conditions. At this stage, the system analyzes the content of the picture to be captured and the specific situation of the environment, such as the complexity of the terrain and the distribution of obstacles, so as to determine a basic route. This process involves image recognition technology to evaluate the elements in the shooting area, and also combines geographic information system (GIS) data to understand more extensive terrain information. Through these means, the initial path setting unit 106 can provide a reasonable and feasible starting point for the subsequent steps.

[0073] The control point setting unit 107 will set several key positions on this path as control points. These control points are not only used to mark important stopping or turning points, but also can be used as calibration references to ensure that the device travels along the expected trajectory. The selection of control points usually takes into account factors such as changes in the shooting perspective, lighting conditions, and the position of the target object, to ensure that each control point has a positive impact on the final film quality. In addition, they also play a monitoring role, enabling the system to make fine-tuning when necessary to keep the device moving steadily along the predetermined route.

[0074] Finally, the optimization and adjustment unit 108 will conduct a detailed optimization of the initial movement path based on the parameters provided by the photographing unit to generate the final shooting path. In this step, the system will comprehensively consider photographic parameters such as the camera focal length, shutter speed, and aperture size, and will also take into account the physical limitations of the device itself, such as the maximum speed and turning radius. Through a comprehensive analysis of the above factors, the optimization and adjustment unit 108 can make fine adjustments to the path, such as adjusting the traveling speed to match the exposure time, or changing the path angle to obtain the best shooting angle. The purpose of doing this is to ensure that each shooting moment can achieve the best effect, whether it is a static photo or a dynamic video, both of which can meet professional-level requirements.

[0075] The moving module 104 includes a path import unit 109, a path decomposition unit 110, a driving information calculation unit 111, a driving unit 112, and a compensation unit 113; the path import unit 109 is used to import the data of the moving path; the path decomposition unit 110 is used to decompose the moving path into a series of discrete trajectory points, and the trajectory points include position coordinates, direction angles, and speed settings; the driving information calculation unit 111 is used to calculate the required speed and turning angle according to the information of each trajectory point, and then convert them into control signals for the motor; the driving unit 112 is used to control the speed and position of the omnidirectional wheels based on the control signals; the compensation unit 113 is used to compensate for the control error of the control signals based on historical data.

[0076] The path import unit 109 is responsible for importing the moving path data planned by the moving path setting module 103 into the moving module 104. This data usually includes a series of coordinate points from the starting point to the ending point and other relevant information, such as speed limits or shooting requirements at specific positions. The path import unit 109 not only needs to ensure the accuracy of data transmission, but also needs to ensure its compatibility with the internal system of the device, so that subsequent units can correctly parse and utilize this information.

[0077] After receiving the path data, the task of the path decomposition unit 110 is to convert the continuous movement path into a series of discrete trajectory points. Each trajectory point contains specific position coordinates (X, Y), direction angle (θ), and speed setting (V). This step is crucial for simplifying complex path planning because it allows the system to process movement instructions in a more fine-grained manner, thereby improving execution accuracy. In addition, the path decomposition unit 110 also considers terrain features or obstacles and makes necessary corrections to the path to ensure safety and efficiency.

[0078] Based on the trajectory point information provided by the path decomposition unit 110, the drive information calculation unit 111 calculates the accurate speed and steering angle required to reach each point. This is a process involving complex algorithms that comprehensively consider factors such as acceleration, deceleration, and maximum turning radius to generate the most suitable motion parameters. These parameters are then converted into specific motor control signals to guide the actions of the omnidirectional wheels. This process requires high-precision time synchronization and real-time computing capabilities to ensure that the device can respond smoothly and quickly to path changes.

[0079] The drive unit 112 receives the control signals from the drive information calculation unit 111 and directly controls the speed and position of the omnidirectional wheels. It includes various types of motor controllers, such as servo motor or stepper motor controllers, each with different characteristics and application scenarios. The drive unit 112 is also responsible for monitoring the actual motion state of the omnidirectional wheels, for example, detecting the actual position and speed through encoder feedback, to ensure that the device moves strictly along the predetermined path. At the same time, it also has a certain self-regulating ability and can quickly respond when encountering external disturbances to maintain the coherence and stability of the movement.

[0080] The compensation unit 113 exists to address inevitable control errors. Even under ideal conditions, due to mechanical wear, sensor noise, or other uncertain factors, the actual movement results will deviate from the expected values. Based on historical data and machine learning algorithms, the compensation unit 113 can identify and quantify these error patterns and then automatically adjust future control signals to minimize the cumulative error. This adaptive mechanism not only improves the long-term operation reliability of the device but also enhances its ability to handle emergencies, such as suddenly appearing obstacles or environmental changes.

[0081] The compensation unit 113 includes a data acquisition subunit 114, an abnormal data removal subunit 115, a data classification subunit 116, a judgment model training subunit 117, an error calculation subunit 118, and an adjustment subunit 119; the data acquisition subunit 114 is configured to acquire historical movement data when driving the omnidirectional wheels, and the movement data includes motor PWM signals, positions and speeds feedback by encoders; the abnormal data removal subunit 115 is configured to remove outliers and missing values in the movement data; the data classification subunit 116 is configured to divide the processed movement data into a training set and a validation set; the judgment model training subunit 117 is configured to train a long short-term memory network using the training set to obtain a judgment model; the error calculation subunit 118 is configured to calculate an adjustment error based on the judgment model and the currently acquired real-time movement data; the adjustment subunit 119 is configured to adjust the control signal in advance based on the adjustment error.

[0082] The task of the data acquisition subunit 114 is to collect relevant movement data from the historical records of driving the omnidirectional wheels. This data mainly includes motor PWM (pulse width modulation) signals, position information and speed information feedback by encoders. By continuously monitoring and recording these parameters, the system can establish a detailed historical database, providing a solid foundation for subsequent analysis and optimization. In addition, this subunit also has the ability to collect real-time data, ensuring that it can respond immediately to the latest changes in the movement state.

[0083] The abnormal data removal subunit 115 is responsible for cleaning outliers and missing values in the historical movement data. Outliers are incorrect readings caused by sensor failures, external interferences or other unforeseen factors; missing values refer to the situation where some data points fail to be recorded successfully. This process is crucial for ensuring the effectiveness and accuracy of data analysis. The subunit will use statistical methods, machine learning algorithms or rule settings to identify and remove these inaccurate data points, thereby improving the quality of subsequent processing.

[0084] The data classification subunit 116 divides the cleaned movement data into two parts: a training set and a validation set. The training set is used to build and optimize the prediction model, while the validation set is used to evaluate the model performance and make necessary adjustments. A reasonable data splitting ratio helps to balance the generalization ability of the model and the risk of overfitting, ensuring its reliability and effectiveness in practical applications. This step also involves preprocessing techniques such as feature selection and dimensionality reduction to simplify the data structure and highlight key attributes.

[0085] The judgment model training subunit 117 uses advanced deep learning algorithms such as long short-term memory network (LSTM) to train the training set, aiming to create a judgment model that can accurately predict the control error. LSTM is particularly suitable for processing time series data because it can effectively capture the long-term dependencies between inputs and outputs. Through repeated iteration and optimization, the finally obtained judgment model should have strong adaptability and prediction accuracy, and be able to give reliable error estimates under different conditions.

[0086] The error calculation subunit 118 calculates the deviation between the current control signal and the ideal path, that is, the adjustment error, based on the trained judgment model and the currently real-time obtained movement data. This process involves inputting the real-time data into the judgment model, predicting the future error trend through the model, and calculating the specific value that needs to be corrected accordingly. In order to achieve accurate error calculation, the subunit will combine various algorithms and technologies, such as Kalman filtering, particle filtering, etc., to improve the accuracy and stability of the prediction.

[0087] The function of the adjustment subunit 119 is to correct the control signal in advance based on the adjustment error provided by the error calculation subunit 118. This step aims to pre-compensate for the emerging errors and make the movement of the omnidirectional wheels closer to the expected path. The adjustment strategies include but are not limited to fine-tuning the PWM signal, adjusting the steering angle, or changing the traveling speed. In addition, the adjustment subunit 119 also needs to consider the dynamic characteristics of the system, such as inertia effects and response delays, to ensure that the adjustment measures can take effect quickly without causing unnecessary jitter or oscillation.

[0088] The shooting module 105 includes a camera unit 120, a camera path planning unit 121, and a monitoring unit 122; the camera unit 120 is used to be fixed on the mobile platform; the camera path planning unit 121 is used to set the shooting path of the camera unit 120 based on the movement path, and the shooting path includes a starting point, an ending point, speed change points, and key frames; the monitoring unit 122 is used to track the shooting process in real time based on the monitor.

[0089] The imaging unit 120 is the core component that actually performs the shooting task and is usually fixedly installed on the support platform 102 of the mobile platform. It is equipped with a high-performance camera that can adapt to various shooting requirements, including but not limited to static photos, video recording, etc. According to specific shooting requirements, the imaging unit 120 is equipped with different types of lenses (such as wide-angle, telephoto, etc.) to capture scenes from different perspectives or within different ranges. In addition, to ensure shooting quality, the imaging unit 120 also integrates advanced features such as an autofocus system, optical image stabilization technology, and high dynamic range (HDR) function to cope with complex and changing lighting conditions and motion states. Its design not only takes into account image quality and stability but also pays attention to compatibility and integration with the mobile platform, ensuring clear and stable images can be obtained even under high-speed or unstable mobile conditions.

[0090] The task of the imaging path planning unit 121 is to formulate a detailed shooting path for the imaging unit 120 based on the given mobile path. This path is not just a simple line from the starting point to the ending point but a complex trajectory containing multiple key elements. Specifically, the shooting path should include the following important components:

[0091] The starting point is the position where shooting begins, determining the starting point of the shooting process. The ending point is the position where shooting ends, marking the completion of the entire shooting sequence. The speed change points set the positions where speed adjustments occur during the shooting process, such as the transition points for accelerating, decelerating, or maintaining a constant speed. Reasonable speed control is crucial for maintaining shooting stability and achieving specific visual effects.

[0092] Key frames refer to those shooting positions or moments that require special attention. They are important scene transition points, special composition angles, or places where a stop is needed for detailed shooting. By predefined these key frames, it can be ensured that no important shooting opportunities are missed, and it helps to organize the materials more easily during post-editing.

[0093] The imaging path planning unit 121 uses advanced algorithms and technologies to optimize this path, taking into account the capacity limitations of the imaging unit 120 (such as the minimum focusing distance), environmental factors (such as light intensity), and the expected visual effects (such as smooth transitions or fast cuts). It also combines machine learning models to predict the best shooting parameters, such as exposure settings, white balance, etc., to further improve shooting quality.

[0094] The function of the monitoring unit 122 is to track and manage the entire shooting process in real time. It usually includes one or more monitors, through which the operator can visually observe the images captured by the camera unit 120, so as to make necessary adjustments in a timely manner. The monitoring unit 122 can not only provide immediate visual feedback, but also record shooting data for subsequent analysis and improvement. In addition, it also supports the remote control function, allowing users to wirelessly control the parameter settings of the camera unit 120 within a certain range, such as zooming and focusing. More importantly, the monitoring unit 122 has an intelligent alarm mechanism. When abnormal situations (such as exceeding the predetermined shooting area, equipment failure, etc.) are detected, it can immediately notify the operator to take measures to ensure the safety and smooth progress of the shooting process.

[0095] Second Embodiment

[0096] Please refer to Figures 6 to 8 , the present invention also provides an anti-collision system, including an omnidirectional wheel walking device, an anti-collision detection module 123, and an obstacle pushing module 124; the anti-collision detection module 123 is used to obtain environmental data during movement and perform anti-collision detection; the obstacle pushing module 124 is used to calculate the size of the obstacle after detecting the obstacle, and if the size of the obstacle is within the preset range, the pushing mechanism is activated to push the obstacle.

[0097] The omnidirectional wheel walking device not only has highly flexible movement capabilities, but also can carry a variety of functional modules. Its unique omnidirectional wheel design allows the device to translate in any direction without the need to rotate, greatly enhancing the mobility. In addition, the device is equipped with a powerful power system and an accurate control system, and can operate stably on various terrains to meet the requirements of different application scenarios.

[0098] The anti-collision detection module 123 is one of the core components of the anti-collision system, responsible for obtaining environmental data in real time during the movement of the device and performing anti-collision detection. This module integrates a variety of sensor technologies, such as lidar (LiDAR), ultrasonic sensors, infrared sensors, etc., and can comprehensively perceive the surrounding environment, including the positions, shapes, and movement trends of static obstacles and dynamic objects. By fusing multi-source sensing information, the anti-collision detection module 123 can build a detailed three-dimensional environmental model to provide a reliable basis for subsequent decision-making.

[0099] Using a high-precision sensor array, continuously scan and collect environmental information around the device. This data contains various parameters such as distance, speed, and angle, which are used to depict a complete environmental profile. Based on the acquired data, the anti-collision detection module 123 uses advanced algorithms for real-time analysis to identify potential collision risks. This includes, but is not limited to, path prediction, collision probability calculation, and emergency obstacle avoidance strategy generation. Once an impending collision is detected, the module will immediately trigger corresponding protection mechanisms, such as deceleration, stopping, or steering.

[0100] When the anti-collision detection module 123 detects an obstacle, the obstacle pushing module 124 will intervene. This module is designed to deal with small obstacles that can be safely removed through physical contact, rather than simply bypassing or stopping. The specific workflow is as follows:

[0101] After detecting the obstacle, the system first precisely measures the size of the obstacle. This is usually done by combining sensor data and image processing techniques to ensure an accurate understanding of the obstacle size. According to pre-set safety standards, it is judged whether the obstacle falls within the size range that can be pushed. This range takes into account factors such as the device's own structural strength and power output to ensure the safety of the pushing process. If the obstacle size meets the preset conditions, the obstacle pushing module 124 will activate the built-in pushing mechanism. The pushing mechanism is a robotic arm, a push rod, or other forms of power devices that can apply appropriate force to move the obstacle out of the predetermined path. At the same time, the system will also monitor the feedback information during the pushing process, such as resistance changes, to ensure that the operation is completed smoothly without causing damage.

[0102] The anti-collision detection module 123 includes an initialization unit 125, a data acquisition unit 126, a data cleaning unit 127, an obstacle recognition unit 128, a distance monitoring unit 129, and a speed adjustment unit 130; the initialization unit 125 is used to initialize and calibrate the sensors; the data acquisition unit 126 is used to acquire environmental data; the data cleaning unit 127 is used to clean noise data; the obstacle recognition unit 128 is used to identify fixed obstacles and moving objects based on the cleaned environmental data using the YOLO algorithm; the distance monitoring unit 129 is used to continuously calculate the distance to fixed obstacles and moving objects; the speed adjustment unit 130 is used to decelerate when the distance to a moving object is less than a preset value; and stop when the distance to a fixed obstacle reaches the preset value. [[ID=Eleven]]

[0103] The task of the initialization unit 125 is to calibrate and initialize the settings of the sensors during each startup or restart. This step is crucial because it ensures that all sensors can work in their optimal state, thus providing accurate and reliable environmental perception capabilities. The initialization process includes:

[0104] Precisely calibrate LiDAR, ultrasonic sensors, infrared sensors, etc. through built-in algorithms or external tools to eliminate deviations caused by installation errors or long-term use. Set parameters such as the working mode and sensitivity threshold of the sensors according to the requirements of the application scenario to adapt to different environmental conditions. Run a series of self-check programs to check whether the sensors are working properly, promptly detect and report any potential problems, and ensure the reliability of the system.

[0105] The data acquisition unit 126 collects data on the surrounding environment in real time. It uses various types of sensors to obtain rich information such as distance, speed, temperature, humidity, etc., and constructs a comprehensive environmental model.

[0106] Specifically:

[0107] Integrate data from different sensors to form a more complete and accurate environmental description. For example, combining the distance data from LiDAR with the image data from a camera can more precisely depict the position and shape of obstacles. To capture fast-changing dynamic scenes, the data acquisition unit 126 uses a high sampling rate to ensure that important information is not missed. For large or complex devices, the data acquisition unit 126 is distributed at multiple locations, each responsible for collecting data in a specific area, and then aggregating it to the central processing unit for unified analysis.

[0108] The data cleaning unit 127 aims to remove noise data and other outliers generated during the acquisition process to improve the accuracy of subsequent processing. This process involves applying various digital filters (such as low-pass filters, high-pass filters) to smooth the signal and reduce the impact of random fluctuations.

[0109] Based on the cleaned environmental data, the obstacle recognition unit 128 uses the advanced YOLO (You Only Look Once) algorithm to identify stationary obstacles and moving objects. The YOLO algorithm is famous for its efficient real-time object detection ability and can quickly and accurately mark each object in the environment. The specific workflow includes: extracting key features from the cleaned image or point cloud data for subsequent classification. Using a pre-trained deep neural network model to distinguish between stationary obstacles (such as walls, furniture) and moving objects (such as pedestrians, other robots), and giving a confidence score. Drawing a bounding box for each recognized object for easy tracking and management.

[0110] The distance monitoring unit 129 continuously calculates the distances between the device and the identified stationary obstacles and moving objects. This process relies on real-time updated position information and relative motion states to ensure prompt response to any approaching trends. Specific functions include continuously refreshing the distance values to obstacles to ensure the timeliness of data. When a certain obstacle is detected to enter the set safe distance range, an alarm signal is immediately issued. Provide the latest distance information to the path planning module to help it optimize the moving path and avoid unnecessary detours or stops.

[0111] The speed adjustment unit 130 adjusts the speed of the device when necessary according to the information provided by the distance monitoring unit 129 to ensure safe driving. Specific measures include: when the distance to a moving object is less than a preset threshold, gradually reduce the speed until reaching the safe driving speed or coming to a complete stop. This progressive deceleration method can effectively avoid the impact caused by sudden braking. When the distance to a stationary obstacle approaches the preset minimum safe distance, immediately stop the device to prevent collisions. At this time, the system will evaluate the nature of the obstacle and the surrounding environment to decide the next action (such as waiting, detouring or requesting manual intervention). Once the obstacle moves out of the safe range or the environment returns to normal, the speed adjustment unit 130 will automatically resume the normal driving speed and continue to perform the task.

[0112] The obstacle recognition unit 128 includes an environmental image acquisition subunit 131, a marking subunit 132, a training subunit 133 and an identification subunit 134; the environmental image acquisition subunit 131 is used to collect environmental images containing different types of obstacles; the marking subunit 132 is used to label the obstacles in the environmental images as stationary obstacles or moving objects; the training subunit 133 is used to train the YOLO model with the labeled environmental images to obtain an obstacle recognition model; the identification subunit 134 is used to identify the target obstacle based on the currently collected environmental image and the obstacle recognition model.

[0113] The task of the environmental image acquisition subunit 131 is to collect images containing different types of obstacles from the device's surrounding environment. To provide a comprehensive and diverse dataset, this subunit is usually equipped with a high-resolution camera or a 3D sensor (such as a depth camera) to capture environmental information from different perspectives. Specific functions include: by installing multiple cameras or a rotating pan-tilt head, ensuring that all directions around the device can be covered to obtain omnidirectional environmental images. In addition to visible light images, it also supports imaging in other spectral ranges such as infrared and ultraviolet to work effectively under low light or other special conditions. Transmit the collected images to the subsequent processing unit in real time to ensure the freshness and timeliness of the data.

[0114] The marking subunit 132 makes detailed annotations on the obstacles in the environmental image. This process is crucial for training a high-quality YOLO model because it provides the "correct answers" required for supervised learning. The specific work content is as follows:

[0115] Professional personnel or trained operators manually mark each obstacle in the image and clearly distinguish between fixed obstacles (such as walls, trees) and moving objects (such as pedestrians, vehicles). This step requires precision and consistency to avoid introducing biases. Use computer vision techniques to assist the annotation process, such as automatically outlining the object contours through edge detection algorithms and then having them confirmed by humans. This method can establish a detailed label library that defines the characteristics and classification criteria of various obstacles, facilitating unified management and subsequent queries.

[0116] The training subunit 133 uses the annotated environmental images to train the YOLO (You Only Look Once) model, and finally obtains an obstacle recognition model that can accurately identify obstacles. The training process involves the following key steps:

[0117] Integrate all the annotated images to form a structured and balanced dataset. Considering the diversity of actual application scenarios, the dataset should cover various types of obstacles and different environmental conditions as much as possible. Select an appropriate YOLO version (such as YOLOv3, YOLOv4, etc.) according to specific requirements and adjust its hyperparameters (such as the number of network layers, activation functions, etc.) to optimize performance. Through multiple iterative trainings, continuously update the model weights to make it gradually adapt to the specific task. Methods such as cross-validation and early stopping strategies are adopted during the training process to prevent overfitting and ensure that the model has good generalization ability. Regularly use the validation set to evaluate the performance of the model, monitor key metrics such as Precision, Recall, F1-score, etc., and promptly discover and solve potential problems.

[0118] Based on the currently acquired environmental image and the trained obstacle recognition model, the recognition subunit 134 performs real-time object recognition tasks. It can quickly and accurately mark the positions and types of fixed obstacles and moving objects in a complex environment. The specific operations include: inputting the latest environmental image into the trained YOLO model, using GPU acceleration for computing to achieve an inference speed in milliseconds to meet the requirements of real-time applications. The model outputs the position coordinates (bounding boxes) and their categories (fixed obstacles or moving objects) of each detected obstacle, along with a confidence score to help subsequent modules make decisions. Continuously update the detection results according to the movement state of the device and environmental changes to ensure that the latest environmental information is always grasped.

[0119] The third embodiment

[0120] Please refer toFigure 9 The present invention also provides an anti-collision method, comprising:

[0121] S301 sets a moving path according to shooting requirements;

[0122] During this initial phase, the system plans the most appropriate movement path based on the specific shooting task requirements. This involves not only selecting the starting and ending points, but also setting keyframes along the way (such as specific viewing angles or composition points). The path planning module takes into account the following factors:

[0123] Shooting Objective: Determine the specific scene or object you want to capture.

[0124] Environmental characteristics: Analyze terrain, lighting conditions, etc. to select the best route.

[0125] Time limit: Arrange a reasonable movement speed and stay time according to the shooting plan.

[0126] S302 generates a driving instruction based on the moving path and drives the corresponding omnidirectional wheel to move;

[0127] Once the path planning is complete, the system generates detailed drive instructions based on this path. These instructions are converted into control signals that act directly on the omnidirectional wheel's power system to achieve precise motion control. The specific process is as follows:

[0128] The continuous movement path is discretized into a series of trajectory points, each containing position coordinates, a heading angle, and a speed setting. Based on the trajectory point information, the required speed and steering angle are calculated and converted into motor control signals. Upon receiving these control signals, the drive unit 112 controls the speed and position of the omnidirectional wheels, ensuring that the device strictly follows the predetermined path.

[0129] S303: Shooting during the movement according to shooting requirements;

[0130] As the device moves along the predetermined path, the camera module 105 begins operating to ensure that high-quality images or videos are captured as required. Key to this stage is the dynamic adjustment of camera unit 120 parameters (such as focal length and exposure) based on the device's actual position and posture to achieve optimal capture. Capture is automatically triggered at preset keyframes to ensure that important moments are not missed. The captured data is synchronized with the timestamp of the movement path to facilitate later editing and analysis.

[0131] S304 acquires environmental data and performs anti-collision detection during movement;

[0132] To ensure the safety of movement, the anti-collision detection module 123 continuously monitors the surrounding environment to promptly detect potential collision risks. The specific operations include integrating data from various sensors such as lidar, ultrasonic sensors, cameras, etc., to construct a comprehensive environmental model. Advanced algorithms (such as deep learning) are used to perform real-time analysis on the environmental data to identify fixed obstacles and moving objects. Continuously calculate the distance to the obstacles, and once approaching the preset safety threshold, immediately take corresponding measures.

[0133] S305 calculates the size of the obstacle after detecting it. If the size of the obstacle is within the preset range, it activates the pushing mechanism to push.

[0134] When the anti-collision detection module 123 detects an obstacle, the obstacle handling mechanism will quickly intervene, evaluate and handle the obstacle to ensure that the device can continue to move forward safely. The specific process is as follows:

[0135] Using sensor data and image processing technology, accurately measure the size of the obstacle and determine whether it is within the range that can be pushed. According to the preset criteria, decide whether the obstacle can be safely pushed. This criterion takes into account factors such as the structural strength of the device itself and the power output. If the size of the obstacle meets the preset conditions, the system will activate the built-in pushing mechanism to apply an appropriate force to move the obstacle away from the predetermined path. At the same time, the system will also monitor the feedback information during the pushing process, such as the change in resistance, to ensure that the operation is completed smoothly without causing damage.

[0136] The above-disclosed is only a preferred embodiment of the present invention. Of course, it cannot be used to limit the scope of the rights of the present invention. Those of ordinary skill in the art can understand all or part of the processes of implementing the above embodiments, and the equivalent changes made according to the claims of the present invention still fall within the scope covered by the invention.

Claims

1. An omnidirectional wheel walking device, comprising a base and a support platform, the support platform and the base are fixedly connected and located at the top of the base, and is characterized in that, further comprising a moving path setting module, a moving module and a shooting module; The moving path setting module is used to set a moving path according to shooting requirements; The moving module is used to generate a driving instruction based on the moving path and drive the corresponding omnidirectional wheel to move; The shooting module is used to perform shooting during movement according to shooting requirements.

2. The omnidirectional wheel walking device according to claim 1, characterized in that, The moving path setting module includes an initial path setting unit, a control point setting unit and an optimization and adjustment unit; The initial path setting unit is used to set an initial movement path based on the shooting picture requirements and the shooting terrain; The control point setting unit is used to set control points on the initial movement path; The optimization and adjustment unit is used to optimize the initial movement path based on the parameters of the photographing unit to obtain a shooting path.

3. The omnidirectional wheel walking device according to claim 2, characterized in that, The moving module includes a path import unit, a path decomposition unit, a driving information calculation unit, a driving unit and a compensation unit; The path import unit is used to import the data of the moving path; The path decomposition unit is used to decompose the moving path into a series of discrete trajectory points, and the trajectory points include position coordinates, direction angles and speed settings; The driving information calculation unit is used to calculate the required speed and steering angle according to the information of each trajectory point, and then convert it into a control signal of the motor; The driving unit is used to control the speed and position of the omnidirectional wheel based on the control signal; The compensation unit is used to compensate the control error of the control signal based on historical data.

4. The omnidirectional wheel walking device according to claim 3, characterized in that, The compensation unit includes a data acquisition sub-unit, an abnormal data removal sub-unit, a data classification sub-unit, a judgment model training sub-unit, an error calculation sub-unit and an adjustment sub-unit; The data acquisition sub-unit is used to acquire historical movement data when driving the omnidirectional wheel to move, and the movement data includes the motor PWM signal, the position and speed fed back by the encoder; The abnormal data removal sub-unit is used to remove the abnormal values and missing values in the movement data; The data classification sub-unit is used to divide the processed movement data into a training set and a verification set; The judgment model training sub-unit is used to train a long short-term memory network with the training set to obtain a judgment model; The error calculation sub-unit is used to calculate an adjustment error based on the judgment model and the currently acquired real-time movement data; The adjustment sub-unit is used to adjust the control signal in advance based on the adjustment error.

5. The omnidirectional wheel walking device, anti-collision system and method according to claim 4, characterized in that, The shooting module includes a camera unit, a camera path planning unit and a monitoring unit; The camera unit is used to be fixed on the moving platform; The camera path planning unit is configured to set the shooting path of the camera unit based on the moving path, and the shooting path includes a starting point, an ending point, a speed change point, and key frames; The monitoring unit is configured to track the shooting process in real time based on the monitor.

6. An anti-collision system, comprising an omnidirectional wheel walking device according to any one of claims 1 to 5, characterized in that it further comprises an anti-collision detection module and an obstacle pushing module; The anti-collision detection module is configured to obtain environmental data during movement and perform anti-collision detection; The obstacle pushing module is configured to calculate the size of the obstacle after detecting the obstacle, and if the size of the obstacle is within a preset range, start the pushing mechanism to push.

7. The anti-collision system according to claim 6, characterized in that the anti-collision detection module includes an initialization unit, a data acquisition unit, a data cleaning unit, an obstacle recognition unit, a distance monitoring unit, and a speed adjustment unit; The initialization unit is configured to perform initialization calibration on the sensor; The data acquisition unit is configured to acquire environmental data; The data cleaning unit is configured to clean the noise data; The obstacle recognition unit is configured to identify fixed obstacles and moving objects based on the cleaned environmental data using the YOLO algorithm; The distance monitoring unit is configured to continuously calculate the distances to the fixed obstacles and moving objects; The speed adjustment unit is configured to decelerate when the distance to the moving object is less than a preset value; and stop when the distance to the fixed obstacle reaches a preset value.

8. The anti-collision system according to claim 7, characterized in that the obstacle recognition unit includes an environmental image acquisition sub-unit, a marking sub-unit, a training sub-unit, and an identification sub-unit; The environmental image acquisition sub-unit is configured to acquire environmental images containing different types of obstacles; The marking sub-unit is configured to label the obstacles in the environmental image as fixed obstacles or moving objects; The training sub-unit is configured to train the YOLO model using the labeled environmental images to obtain an obstacle recognition model; The identification sub-unit is configured to identify target obstacles based on the currently acquired environmental image and the obstacle recognition model.

9. A collision prevention method, applied to a collision prevention system according to claim 8, Characterized in that it includes: setting a moving path according to the shooting requirements; generating a driving instruction based on the moving path and driving the corresponding omnidirectional wheel to move; performing shooting during movement according to the shooting requirements; acquiring environmental data during movement and performing anti-collision detection; calculating the size of the obstacle after detecting the obstacle, and if the size of the obstacle is within a preset range, start the pushing mechanism to push.

Images