A method, system, terminal and robot for escaping from trouble for a tracked robot

Abstract

The application relates to the field of robot control, and specifically provides a caterpillar robot escape method, a system, a terminal and a robot, the method comprising the following steps: establishing a coordinate system with the robot center as the origin, and dividing the surrounding space into four direction channels of front, back, left and right; sensing the environment by using a holder sensor, updating the passable state of each channel according to the position and distance of the obstacle; monitoring the path planning state and the motion state of the robot in real time, and judging whether the escape trigger condition is met; when the trigger condition is met, the latest channel state is called according to the type, and the corresponding escape strategy is executed. The application realizes the control from sensing to classifying and executing the escape by maintaining the passable state of the four direction channels in real time and based on the passable state and the robot state, and improves the directionality, safety and disposal efficiency of the escape decision for different difficulties.

Description

Technical Field

[0001] This invention relates to the field of robot control, and specifically to a method, system, terminal, and robot for a tracked robot to escape from obstacles. Background Technology

[0002] When tracked mobile robots perform autonomous tasks such as inspection and transportation, they face the challenge of escaping when they become trapped and unable to continue along their original path due to environmental changes or abnormalities in their own condition. In related technologies, the robot's escape behavior largely relies on pre-set, simple reaction logic. For example, after detecting obstruction, it might rotate randomly in place and try again; or after detecting motor overload, it might execute a fixed backward movement. These methods lack assessment of the real-time spatial structure of the robot's surroundings, resulting in blind escape actions. This can easily lead to the robot becoming increasingly entangled in complex obstacles or causing mechanical damage. Furthermore, the triggers lack differentiation of the root cause of the predicament; whether the path is blocked, the ground is slippery, or the mechanism is jammed, the same reaction is triggered, leading to low processing efficiency. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides a method, system, terminal, and robot for escaping obstacles in a tracked robot. By maintaining the passability status of the four-way passage in real time and based on the passability status and the robot's status, it achieves control from perception to classification and execution of escape, thereby improving the directionality, safety, and efficiency of escape decision-making and handling different difficulties.

[0004] In a first aspect, the technical solution of the present invention provides a method for a tracked robot to escape from a difficult situation, comprising the following steps: Establish a robot coordinate system with the robot's center as the origin, and divide the space around the robot into four directional channels: front, back, left, and right. The robot uses its onboard gimbal sensors to perceive the environment, determine the position of obstacles in the robot's coordinate system, and update the passability status of the corresponding directional channel based on the direction channel to which the obstacle belongs and its distance from the robot. The robot's path planning and motion status are acquired in real time to determine whether the conditions for escaping obstacles are met. When the escape trigger condition is triggered, the corresponding escape strategy is executed based on the type of trigger condition and the latest updated passability status of each directional channel.

[0005] Secondly, the technical solution of the present invention provides an escape system for a tracked robot, comprising: The directional channel division unit is used to establish a robot coordinate system with the robot center as the origin, and divide the space around the robot into four directional channels: front, back, left, and right. The directional channel status update unit is used to use the gimbal sensor on the robot to perceive the environment, determine the position of the obstacle in the robot coordinate system, and update the passability status of the corresponding directional channel according to the directional channel to which the obstacle belongs and the distance from the robot. The escape trigger condition judgment unit is used to acquire the robot's path planning status and motion status in real time to determine whether the escape trigger condition is met. The escape strategy execution unit is used to execute the corresponding escape strategy based on the latest updated passability status of each directional channel when the escape trigger condition is triggered, according to the type of trigger condition.

[0006] Thirdly, the technical solution of the present invention provides a terminal, comprising: Memory, used to store the obstacle avoidance program for the tracked robot; A processor, configured to implement the steps of the tracked robot's escape procedure as described in any of the preceding claims, in executing the tracked robot's escape procedure.

[0007] Fourthly, the technical solution of the present invention provides a robot equipped with the aforementioned terminal.

[0008] As can be seen from the above technical solutions, this application has the following advantages: By dividing the continuous space around the robot into four directional channels—front, back, left, and right—and maintaining their passability, the complex problem of finding the best escape space is transformed into a problem of querying and judging the state of a finite number of discrete channels. This provides a concise, clear, and real-time updated environmental situation model for escape decision-making, making the robot's escape actions directional actions with clear directional choices, thus improving the rationality and safety of escape decision-making. Real-time environmental perception and channel status updates are performed, and robot status monitoring and trigger judgment are executed. Then, based on the channel status strategy, the robot can escape from the predicament. This ensures that the escape action is based on the latest environmental information and its own status, avoids the disconnect between perception, decision-making and execution, and enhances the robot's ability to adapt to dynamically changing environments. When different escape trigger conditions are met, the latest channel status is invoked according to the type of trigger condition, and the corresponding strategy is executed. Multiple specific escape strategies can be integrated and scheduled, such as escape strategies for planning anomalies, jamming, and turning anomalies, to achieve classified processing and improve the targeting and efficiency of processing. Attached Figure Description

[0009] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic flowchart of a method for an escape mechanism of a tracked robot provided in an embodiment of the present invention.

[0011] Figure 2 This is a schematic block diagram of an escape system for a tracked robot provided in an embodiment of the present invention.

[0012] Figure 3 This is a schematic diagram of the structure of a terminal provided in an embodiment of the present invention. Detailed Implementation

[0013] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0015] Figure 1 This is a schematic flowchart illustrating a method for an escape mechanism of a tracked robot according to an embodiment of the present invention. Figure 1 The executing entity can be an escape system for a tracked robot. The escape method for a tracked robot provided in this embodiment is executed by a computer device; correspondingly, the escape system for the tracked robot runs within the computer device. Depending on different requirements, the order of the steps in this flowchart can be changed, and some steps can be omitted.

[0016] like Figure 1 As shown, the method includes the following steps.

[0017] S1: Establish a robot coordinate system with the robot's center as the origin, and divide the space around the robot into four directional channels: front, back, left, and right.

[0018] S2 uses the gimbal sensor on the robot to perceive the environment, determine the position of the obstacle in the robot's coordinate system, and update the passability status of the corresponding directional channel based on the direction channel to which the obstacle belongs and the distance from the robot.

[0019] S3 acquires the robot's path planning and motion status in real time to determine whether the conditions for escaping obstacles are met.

[0020] S4. When the escape trigger condition is triggered, the corresponding escape strategy is executed according to the type of trigger condition and the latest updated passability status of each directional channel.

[0021] As a refinement and extension of the specific implementation of the above embodiments, in order to fully explain the specific implementation process of this embodiment, the following will provide possible embodiments to describe the specific implementation of the above steps in a non-limiting manner.

[0022] In this embodiment, step S1 first establishes a robot coordinate system. To facilitate a unified description of the relative relationship between the robot and its surrounding environment, a robot body coordinate system (referred to as the robot coordinate system) is defined. The coordinate system has its origin at the robot's geometric center. The geometric center is the intersection of the longitudinal and transverse axes of symmetry of the two tracks. The axis points directly in front of the robot's front end. The axis points to the left side of the robot, and the ZR axis is perpendicular to the left side. The plane points upwards, forming a right-handed coordinate system. This coordinate system is fixed to the robot's body and moves and rotates with the robot.

[0023] Based on robot coordinate system The horizontal direction around the robot (i.e. A two-dimensional space on a plane, with the origin as the reference point. Centered on the axis, the area is evenly divided into four independent sectors according to azimuth angle; each sector is called a "directional channel." The division rules are as follows: Forward Channel (F): Covers the azimuth angle satisfy This area primarily corresponds to the robot's normal forward movement direction and adjacent areas. Left Channel (L): Covers azimuth angle satisfy The area; Backward Channel (B): Covers the azimuth angle satisfy The area; Right Channel (R): Covers the azimuth angle satisfy The area.

[0024] Among them, azimuth angle The calculation method is as follows: In the robot coordinate system, for any point in space... Its azimuth , It is the arctangent function in the four quadrants.

[0025] Each directional passage is associated with a dynamically updated state attribute, called the "passable state." This state is a binary variable, defined as follows: "Clear": This means that no obstacles that would prevent the robot from passing through are detected within the preset effective detection distance of the passage (e.g., 4 meters). "Blocked": This indicates that there is at least one obstacle within the preset effective detection range of this passage, and that the obstacle is within the robot's center. The horizontal distance is less than the "warning distance" set for the passage (e.g., 1.5 meters).

[0026] The partitioning method in this embodiment discretizes the continuous, infinite obstacle avoidance space around the robot into four logical units with clear directional semantics. This reduces the computational complexity of real-time decision-making, and each channel is associated with the robot's basic motion direction. This allows the escape strategy to be directly mapped to a set of preset, verified basic action primitives, simplifying the conversion process from high-level decision-making to low-level execution. It eliminates the uncertainty in the intermediate planning stage, improves the reliability and response speed of the entire escape control chain, and enhances the predictability of behavior, thereby facilitating the understanding of the robot's intentions by surrounding intelligent agents.

[0027] In this embodiment, step S2 updates the passability status of the corresponding directional channel, specifically including the following steps S2.1 to S2.5.

[0028] S2.1, acquire the real-time pose parameters of the gimbal sensor and the distance measurement information of the obstacle.

[0029] Define a fixed offset for the mounting position of the gimbal sensor in the robot coordinate system. Establish the gimbal coordinate system To the robot coordinate system The conversion relationship is determined by the horizontal rotation angle of the gimbal. and vertical pitch angle Confirmed. Horizontal rotation angle. It is the angle of rotation of the gimbal base relative to the robot's zero position around the vertical axis, typically ranging from 0° to 360°, the vertical pitch angle. This refers to the angle at which the sensor module mounted on the gimbal rotates relative to the horizontal plane around its horizontal axis, ranging from, for example, -90° (view downwards) to +40° (view upwards). It offers a high degree of freedom, and because it is mounted at the highest point in the robot's center, it can fully perceive all information about the robot's own position. During normal tasks, the dual-spectrum camera performs a cyclical rotation, enabling real-time monitoring of the scene.

[0030] In some optional implementations, the gimbal sensor is a dual-spectrum gimbal, integrating a visible light imaging module and an infrared thermal imaging module. This enables the robot to have all-weather environmental perception capabilities: in well-lit conditions, it primarily relies on visible light images for high-precision identification; in low-light or no-light environments, it can continuously perceive the environment and heat-generating obstacles based on infrared thermal images, thereby ensuring the continuity and reliability of updates to the passability status of directional paths. Simultaneously, the dual-spectrum data also directly serves the robot's core inspection tasks, such as equipment infrared temperature measurement, realizing the reuse of sensing resources and the integration of system functions.

[0031] In practice, the gimbal sensor is controlled to perform a scanning action to obtain the original perception data of the obstacle target in the gimbal coordinate system, including real-time pose parameters and distance measurement information of the obstacle.

[0032] Real-time pose parameters include the horizontal deflection angle of the sensor beam pointing relative to the zero position of the gimbal coordinate system. with vertical deflection angle .

[0033] Obstacle ranging information refers to the straight-line distance from the sensor to the obstacle. When using lidar, the straight-line distance... The distance value calculated for the direct-return laser beam's time of flight (ToF); when using binocular vision, the straight-line distance. The distance value is obtained by converting disparity information calculated through stereo matching; when using a monocular vision + depth estimation model, the straight-line distance is... This is an approximate distance value inferred through models such as neural networks.

[0034] S2.2, based on pose parameters and ranging information, calculates the coordinates of the obstacle in the robot coordinate system through coordinate transformation.

[0035] Based on the original sensing data, calculate the spherical coordinates of the obstacle in the gimbal coordinate system. And convert it to rectangular coordinates in the gimbal coordinate system. :

[0036]

[0037]

[0038] Subsequently, through the coordinate transformation matrix Translation vector Transform the obstacle coordinates to the robot coordinate system to obtain the coordinates. :

[0039] S2.3, Based on the azimuth angle of the coordinates, determine the front, back, left, and right direction channels to which the obstacle belongs.

[0040] Based on the projection position of the obstacle on the robot's horizontal plane Calculate its azimuth. .

[0041] Based on azimuth Determine the direction of the obstacle within the preset angle range: like If so, it is determined to be a forward passage; like If so, it is determined to be the left passage; like If so, it is determined to be a backward passage; like If so, it is determined to be the right-side passage.

[0042] S2.4 Calculate the horizontal distance from the obstacle to the center of the robot and compare this distance with the preset warning distance for the corresponding directional passage.

[0043] Calculate obstacles to the robot center Euclidean distance on the horizontal plane : . A warning distance is preset for each directional passage. Preferred =1.5m.

[0044] S2.5, Based on the comparison results, update the passability status of the passage in that direction: if the horizontal distance is less than the warning distance, update the passability status of the passage to impassable; otherwise, if no obstacle meets the condition within the preset update cycle, update the passability status of the passage to passable.

[0045] Specifically, if If the obstacle is blocked, the current passable status of the passage in that direction will be marked as "impassable", and the current minimum distance value will be recorded. If, within the preset update cycle, a channel in a certain direction does not detect a distance less than [a certain value], [the following is true]. If an obstacle is found, the passage's passability status will be reset to "passable".

[0046] For a channel in the same direction, if multiple obstacles are detected within a single update cycle, the distance to the obstacle closest to the robot's center is selected as the distance to the robot's center. and based on that and The comparison results are used to determine and update the passability status of the channel.

[0047] In this embodiment, step S3, determining whether the escape trigger condition is met, specifically includes: acquiring the robot's path planning status in real time, and determining whether the path planning abnormality trigger condition is met based on the path planning status; acquiring the robot's motion status in real time, and determining whether the jamming trigger condition or the turning abnormality trigger condition is met based on the motion status.

[0048] Regarding the planning anomaly triggering condition, the path planning status includes a planning success flag and a feasible path sequence. If the planning success flag indicates failure or the feasible path sequence remains empty for M consecutive control cycles, then the planning anomaly triggering condition is determined to be met; where M is an integer greater than 1.

[0049] Specifically, the system monitors and acquires the planning success flag and feasible path sequence output from the robot path planning module in real time. The planning success flag is a Boolean state variable. When the planner successfully calculates a feasible path from the starting point to the destination based on the current map information, robot pose, and target point, the flag is set to "True" (True or 1); when the planner fails to calculate a feasible path for any reason, the flag is set to "False" (False or 0). The feasible path sequence is a data structure, an ordered set of pose points, representing the continuous trajectory from the current position to the target point that the planner has successfully calculated and that the robot can track and execute. When planning fails, the sequence is empty.

[0050] A status monitoring loop is run at fixed time intervals. In each control cycle, the planning success flag and feasible path sequence are read once. Simultaneously, a counter is maintained. In each control cycle, if the read planning success flag is "false," the counter is incremented by 1; if the planning success flag is "true," the counter is cleared. When the counter is not less than M, meaning the planning success flag indicates failure for M consecutive control cycles, the planning anomaly trigger condition is determined to be met.

[0051] The determination of an empty path sequence and the determination of the flag are performed in parallel and independently. In some architectures, the planner may output an empty path sequence due to internal logic errors, even when the flag does not explicitly report an error. In each control cycle, the system checks the feasible path sequence. If the sequence remains empty, after confirmation for the same M control cycles, it is determined that the planning anomaly triggering condition is met.

[0052] The integer M is a configurable fault tolerance threshold to prevent accidental escape attempts caused by momentary disturbances or sensor noise leading to single planning failures. For example, in a dynamic environment, a pedestrian briefly passing in front of the robot may cause a single planning failure, but planning resumes immediately after the pedestrian moves away. Brief planning failures do not require triggering complex escape procedures. Only when planning failures persist for a certain period of time, indicating that the obstacle is persistent and structural, is the robot confirmed to be trapped in a planning anomaly.

[0053] For escape from abnormal movement, determine whether the triggering conditions are met, specifically including the following steps S311 to S316.

[0054] S311, acquire the first set of sensor data and the second set of sensor data in real time. The first set of sensor data includes: the number of pulses fed back by the left and right track encoders, and the real-time current value of the drive motor; the second set of sensor data includes: the Z-axis angular velocity measured by the inertial measurement unit, and the calculated robot attitude angle.

[0055] S312, based on the number of pulses fed back from the left and right track encoders, calculates within the time window. Left-side average linear velocity and the average linear velocity on the right ; Calculate the moving average value of the real-time current of the drive motor ; Calculate the absolute value of the robot's roll angle based on the robot's attitude angle. .

[0056] Specifically, encoders coaxially mounted on the left and right drive motors provide real-time feedback on the angular displacement of the track rotation by analyzing the pulse count within a time window. By differentiating and converting the counts within the range, the average linear velocity on the left side can be obtained. Compared with the average linear velocity on the right .

[0057] The current feedback from the motor driver is read, and this value directly reflects the motor's output torque. A smoothed average current value is obtained by applying a moving average filter to the continuous current values. , is used to characterize the average load over a period of time.

[0058] The robot's attitude is estimated in real time by fusing gyroscope (angular velocity) and accelerometer data from the IMU, and is represented by Euler angles (roll, pitch, and yaw). This embodiment primarily focuses on the roll angle, its absolute value. It reflects the degree to which the robot tilts to the left or right.

[0059] S313, if the following conditions are met simultaneously, the jamming trigger condition is determined to be met: a) ;in, The preset lower speed threshold; b) ;in This is the preset upper limit threshold for current; c) The duration exceeded ,in The preset attitude angle threshold, This is a preset duration threshold.

[0060] "Stuck" refers to a situation where the robot's tracks slip severely between themselves and the ground, or where obstacles prevent the driving force from being effectively converted into movement. This embodiment accurately identifies the stuck state by fusing speed, current, and attitude data. Condition a) indicates that when the actual movement speed of both left and right tracks is lower than a preset lower speed threshold, the robot is almost stationary. Condition b) indicates that the average current of the motor is consistently higher than a preset upper current threshold, indicating that the drive system is outputting high torque in an attempt to push the robot, but has failed. Condition c) indicates that the absolute value of the robot's roll angle exceeds a preset attitude angle threshold, and this abnormal attitude persists for a period of time, indicating that the robot is not on flat ground, but may have one track stuck in a ditch or straddling an obstacle, causing the robot to tilt continuously. Only when all three conditions are met simultaneously will the system determine that the robot is "stuck," avoiding misjudgment.

[0061] S314, based on the Z-axis angular velocity, obtains the actual steering angular velocity after low-pass filtering. .

[0062] S315, Calculate the absolute value of the angular velocity deviation. ,in, This represents the robot's theoretical turning angular velocity.

[0063] S316, if The duration exceeds the preset abnormal duration. If so, the steering anomaly trigger condition is met; where, This is the preset angular velocity deviation threshold.

[0064] "Steering anomaly" refers to the robot's inability to effectively execute steering commands, usually caused by one track getting stuck, foreign objects becoming entangled, or extremely asymmetrical ground friction coefficients. This embodiment identifies this by monitoring the deviation between the command and the executed angular velocity.

[0065] First, the raw Z-axis angular velocity measured by the IMU is low-pass filtered to obtain a smoothed actual steering angular velocity. This is to filter out high-frequency noise and vehicle vibration interference. Then, the theoretical steering angular velocity command currently being sent to the chassis is obtained from the motion controller. Calculate the absolute value of the difference between the instruction and the actual value. If the absolute value of the angular velocity deviation continues to exceed the preset angular velocity deviation threshold... For example, 0.1 radians / second, and the duration exceeds For example, if the duration is 1.5 seconds, the abnormal steering trigger condition is determined to be met. This embodiment uses a duration threshold. To avoid false triggering due to momentary slippage or sensor noise, a steering mechanism malfunction is only confirmed when there is a persistent and uncorrectable deviation between the steering command and the actual response. In this embodiment, step S4 executes the corresponding escape strategy based on the type of triggering condition and the latest updated passability status of each directional channel.

[0066] If the trigger condition is a planning anomaly, the first escape strategy is executed. This strategy is configured to: query the passability status of the left, right, and rear directions in a preset priority order, select the first direction with the status "passable" as the target escape direction, and generate the first control command sequence: first, control the robot to rotate around its center in place until its front is aligned with the target escape direction; second, control the robot to move laterally a preset distance in the target escape direction; after completing the lateral movement, send a path replanning request to replan the global path with the robot's current position as the new starting point.

[0067] Specifically, in application scenarios with heavy vehicle traffic and pedestrian barriers, obstacles can completely block the robot's path. The robot is trapped at the edge of the inflated area and unable to plan a reasonable path. Traditionally, the robot would rotate in place to find a feasible path through trial and error. This method cannot immediately rotate to a feasible position, and the rotational trial-and-error behavior is also somewhat dangerous. Even worse, the robot may become completely trapped within the inflated area of ​​the grid map during global path planning, rendering both global and local path planning impossible and preventing the robot from operating normally. Traditionally, the robot rotates itself to find a feasible path, but the obstacles are not removed, and the robot finds itself still within the inflated area, leading to task blockage. This embodiment will achieve the escape behavior based on an updated four-channel drivability feature. If the robot is obstructed, it sequentially searches the left, right, and rear passage areas of the accessibility table for obstacles. If no obstacle is found on the left, the robot rotates to the left area and reactivates global path planning to bypass the obstacle and continue the task. If an obstacle exists on the left, it searches the right and rear sides in the same way. If both are obstructed, a warning message is issued, and human intervention is required to extricate the robot. If the robot is completely trapped in the expansion area, it sequentially searches the left, right, and rear passage areas of the accessibility table for obstacles. If no obstacle is found on the left, the robot rotates to the left area and then moves forward autonomously by 0.2m. At this point, global path planning is reactivated again to bypass the obstacle and continue the task. If an obstacle exists on the left, it searches the right and rear sides in the same way. If both are obstructed, a warning message is issued, and human intervention is required to extricate the robot.

[0068] If the trigger condition is a jam, then the second escape strategy is executed. This strategy is configured to: query the passability status of the front and rear channels; if the status of both the front and rear channels is "passable", then generate a second control command sequence: with a preset period T1, alternately control the left track and the right track to perform short drives in opposite directions for a duration of Δt1, forming a periodic alternating forward and reverse action, which is executed continuously for N1 cycles; where N1 is a positive integer.

[0069] If the trigger condition is a steering anomaly, the third escape strategy is executed. This strategy is configured as follows: generate gimbal control commands to drive the gimbal sensors to acquire images of the track areas on both sides of the robot, call a preset visual recognition model to process the acquired images, identify whether there are preset types of entangled foreign objects in the images, generate an alarm signal and upload it if entangled foreign objects are identified, and control the robot to enter a safe stop state. If no entangled foreign objects are identified, generate a third control command sequence: apply speed pulse commands of equal magnitude and opposite direction to the left and right tracks alternately at a preset high-frequency period T2, and continue to execute for N2 cycles; where N2 is a positive integer.

[0070] A pre-built visual recognition model is used for online real-time detection of foreign object entanglement in the track and drive wheel areas of the tracked robot. It can complete the reasoning of a single frame image within the robot's control cycle to meet the immediacy requirements of escape decision-making. Preferably, a cropped and optimized single-stage target detection algorithm is used, such as a lightweight version of the YOLO (You Only Look Once) series or MobileNet-SSD.

[0071] Based on engineering experience, several categories of the most common entangled foreign objects have been defined, such as: plastic_bag (plastic bag, plastic film), rope_string (rope, cable, straw rope), wire_metal (iron wire, metal wire), cloth_fabric (cloth strips, textiles), and vegetation (grass, vines, suitable for outdoor scenarios). These categories can be expanded or simplified according to the specific application scenario of the robot.

[0072] The training data originated from the robot's development and debugging phase. Using its dual-spectrum gimbal, image data was intentionally simulated or actually collected under various typical scenarios when the tracks were entangled or scraped by various foreign objects. Offline enhancement was performed on the collected images, including simulating different lighting conditions, weather conditions, shooting angles, and partial occlusion, to improve the model's generalization ability. In the images, rectangular bounding boxes were used to select the regions of foreign objects entangled on the tracks, drive wheels, driven wheels, or tension wheels, and their categories were labeled.

[0073] Once the system determines that the "steering anomaly trigger condition" is met, the model is invoked on demand: the gimbal is rotated to a specific preset position, aligning its visible light camera with the robot's left and right track drive areas for rapid image capture, obtaining high-definition close-up images; the acquired images are scaled to the model's specified input size and fed into a preset visual recognition model for forward inference; the model output includes the predicted bounding box, class confidence score, and class label. A confidence threshold is set; any prediction with a confidence score higher than this threshold is considered a valid detection.

[0074] If any bounding box in the model output belongs to a predefined category of entangled foreign object and the confidence level meets the requirement, the system immediately generates a "Foreign Object Entanglement Confirmation" signal. This signal will trigger a high-priority safety shutdown command and simultaneously send an alarm message, including an abnormal image and annotation results, to the remote monitoring terminal via the communication link, requesting manual intervention. At this time, the system will not execute the "oscillation escape strategy" to prevent further damage.

[0075] If the model does not output any detection results higher than the confidence threshold, the system determines that "no non-target foreign object entanglement exists." This means that the abnormal steering may be caused by other reasons, such as severe track slippage, loose gravel on the ground, or internal mechanical failure of the drive mechanism. In this case, the system will safely execute the third escape strategy, namely high-frequency oscillation escape action, attempting to get rid of any small obstacles that may exist, are not within the model category, or have inconspicuous visual features, through vibration.

[0076] Specifically, the robot monitors the differences in track movement in real time and integrates IMU data and track encoder data to determine if the robot is in an abnormal movement state. If the track encoder speed is zero or extremely low, but the robot motor is still outputting high torque, and the IMU shows abnormal robot posture (such as a continuous increase in pitch or roll angle), it is determined to be "stuck," and the robot is considered to be in a state of insufficient driving force for climbing or crossing obstacles. This embodiment checks the passability of the robot's front and rear passages. If there are no obstacles in the front and rear passage areas, an alternating forward and reverse rotation escape strategy is adopted. By controlling the tracks on both sides to alternately rotate forward and reverse, the robot produces a "crawling" or "swaying" motion, gradually changing the robot's contact state with the ground, and using the grip of the tracks to gradually "climb" out of the predicament. If obstacles exist in both the front and rear passage areas, the task is terminated, requiring manual intervention to extricate the robot. If the system detects a persistent and uncorrectable deviation between the robot's actual rotation angle and the commanded angle, or if frequent pauses and vibrations occur during rotation, the system determines that the robot may be stuck by obstacles and unable to complete the turn smoothly, assuming that the robot's tracks may be entangled in debris and trapped. In this embodiment, the gimbal will rotate 360° to monitor the track information on both sides, and the YOLO algorithm will be applied to detect whether there are abnormal types of obstacles such as plastic bags or ropes. If abnormal types of obstacles are detected, the inspection task will be canceled and the situation will be reported. If no abnormal types of obstacles are detected, an oscillation extrication strategy will be adopted. By rapidly and slightly alternating the rotation speed and direction of the tracks on both sides, the robot will generate high-frequency micro-vibrations, thereby shaking off the debris entangled in the tracks or drive wheels and restoring the normal movement capability of the tracks.

[0077] In some alternative implementations, the path planning state is obtained based on an enhanced environment map containing spatial height information, through a path planning algorithm; wherein the path planning process includes steps 1 to 4.

[0078] Step 1: Acquire data collected by the lidar and inertial measurement unit on the robot, and construct a three-dimensional environment map using the laser-inertial odometry method; project the point cloud data in the three-dimensional environment map onto the horizontal plane, and record the lowest height value of the corresponding area for the projected two-dimensional grid to form an enhanced environment map.

[0079] To enable accessibility assessment of complex terrains such as steps, potholes, and slopes, this implementation constructs an enhanced environmental map that includes spatial elevation information.

[0080] The robot is equipped with a 16-line LiDAR and a 9-axis inertial measurement unit (IMU). Through a laser-inertial odometry (LIO) algorithm, the continuous frame point cloud data from the LiDAR and the high-frequency inertial data from the IMU are tightly coupled and optimized to obtain a high-precision robot pose in real time, and simultaneously construct a 3D point cloud map with centimeter-level accuracy. This map fully records the 3D geometry of the environment.

[0081] Project all points from the 3D point cloud map onto a horizontal plane (XY plane) along the direction of gravity (Z-axis); divide the horizontal plane into uniform 2D grids. For all points that fall within the same grid after projection, calculate and record the minimum height value of all points within that grid area. The final result is an enhanced map, where each grid cell not only contains the traditional "occupied," "free," and "unknown" states, but also additionally stores... property.

[0082] Step 2: On the enhanced environment map, perform global path planning based on the hybrid A* search algorithm to obtain the initial global path.

[0083] To ensure the safe passage of tracked robots, accessibility preprocessing is performed before global planning on the enhanced environment map.

[0084] The drivability preprocessing includes: traversing each "free" cell in the enhanced map before invoking the planning algorithm; calculating the drivability of that cell with all "free" cells in its eight neighborhoods. The absolute value ΔH of the difference between the values. If any ΔH is greater than a preset height change threshold, such as 15 cm, the grid is determined to be impassable, and its status is temporarily changed to "occupied". This operation aims to avoid terrain with height changes that may cause the robot to overturn or bottom out, such as the edges of steep slopes or deep pits.

[0085] On the preprocessed map, a hybrid A* search algorithm is employed. This algorithm searches in the state space (x, y, θ) and can directly generate smooth paths that satisfy the robot's nonholonomic motion constraints. This effectively avoids the post-processing smoothing step required after the traditional A* algorithm generates a polyline path. Furthermore, the generated path better conforms to the kinematic characteristics of the tracked robot and can be directly used for tracking control.

[0086] Step 3: Control the robot to move along the initial global path and acquire sensor data of the surrounding environment in real time; project the obstacles perceived in real time onto the local environment map and perform dilation processing.

[0087] As the robot moves along the global path, it continuously uses sensors such as LiDAR and binocular cameras to perceive the surrounding environment in real time. The perceived obstacle positions are projected and updated in real time onto a local cost map centered on the robot. Based on the robot's actual contours, obstacles are inflated, essentially defining a "no-go zone" around the obstacle to ensure the robot's safety envelope does not come into contact with it.

[0088] Step 4: Based on the local environment map, predict the robot's trajectory within a preset distance in the future; if the predicted trajectory will collide with the expanded obstacle, it is determined that there is a collision risk, triggering a global path replanning starting from the robot's current position.

[0089] The local planner employs Model Predictive Control (MPC) or Time Elastic Band (TEB) algorithms to track the global path and predict the robot's trajectory within a preset look-ahead distance based on the current local cost map. It continuously checks whether this predicted trajectory crosses "no-go zones" (expanded obstacle areas) in the local cost map. If a collision risk is detected, the system immediately determines that the current global path is no longer feasible in that local area. Instead of attempting a forced detour, the local planner sends an interruption signal to the upper layer, triggering a completely new global path replanning. This replanning uses the robot's current position as a new starting point and incorporates the latest environmental information—the updated enhanced map—to attempt to plan an alternative path that bypasses the current congestion.

[0090] The path planning process described above is the primary source of planning anomaly triggering conditions. In step S4, when local sensing detects a complete blockage ahead and global path replanning fails consecutively, the path planning module will enter a state of continuous failure, satisfying the condition that "within M consecutive control cycles, the planning success flag indicates failure," thus triggering a planning anomaly escape. When constructing the enhanced map, if the robot is in an area completely surrounded by terrain with significant elevation differences, all directional grids may be marked as "occupied" after preprocessing, causing global planning in step 2 to fail from the starting point. This will also quickly trigger planning anomaly conditions.

[0091] The above text describes in detail an embodiment of a method for escaping a tracked robot from a difficult situation. Based on the escaping method for a tracked robot described in the above embodiment, this invention also provides an escaping system for a tracked robot corresponding to the method.

[0092] Figure 2This is a schematic block diagram of an escape system for a tracked robot provided in an embodiment of the present invention. In this embodiment, the escape system 200 of the tracked robot can be divided into multiple functional units according to the functions it performs. A unit, as referred to in this invention, is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and is stored in memory.

[0093] The directional channel division unit 210 is used to establish a robot coordinate system with the robot center as the origin and divide the space around the robot into four directional channels: front, back, left, and right.

[0094] The directional channel status update unit 220 is used to use the gimbal sensor on the robot to perceive the environment, determine the position of the obstacle in the robot coordinate system, and update the passability status of the corresponding directional channel according to the directional channel to which the obstacle belongs and the distance from the robot.

[0095] The escape trigger condition judgment unit 230 is used to acquire the robot's path planning status and motion status in real time to determine whether the escape trigger condition is met.

[0096] The escape strategy execution unit 240 is used to execute the corresponding escape strategy according to the type of the trigger condition and the latest updated passability status of each directional channel when the escape trigger condition is triggered.

[0097] The tracked robot escaping system of this embodiment is used to implement the aforementioned tracked robot escaping method. Therefore, the specific implementation of this system can be found in the embodiment section of the tracked robot escaping method above. Thus, the specific implementation can be referred to the description of the corresponding embodiments, and will not be elaborated here.

[0098] Furthermore, since the tracked robot's obstacle-avoidance system in this embodiment is used to implement the aforementioned obstacle-avoidance method for the tracked robot, its function corresponds to the function of the aforementioned method, and will not be repeated here.

[0099] Figure 3 This is a schematic diagram of the structure of a terminal 300 provided in an embodiment of the present invention, including: a processor 310, a memory 320, and a communication unit 330. The processor 310 is used to implement the process steps of the above-described tracked robot escape method embodiment when implementing the tracked robot escape program stored in the memory 320.

[0100] This invention also provides a robot configured with the aforementioned terminal 300.

[0101] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for escaping obstacles for a tracked robot, characterized in that, Includes the following steps: Establish a robot coordinate system with the robot's center as the origin, and divide the space around the robot into four directional channels: front, back, left, and right. The robot uses its onboard gimbal sensors to perceive the environment, determine the position of obstacles in the robot's coordinate system, and update the passability status of the corresponding directional channel based on the direction channel to which the obstacle belongs and its distance from the robot. The robot's path planning and motion status are acquired in real time to determine whether the conditions for escaping obstacles are met. When the escape trigger condition is triggered, the corresponding escape strategy is executed based on the type of trigger condition and the latest updated passability status of each directional channel.

2. The method for escaping obstacles for a tracked robot according to claim 1, characterized in that, The robot uses its onboard gimbal sensors to perceive the environment, determine the position of obstacles in the robot's coordinate system, and update the passability status of the corresponding directional channel based on the direction of the obstacle and its distance from the robot. Specifically, this includes: Acquire the real-time pose parameters of the gimbal sensor and the distance measurement information of obstacles; Based on pose parameters and ranging information, the coordinates of the obstacle in the robot coordinate system are calculated through coordinate transformation. Based on the azimuth of the coordinates, determine the front, back, left, or right direction of the passage to which the obstacle belongs; Calculate the horizontal distance from the obstacle to the center of the robot and compare this distance with the preset warning distance for the corresponding directional path; Based on the comparison results, the passability status of the passage in that direction is updated: if the horizontal distance is less than the warning distance, the passability status of the passage is updated to impassable; otherwise, if no obstacle meets the condition within the preset update cycle, the passability status of the passage is updated to passable.

3. The method for escaping obstacles for a tracked robot according to claim 2, characterized in that, Real-time monitoring of the robot's path planning and motion status to determine whether the escape trigger conditions are met, specifically including: The robot's path planning status is acquired in real time, and the path planning status is used to determine whether the conditions for triggering planning anomalies are met. The robot's motion status is acquired in real time, and the motion status is used to determine whether the jamming trigger condition or the abnormal turning trigger condition is met.

4. The method for escaping obstacles for a tracked robot according to claim 3, characterized in that, The path planning status includes a planning success flag and a feasible path sequence. If the planning success flag indicates failure or the feasible path sequence remains empty for M consecutive control cycles, the planning anomaly triggering condition is determined to be met; where M is an integer greater than 1.

5. The method for escaping obstacles for a tracked robot according to claim 3, characterized in that, The robot's motion state is acquired in real time, and based on the motion state, it is determined whether the jamming trigger condition or the abnormal turning trigger condition is met. Specifically, this includes: The system acquires data from the first set of sensors and the second set of sensors in real time. The first set of sensor data includes the number of pulses fed back by the encoders on the left and right tracks, as well as the real-time current value of the drive motor. The second set of sensor data includes the Z-axis angular velocity measured by the inertial measurement unit, as well as the calculated robot attitude angle. Based on the number of pulses fed back from the left and right track encoders, the calculation is performed within the time window. Left-side average linear velocity and the average linear velocity on the right ; Calculate the moving average value of the real-time current of the drive motor ; Calculate the absolute value of the robot's roll angle based on the robot's attitude angle. ; If the following conditions are met simultaneously, the blocking trigger condition is determined to be satisfied: a) ;in, The preset lower speed threshold; b) ;in This is the preset upper limit threshold for current; c) The duration exceeded ,in The preset attitude angle threshold, This is a preset duration threshold; The actual steering angular velocity is obtained by low-pass filtering based on the Z-axis angular velocity. ; Calculate the absolute value of the angular velocity deviation. ,in, This represents the robot's theoretical turning angular velocity; like The duration exceeds the preset abnormal duration. If so, the steering anomaly trigger condition is met; where, This is the preset angular velocity deviation threshold.

6. The method for escaping obstacles for a tracked robot according to claim 3, characterized in that, When the escape trigger condition is triggered, the corresponding escape strategy is executed based on the type of trigger condition and the latest updated passability status of each directional passage, specifically including: If the trigger condition is a planning anomaly, the first escape strategy is executed. This strategy is configured to: query the passability status of the left, right, and rear directions in a preset priority order, select the first direction with the status "passable" as the target escape direction, and generate the first control command sequence: first, control the robot to rotate around its center until its front is aligned with the target escape direction; second, control the robot to move laterally a preset distance in the target escape direction; after completing the lateral movement, send a path replanning request to replan the global path with the robot's current position as the new starting point. If the trigger condition is a jam, then the second escape strategy is executed. This strategy is configured to: query the passability status of the forward and backward passages; if both the forward and backward passages are "passable", then generate a second control command sequence: with a preset period T1, alternately control the left track and the right track to perform short drives in opposite directions for a duration of Δt1, forming a periodic forward and reverse alternating action, which is executed continuously for N1 periods; where N1 is a positive integer. If the trigger condition is a steering anomaly, the third escape strategy is executed. This strategy is configured as follows: generate gimbal control commands to drive the gimbal sensors to acquire images of the track areas on both sides of the robot, call a preset visual recognition model to process the acquired images, identify whether there are preset types of entangled foreign objects in the images, generate an alarm signal and upload it if entangled foreign objects are identified, and control the robot to enter a safe stop state. If no entangled foreign objects are identified, generate a third control command sequence: apply speed pulse commands of equal magnitude and opposite direction to the left and right tracks alternately at a preset high-frequency period T2, and continue to execute for N2 cycles; where N2 is a positive integer.

7. The method for escaping obstacles of a tracked robot according to any one of claims 3-6, characterized in that, The path planning state is obtained based on an enhanced environment map containing spatial height information, through a path planning algorithm; the path planning process includes: Data collected by the lidar and inertial measurement unit on the robot is acquired, and a three-dimensional environment map is constructed using the laser-inertial odometry method. The point cloud data in the three-dimensional environment map is projected onto the horizontal plane, and the lowest height value of the corresponding area is recorded for the projected two-dimensional grid to form an enhanced environment map. On the enhanced environment map, global path planning is performed based on the hybrid A* search algorithm to obtain the initial global path; The robot is controlled to move along an initial global path and acquire sensor data of the surrounding environment in real time; the obstacles perceived in real time are projected onto a local environment map and inflated. Based on the local environment map, the robot's trajectory is predicted within a preset distance in the future; if the trajectory is predicted to collide with the expanded obstacle, it is determined that there is a collision risk, triggering a global path replanning starting from the robot's current position.

8. An escape system for a tracked robot, characterized in that, include: The directional channel division unit is used to establish a robot coordinate system with the robot center as the origin, and divide the space around the robot into four directional channels: front, back, left, and right. The directional channel status update unit is used to use the gimbal sensor on the robot to perceive the environment, determine the position of the obstacle in the robot coordinate system, and update the passability status of the corresponding directional channel according to the directional channel to which the obstacle belongs and the distance from the robot. The escape trigger condition judgment unit is used to acquire the robot's path planning status and motion status in real time to determine whether the escape trigger condition is met. The escape strategy execution unit is used to execute the corresponding escape strategy based on the latest updated passability status of each directional channel when the escape trigger condition is triggered, according to the type of trigger condition.

9. A terminal, characterized in that, include: Memory, used to store the obstacle avoidance program for the tracked robot; A processor, configured to implement the steps of the tracked robot's escape procedure as described in any one of claims 1 to 7.

10. A robot, characterized in that, The device is equipped with the terminal described in claim 9.

Images