Patrol control method and system for quadruped robot dog

Abstract

The invention provides a quadruped robot dog patrol control method and system, and belongs to the technical field of safety protection. Comprising the following steps: marking static obstacles and passing areas on a point cloud map of a fortification area, and writing the static obstacles and passing areas into a four-footed robot dog; the quadruped robot dog regularly executes a global patrol task, constructs a patrol path and patrols all passing areas; when the global patrol task is executed, the quadruped robot dog also recognizes personnel identities in the view range, if the personnel are not authenticated, the global patrol task is interrupted, the non-authenticated personnel are tracked and repelled until the non-authenticated personnel leave the boundary of the fortified area, and the quadruped robot dog returns to execute the global patrol task after the non-authenticated personnel are confirmed to leave; the quadruped robot dog further responds to the auxiliary task request of the authentication personnel, dynamically plans a temporary path according to the auxiliary task requirement and the target position, and returns to the initial position and state before the auxiliary task request after reaching the position of the authentication personnel and completing the auxiliary task request.

Description

Technical Field

[0001] This invention relates to the field of safety protection technology, and in particular to a method and system for patrol control of a quadruped robot dog. Background Technology

[0002] Mobile robots can be categorized into wheeled robots, legged robots, and tracked robots. Wheeled robots have simple structures and controls, but poor obstacle-crossing capabilities. Tracked robots have low ground pressure and good road adaptability, but they are heavy, vibrate significantly during movement, and are less flexible. With the continuous development of robotics technology, legged quadrupedal robot dogs have the ability to walk on different road surfaces and cross obstacles, adapting to various tasks in different situations. Their limbs can flexibly adjust their posture, exhibiting excellent motion tolerance and enabling them to complete complex tasks.

[0003] Currently, quadruped robot dogs have limited applications in autonomous navigation and security control, and their flexibility is insufficient, often requiring human guidance when performing security patrol tasks. Therefore, it is essential to provide a quadruped robot dog patrol control method and system that enhances the robot dog's ability to independently complete security patrol tasks and respond promptly to assistance requests from certified personnel. Summary of the Invention

[0004] In view of this, the present invention proposes a patrol control method and system for a quadruped robot dog that can autonomously plan patrol routes in a fortified area and reliably respond to interruptions.

[0005] This invention provides a method for controlling the patrol of a quadruped robot dog, comprising the following steps:

[0006] S1: A point cloud map of the fortified area. Static obstacles and passageways are marked on the point cloud map and written into the quadruped robot dog.

[0007] S2: The quadruped robot dog regularly performs full-area patrol tasks, constructs patrol routes, and inspects all passable areas;

[0008] S3: When performing a full-area patrol mission, the quadruped robot dog also identifies the identity of personnel within its field of vision. If the personnel are certified, the robot dog ignores them; if the personnel are not certified, the robot dog interrupts the full-area patrol mission and tracks and drives away the uncertified personnel until they leave the boundary of the protected area. After confirming that the uncertified personnel have left, the quadruped robot dog returns to perform the full-area patrol mission.

[0009] S4: When the quadruped robot dog is in a state other than tracking and driving away, it also responds to the auxiliary task request of the certifier. According to the auxiliary task requirements and target location, it dynamically plans a temporary path to reach the certifier's location. After completing the auxiliary task request, the quadruped robot dog returns to the starting position and state before the auxiliary task request.

[0010] Based on the above technical solution, preferably, step S1 also includes an image preprocessing step, specifically: acquiring environmental images through the camera mounted on the quadruped robot dog; obtaining the camera's pose in the coordinate system of the quadruped robot dog based on the environmental images; constructing an indoor dense point cloud map using visual SLAM; performing point cloud preprocessing by using voxel grid downsampling to remove outliers; selecting a point cloud at any location as the center point; calculating the local plane normal vector within the neighborhood of the center point; considering adjacent local plane normal vectors with an angle not exceeding 15° and a curvature not exceeding 0.05 as coplanar; fusing coplanar local planes to obtain a new local plane; and using non-fusing local planes as new planes; traversing all point clouds and establishing local planes for all of them.

[0011] Preferably, the step S1 of marking static obstacles and passage areas on the point cloud map is based on the following rules for determining passage areas: 1) When the angle of inclination of a local plane relative to the horizontal plane is less than 30°, or the height difference between adjacent local planes does not exceed 0.15m, it is determined to be a passage area; 2) When the angle of inclination of a local plane relative to the horizontal plane is between [30°, 45°], it is further determined whether there is a stepped distribution of local planes with a continuous tread depth greater than 0.25m. If there are stepped distribution local planes that meet the requirements, it is determined that there is a staircase. Steps are used as passage areas; if no local plane with a step distribution that meets the requirements does not exist, it is judged as an obstacle; 3) when the angle of inclination of a local plane relative to the horizontal plane is greater than 45°, it is judged as an obstacle; 4) when the point cloud density of a local plane is lower than the point cloud threshold, it is judged as an obstacle; 5) for holes with a radius not exceeding 0.15m, morphological processing is performed, and after closure, they are merged with the nearest local plane; for holes with a radius greater than 0.15m, they are judged as obstacles; the passage areas and obstacles corresponding to the local planes are categorized and written into the quadruped robot dog.

[0012] More preferably, step S2 consists of: after determining the passable area and obstacles, let the set of passable areas be R = {R1, R2, ..., R...} n Let R be any two adjacent connected passage regions. i and R j Let G, i, j∈1,2,...,n. ij For the passage area R i and R j For each path between the points, set the travel cost, movement path constraint, and coverage penalty for all paths. Construct the final total path cost Γ based on the travel cost, movement path constraint, and coverage penalty. Optimize the final total path cost Γ to obtain the path with the minimum final total path cost as the patrol path.

[0013] A further optimized approach is to set the cost of all paths to ∑w. ij Where any two adjacent connected passage regions R i and R j The path travel cost is w ij ;

[0014] The path constraint term is to define any two adjacent connected travel regions R. i and R j It is obtained by summing the integrals of the squares of the path curvatures between them and adjusting the proportions using the first adjustment coefficient;

[0015] The penalty term is an exponential function constructed based on the number of access areas used by the patrol route relative to the total number of access areas, and is proportionally adjusted by a second adjustment coefficient.

[0016] Further optimization involves finding the patrol path with the minimum total cost Γ of the final path, which includes the following:

[0017] A) If the overall outline of the fortified area is a rectangle or several passage areas are fitted into rectangles, and the aspect ratio of the rectangle does not exceed 3:1, and the rectangle has at least one entrance and exit, then several virtual quadrilaterals offset from the boundary of the rectangle are constructed by using an equally spaced loop scan method. The spacing d between adjacent virtual rectangles is equal. The final path of the quadruped robot dog is to traverse the vertices of each virtual quadrilateral in descending order using a loop trajectory. If there is an obstacle at a vertex or on the edge of a virtual quadrilateral, the quadruped robot dog moves towards a virtual quadrilateral with a smaller outline to avoid it.

[0018] B) If the overall outline of the protected area is a long rectangle or several passage areas are fitted into a long rectangle, and the aspect ratio of the long rectangle is greater than 3:1, and the long rectangle has at least one entrance or exit, then the final path of the quadruped robot dog is a straight path or a serpentine path, and the curvature of the serpentine path does not exceed the reciprocal of the minimum non-stationary turning radius of the quadruped robot dog.

[0019] C) If the overall outline of the fortified area is fitted to a circle or annulus, and several passage areas are fitted to a circle or annulus, then the final path of the quadruped robot dog is to move towards the center of the circle or annulus in the form of an Archimedean spiral.

[0020] D) If the overall outline of the fortified area or several passage areas are irregular areas, first extract the central axis of the irregular area, take the central axis as the main area, find the branch areas connected to the central axis, and form several branch paths connected to the central axis; fit the shape of each branch path, and combine the conditions of A), B), or C) to obtain the final path.

[0021] Preferably, the extraction of the central axis of the irregular region involves constructing a set of points equidistant from the boundary of the irregular region, using points near the entrance / exit as the starting point, fitting a curve to the starting point and non-current points to obtain the current curve, and using the portion connected to the current curve as the branch region. Then, fitting the set of points equidistant from the branch region boundary into a branch curve that intersects the current curve, and symmetrically constructing several offset curves on both sides of the current curve with equal spacing between adjacent offset curves. The longest current curve is obtained, or the current curve with the fewest deviations in the number of intersection points between the offset curves on both sides of the current curve and different branch curves is used as the central axis of the irregular region.

[0022] More preferably, step S3 involves the quadruped robot dog verifying the identity of personnel within its field of vision by reading RFID tags on employee badges or using facial recognition during the full-area patrol mission. If the identity verification is successful, the robot dog ignores the personnel and marks the vicinity of their location as an obstacle. If the identity verification fails, the personnel are considered unverified, the quadruped robot dog interrupts the full-area patrol mission, and executes the following deportation gradient strategy:

[0023] Non-verifying personnel should not stay for more than 30 seconds. The quadruped robot dog will activate its audible and visual warnings and maintain a safe distance from non-verifying personnel.

[0024] If an unverified person stays for more than 60 seconds, the quadruped robot dog will activate the forced removal mode. The quadruped robot dog will activate audible and visual warnings and its built-in voice prompts the unverified person to move to the nearest entrance / exit to leave the protected area, and will report the intrusion to the remote control center.

[0025] When an unverified person leaves their current location but does not move towards the nearest entrance or exit, the quadruped robot dog activates an audible and visual warning and its built-in voice prompts the unverified person to change direction and intercepts them in front of their current direction of movement. If the unverified person approaches the quadruped robot dog and is less than 2 / 3 of the safe distance, the quadruped robot dog retreats and maintains a safe distance from the unverified person.

[0026] If a non-verification personnel leaves the protected area for more than 5 seconds and the distance from the boundary of the protected area continues to increase, the quadruped robot dog returns to the final path position where the global patrol mission was interrupted and continues to perform the global patrol mission.

[0027] More preferably, the temporary path mentioned in step S4 is as follows: after the quadruped robot dog receives the assistance task request, it freezes the current state except for the tracking and expulsion operation, calculates the cost of reaching the target position from the current position, selects the path with the lowest cost as the temporary path, and performs gait adaptive adjustment; after completing the assistance task request, it returns along the temporary path.

[0028] On the other hand, the present invention also provides a quadruped robot dog perception and following control system for implementing the above-mentioned method, comprising:

[0029] The point cloud map annotation unit is configured on the quadruped robot dog to obtain the point cloud map of the fortified area, mark static obstacles and passage areas on the point cloud map, and write the positions of the obstacles and passage areas in the world coordinate system into the quadruped robot dog.

[0030] The patrol route construction unit establishes patrol routes for all-area patrol missions, covering all accessible areas, based on the fortified area.

[0031] The interrupt unit is used when the quadruped robot dog detects an uncertified person in its path while performing a full-domain patrol task. It then interrupts the full-domain patrol task, tracks and removes the uncertified person, and then returns to perform the full-domain patrol task. Alternatively, when the quadruped robot dog is in a state other than tracking and removing the uncertified person, it can also respond to the auxiliary task request of certified personnel, generate a temporary path and reach the target location. After completing the auxiliary task request, it returns to the initial state.

[0032] The quadruped robot dog patrol control method and system provided by this invention have the following advantages compared with the prior art:

[0033] (1) This invention provides a material constraint basis for subsequent path planning by establishing a high-precision map of passable areas, thereby improving the movement safety and motion accuracy of the quadruped robot dog; by reducing navigation complexity through planar fusion, and by setting the judgment criteria for passable areas, the terrain misjudgment rate can be significantly reduced, ensuring that the path conforms to the movement capabilities of the quadruped robot dog.

[0034] (2) For the patrol path of the whole-domain patrol mission, the final path total cost is first constructed by the path passage cost, curvature integral term and coverage penalty term. Then, the path is adapted according to the overall shape of the defense area or the local shape of the passage area, such as rectangular, narrow area, circle and ring and irregular area, which can improve the adaptability to different areas and improve the efficiency of the final path generation.

[0035] (3) For non-certified personnel, a tiered expulsion strategy was developed to respond in different levels and ensure the safety of the quadruped robot dog.

[0036] (4) For sudden auxiliary task requests, the cost of reaching the target location from the current location is constructed, taking into full account the distance to the target location, the distance from the obstacle and the curvature of the temporary path, so as to improve the safety and reliability of the temporary path, so as to better respond to the auxiliary task requests. After completing the auxiliary task request, the interruption point can be returned directly along the temporary path. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating the steps of the quadruped robot dog patrol control method and system of the present invention.

[0039] Figure 2 This is a schematic diagram of the loop-shaped trajectory path within a rectangular area of ​​the quadruped robot dog patrol control method and system of the present invention;

[0040] Figure 3 This is a schematic diagram of a straight or serpentine path for the quadruped robot dog patrol control method and system of the present invention;

[0041] Figure 4 This is a schematic diagram of the path in an irregular area of ​​the quadruped robot dog patrol control method and system of the present invention. Detailed Implementation

[0042] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0043] Quadruped robot dogs have limited applications in autonomous navigation and security control, and lack flexibility, often requiring manual guidance during security patrols. Therefore, if... Figure 1 As shown, the present invention provides a method for controlling the patrol of a quadruped robot dog, comprising the following steps:

[0044] S1: A point cloud map of the defended area. Static obstacles and passageways are marked on the point cloud map and written into the quadruped robot dog.

[0045] This step includes an image preprocessing step, specifically: acquiring environmental images using a camera mounted on a quadruped robot dog; obtaining the camera's pose in the quadruped robot dog's coordinate system based on the environmental images; constructing an indoor dense point cloud map using visual SLAM; performing point cloud preprocessing by voxel grid downsampling and removing outliers; selecting a point cloud at any location as the center point; calculating the local plane normal vector within the neighborhood of the center point; considering adjacent local plane normal vectors with an angle not exceeding 15° and a curvature not exceeding 0.05 as coplanar; fusing coplanar local planes to obtain a new local plane; and using non-fusing local planes as new planes; and traversing all point clouds and establishing local planes for all of them.

[0046] Voxel downsampling reduces map data volume and improves real-time processing depth. Local plane fusion reduces navigation complexity for quadruped robots. By constructing local planes, the surface of an object can be fitted, thus distinguishing its basic shape and providing a reference benchmark for subsequent obstacle and passage area segmentation. This step is well-suited for resource-constrained embedded systems, improving the efficiency of point cloud map construction.

[0047] The step S1, which involves marking static obstacles and passable areas on the point cloud map, establishes the following rules for determining passable areas: 1) When the angle of inclination of a local plane relative to the horizontal plane is less than 30°, or the height difference between adjacent local planes does not exceed 0.15m, it is determined to be a passable area; 2) When the angle of inclination of a local plane relative to the horizontal plane is between [30°, 45°], it is further determined whether there is a local plane with a continuous tread depth greater than 0.25m. If a local plane with a stepped distribution that meets the requirements exists, it is determined that a staircase exists. As a passage area; if there is no local plane with a stepped distribution that meets the requirements, it is judged as an obstacle; 3) when the angle of inclination of the local plane relative to the horizontal plane is greater than 45°, it is judged as an obstacle; 4) when the point cloud density of the local plane is lower than the point cloud threshold, it is judged as an obstacle; 5) for holes with a radius not exceeding 0.15m, after morphological processing, they are closed and merged with the nearest local plane; for holes with a radius greater than 0.15m, they are judged as obstacles; after classifying and labeling the passage area and obstacles corresponding to the local plane, they are written into the quadruped robot dog.

[0048] The tilt angle threshold here is to accommodate the climbing ability of the quadruped robot dog; the continuous tread depth detection of the stairs is to confirm the actual existence of the staircase structure; small holes are considered features that do not affect passage, while larger holes are considered obstacles. If the point cloud density is too low in a local area, it indicates the possible presence of unreliable areas such as water accumulation, which should be avoided as much as possible. In this embodiment, the number of point cloud thresholds for the local plane is 50. Through the above processing, the passage area and obstacles are marked in the point cloud region, and the passage areas can be further connected to construct patrol paths or temporary paths.

[0049] S2: The quadruped robot dog regularly performs full-area patrol tasks, constructs patrol routes, and inspects all passable areas.

[0050] After determining the passable areas and obstacles, let the set of passable areas be R = {R1, R2, ..., R...} n Let R be any two adjacent connected passage regions. i and R j Let G, i, j∈1,2,...,n. ij For the passage area R i and R j For each path between the points, set the travel cost, movement path constraint, and coverage penalty for all paths. Construct the final total path cost Γ based on the travel cost, movement path constraint, and coverage penalty. Optimize the final total path cost Γ to obtain the path with the minimum final total path cost as the patrol path.

[0051] The above can be understood as each passage area being an isolated island, requiring the establishment of a final path connecting as many passage areas as possible. Each passage area is connected by a path, and the final path should minimize repetition. If each path is an arc, it should be as smooth as possible, otherwise it may affect the safety of the quadruped robot dog's movement.

[0052] Specifically, let the cost of all paths be ∑w ij Any two adjacent connected passage regions R i and R j The path travel cost is w ij =A×||C(R) i )-C(R j )||+B×D(G ij A and B are weighted terms, A + B = 1, C(R) i ) and C(R j ) represent the geometric centers of the passage area, and ||·|| represents the distance calculation; parameter D(G) ij )=ω width ×f width +ωslope ×f slope +ω terr ×f terr ω width ω slope ω terr These are the width weight, slope weight, and terrain weight, respectively, ω width +ω slope +ω terr =1; f width For the cost of width, W is the path width. robot Let f be the width of the quadruped robot dog, k be the sensitivity coefficient, and f be the width cost. width A value of 0 indicates that the quadruped robot can pass smoothly along the path, while a value of 1 indicates that the path width is insufficient and the quadruped robot cannot pass; f slope For the sake of slope, θ is the average slope angle of the path relative to the ground; f terr As a consequence of terrain, f terr = a × Rough + b × (1 - μ), where a and b are weighting terms, Rough is the roughness value, and μ is the friction coefficient. This considers the travel area R. i and R j The path length, width, slope, and terrain cost are considered in relation to the path's passability. The width cost takes into account that the path should have a certain margin so that the quadruped robot dog carrying cargo can pass or turn smoothly. The slope cost evaluates the impact of the path's pitch angle on the robot dog's posture. The terrain cost considers the impact of ground roughness and friction on the quadruped robot dog. To increase the dynamic flexibility of the terrain cost, the number of obstacles per unit area and their weights can also be added to better describe the ease or difficulty of the path for the quadruped robot dog to pass.

[0053] Motion path constraint term λ∑k 2 ds(s), where λ is the first adjustment coefficient and k(s) is the number of adjacent connected passage regions R. i and R j The path curvature is defined as follows: s is the arc length parameter of the path, and ds is the change in arc length. The purpose of the motion path constraint term is to penalize high curvature paths, making the final path smoother and avoiding sharp bends in local paths. The summation of the square integral of the path curvature reflects the steering energy consumption. The path curvature k(s) is positively correlated with the joint torque of the quadruped robot dog, so the smaller this term is, the better.

[0054] Penalty items p is the second adjustment coefficient, R total R represents the total number of accessible areas. pathThe number of access areas used for the patrol route; this penalty is to ensure that the final route covers as many access areas as possible. The more access areas the final route passes through and the higher the proportion of the total number of access areas, the smaller the value of the penalty.

[0055] Based on the above calculations, the final total path cost is:

[0056] After constructing the final path total cost, targeted optimization is needed for different region shapes. By optimizing the final path total cost Γ, the path with the minimum final path total cost can be obtained more quickly as the patrol path, including the following:

[0057] A) such as Figure 2 As shown, if the overall outline of the fortified area is a rectangle or several passage areas are fitted into rectangles, and the aspect ratio of the rectangle does not exceed 3:1, and the rectangle has at least one entrance / exit, then several virtual quadrilaterals offset from the rectangular boundary are constructed by using an equally spaced spiral scan method. The spacing d between adjacent virtual rectangles is equal. The final path of the quadruped robot dog is to traverse the vertices of each virtual quadrilateral in descending order using a spiral trajectory. If there is an obstacle at a vertex or on the edge of a virtual quadrilateral, the quadruped robot dog moves towards a virtual quadrilateral with a smaller outline to avoid it.

[0058] The spacing d between adjacent virtual rectangles needs to ensure that when the quadruped robot dog moves along the loop trajectory, the robot dog's field of vision on the current virtual quadrilateral overlaps with the field of vision on the adjacent virtual quadrilateral by at least 10%.

[0059] In addition to the rectangle mentioned in scenario A), ellipses with a major-to-minor axis ratio not exceeding 3, sectors with a chord length-to-width ratio not exceeding 3, and new protruding structures obtained by merging multiple rectangles through Boolean operations can also use the strategy in scenario A) to construct virtual quadrilaterals or virtual sector structures with offset contours. After the quadruped robot dog completes its patrol along the above trajectory, it returns to the entrance / exit in a straight line or arc, or enters the next passage area.

[0060] B) such as Figure 3 As shown, if the overall outline of the protected area is a long and narrow rectangle or several passage areas are fitted into a long and narrow rectangle, and the aspect ratio of the long and narrow rectangle is greater than 3:1, and the long and narrow rectangle has at least one entrance and exit, then the final path of the quadruped robot dog adopts a straight path or a serpentine path. The curvature of the serpentine path does not exceed the reciprocal of the minimum non-stationary turning radius of the quadruped robot dog.

[0061] For narrow areas, a straight path can be used; for passages of a certain width, a serpentine, curved path can be used.

[0062] Similarly, the overall outline of a fortified area or the local shape of a passageway, such as an ellipse with a major-to-minor axis ratio greater than 3 or a narrow sector with a chord length-to-width ratio greater than 3, can also be handled using the method in scenario B).

[0063] C) If the overall outline of the protected area is fitted to a circle or annulus, and several passage areas are fitted to a circle or annulus, then the quadruped robot dog's final path adopts an Archimedean spiral motion toward the center of the circle or annulus; the Archimedean spiral motion can pass through most of the passage areas of the circle or annulus very well.

[0064] D) such as Figure 4 As shown, if the overall outline of the fortified area or several passage areas are irregular areas, the central axis of the irregular area is first extracted. The central axis is used as the main area, and branch areas connected to the central axis are found to form several branch paths connected to the central axis. The shape of each branch path is fitted, and each branch path is combined to satisfy the conditions A), B), or C) to obtain the final path.

[0065] The extraction of the central axis of the irregular region involves constructing a set of points equidistant from the boundary of the irregular region. Points near the entrance / exit are used as starting points. A curve is fitted to the starting point and non-current points to obtain the current curve. The portion connected to the current curve is designated as a branch region. The set of points equidistant from the branch region boundary is then fitted to a branch curve intersecting the current curve. Several offset curves are symmetrically constructed on both sides of the current curve, with equal spacing between adjacent offset curves. The longest current curve, or the current curve with the fewest deviations in the number of intersection points between the offset curves on both sides and different branch curves, is selected as the central axis of the irregular region. Figure 4 The curve at the center is taken as the current curve corresponding to the central axis, and the curves on both sides of the current curve are branch curves.

[0066] If the number of intersection points between the offset curves on both sides of the current curve deviates significantly, it indicates that the selected endpoint of the current curve is not suitable and needs to be re-selected. For branch curves that are too short, if the passage area involved is minimal, they can be ignored by pruning.

[0067] S3: When performing a full-area patrol mission, the quadruped robot dog also identifies the identity of personnel within its field of vision. If the personnel are certified, the robot dog ignores them; if the personnel are not certified, the robot dog interrupts the full-area patrol mission and performs tracking and removal operations on the uncertified personnel until they leave the boundary of the protected area. After confirming that the uncertified personnel have left, the quadruped robot dog returns to perform the full-area patrol mission.

[0068] Step S3 involves the quadruped robot dog verifying the identity of personnel within its field of vision by reading RFID tags on employee badges or using facial recognition. If the identity verification is successful, the robot dog ignores the personnel and marks the vicinity of their location as an obstacle. If the identity verification fails, the personnel are considered unverified, the quadruped robot dog interrupts the full-domain patrol mission, and executes the following removal gradient strategy:

[0069] Non-verifying personnel should not stay for more than 30 seconds. The quadruped robot dog will activate its audible and visual warnings and maintain a safe distance from non-verifying personnel.

[0070] If an unverified person stays for more than 60 seconds, the quadruped robot dog will activate the forced removal mode. The quadruped robot dog will activate audible and visual warnings and its built-in voice prompts the unverified person to move to the nearest entrance / exit to leave the protected area, and will report the intrusion to the remote control center.

[0071] When an unverified person leaves their current location but does not move towards the nearest entrance or exit, the quadruped robot dog activates an audible and visual warning and its built-in voice prompts the unverified person to change direction and intercepts them in front of their current direction of movement. If the unverified person approaches the quadruped robot dog and is less than 2 / 3 of the safe distance, the quadruped robot dog retreats and maintains a safe distance from the unverified person.

[0072] If a non-verification personnel leaves the protected area for more than 5 seconds and the distance from the boundary of the protected area continues to increase, the quadruped robot dog returns to the final path position where the global patrol mission was interrupted and continues to perform the global patrol mission.

[0073] In this embodiment, a tiered disengagement strategy is implemented to maintain a safe distance between the quadruped robot dog and non-verification personnel at all times, preventing malicious damage to the quadruped robot dog. The safe distance is no less than 3 meters.

[0074] S4: When the quadruped robot dog is in a state other than tracking and driving away, it also responds to the auxiliary task request of the certifier. According to the auxiliary task requirements and target location, it dynamically plans a temporary path to reach the certifier's location. After completing the auxiliary task request, the quadruped robot dog returns to the starting position and state before the auxiliary task request.

[0075] The temporary path mentioned in step S4 is defined as follows: after the quadruped robot dog receives an assistance task request, it freezes its current state (excluding the tracking and disengagement operation), calculates the cost (Cost) of reaching the target position from the current position, selects the path with the lowest cost as the temporary path, and performs gait adaptive adjustment; after completing the assistance task request, it returns along the temporary path; the cost of reaching the target position... Where P is the distance weight, D is the path length from the current position to the target position, Q is the obstacle weight, and d obsS represents the minimum distance between the path and obstacles in each passage area, where S is the curvature weight and k is the value of the path. max (s) represents the maximum value of the path curvature from the current position to the target position, P+Q+S=1.

[0076] This approach fully considers the target location distance, the distance from obstacles, and the curvature of the temporary path, thereby improving the safety and reliability of the temporary path and enabling better response to auxiliary task requests.

[0077] Gait adaptive adjustment allows the quadruped robot dog to patrol at high speed in open areas far from obstacles, such as more than 5 meters away, with a movement speed of no less than 1 m / s; if the quadruped robot dog is currently in a curved area, the movement speed will not exceed 0.7 m / s; if the quadruped robot dog is currently in a narrow passage with a passage width of no less than 1.3 times the width of the quadruped robot dog, the movement speed will not exceed 0.5 m / s.

[0078] On the other hand, the present invention also provides a quadruped robot dog perception and following control system for implementing the above-mentioned method, comprising:

[0079] The point cloud map annotation unit is configured on the quadruped robot dog to obtain the point cloud map of the fortified area, mark static obstacles and passage areas on the point cloud map, and write the positions of the obstacles and passage areas in the world coordinate system into the quadruped robot dog.

[0080] The patrol route construction unit establishes patrol routes for all-area patrol missions, covering all accessible areas, based on the fortified area.

[0081] The interrupt unit is used when the quadruped robot dog detects an uncertified person in its path while performing a full-domain patrol task. It then interrupts the full-domain patrol task, tracks and removes the uncertified person, and then returns to perform the full-domain patrol task. Alternatively, when the quadruped robot dog is in a state other than tracking and removing the uncertified person, it can also respond to the auxiliary task request of certified personnel, generate a temporary path and reach the target location. After completing the auxiliary task request, it returns to the initial state.

[0082] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling the patrol of a quadruped robot dog, characterized in that, Includes the following steps: S1: A point cloud map of the fortified area. Static obstacles and passageways are marked on the point cloud map and written into the quadruped robot dog. S2: The quadruped robot dog regularly performs full-area patrol tasks, constructs patrol routes, and inspects all passable areas; S3: When performing a full-area patrol mission, the quadruped robot dog also identifies the identity of personnel within its field of vision. If the personnel are certified, the robot dog ignores them; if the personnel are not certified, the robot dog interrupts the full-area patrol mission and tracks and drives away the uncertified personnel until they leave the boundary of the protected area. After confirming that the uncertified personnel have left, the quadruped robot dog returns to perform the full-area patrol mission. S4: When the quadruped robot dog is in a state other than tracking and driving away, it also responds to the auxiliary task request of the certifier. According to the auxiliary task requirements and target location, it dynamically plans a temporary path to reach the certifier's location. After completing the auxiliary task request, the quadruped robot dog returns to the starting position and state before the auxiliary task request.

2. The quadruped robot dog patrol control method according to claim 1, characterized in that, Step S1 also includes an image preprocessing step, specifically: acquiring environmental images using a camera mounted on a quadruped robot dog; obtaining the camera's pose in the quadruped robot dog's coordinate system based on the environmental images; constructing an indoor dense point cloud map using visual SLAM; performing point cloud preprocessing by voxel grid downsampling and removing outliers; selecting any point cloud location as the center point; calculating the local plane normal vector within the neighborhood of the center point; considering adjacent local plane normal vectors with an angle not exceeding 15° and a curvature not exceeding 0.05 as coplanar; fusing coplanar local planes to obtain a new local plane; and using non-fusing local planes as new planes; and traversing all point clouds and establishing local planes for all of them.

3. The quadruped robot dog patrol control method according to claim 2, characterized in that, The step S1, which involves marking static obstacles and passable areas on the point cloud map, establishes the following rules for determining passable areas: 1) When the angle of inclination of a local plane relative to the horizontal plane is less than 30°, or the height difference between adjacent local planes does not exceed 0.15m, it is determined to be a passable area; 2) When the angle of inclination of a local plane relative to the horizontal plane is between [30°, 45°], it is further determined whether there is a local plane with a continuous tread depth greater than 0.25m. If a local plane with a stepped distribution that meets the requirements exists, it is determined that a staircase exists. As a passage area; if there is no local plane with a stepped distribution that meets the requirements, it is judged as an obstacle; 3) when the angle of inclination of the local plane relative to the horizontal plane is greater than 45°, it is judged as an obstacle; 4) when the point cloud density of the local plane is lower than the point cloud threshold, it is judged as an obstacle; 5) for holes with a radius not exceeding 0.15m, after morphological processing, they are closed and merged with the nearest local plane; for holes with a radius greater than 0.15m, they are judged as obstacles; after classifying and labeling the passage area and obstacles corresponding to the local plane, they are written into the quadruped robot dog.

4. The quadruped robot dog patrol control method according to claim 3, characterized in that, Step S2 is as follows: After determining the passable area and obstacles, let the set of passable areas be R = {R1, R2, ..., R...} n Let R be any two adjacent connected passage regions. i and R j Let G, i, j∈1,2,...,n. ij For the passage area R i and R j For each path between the points, set the travel cost, movement path constraint, and coverage penalty for all paths. Construct the final total path cost Γ based on the travel cost, movement path constraint, and coverage penalty. Optimize the final total path cost Γ to obtain the path with the minimum final total path cost as the patrol path.

5. The quadruped robot dog patrol control method according to claim 4, characterized in that, Let the cost of all paths be ∑w ij Where any two adjacent connected passage regions R i and R j The path travel cost is w ij ; The path constraint term is to define any two adjacent connected travel regions R. i and R j It is obtained by summing the integrals of the squares of the path curvatures between them and adjusting the proportions using the first adjustment coefficient; The penalty term is an exponential function constructed based on the number of access areas used by the patrol route relative to the total number of access areas, and is proportionally adjusted by a second adjustment coefficient.

6. The quadruped robot dog patrol control method according to claim 5, characterized in that, By optimizing the total cost Γ of the final path, the path with the minimum total cost is obtained as the patrol path, including the following: A) If the overall outline of the fortified area is a rectangle or several passage areas are fitted into rectangles, and the aspect ratio of the rectangle does not exceed 3:1, and the rectangle has at least one entrance and exit, then several virtual quadrilaterals offset from the boundary of the rectangle are constructed by using an equally spaced loop scan method. The spacing d between adjacent virtual rectangles is equal. The final path of the quadruped robot dog is to traverse the vertices of each virtual quadrilateral in descending order using a loop trajectory. If there is an obstacle at a vertex or on the edge of a virtual quadrilateral, the quadruped robot dog moves towards a virtual quadrilateral with a smaller outline to avoid it. B) If the overall outline of the protected area is a long rectangle or several passage areas are fitted into a long rectangle, and the aspect ratio of the long rectangle is greater than 3:1, and the long rectangle has at least one entrance or exit, then the final path of the quadruped robot dog is a straight path or a serpentine path, and the curvature of the serpentine path does not exceed the reciprocal of the minimum non-stationary turning radius of the quadruped robot dog. C) If the overall outline of the fortified area is fitted to a circle or annulus, and several passage areas are fitted to a circle or annulus, then the final path of the quadruped robot dog is to move towards the center of the circle or annulus in the form of an Archimedean spiral. D) If the overall outline of the fortified area or several passage areas are irregular areas, first extract the central axis of the irregular area, take the central axis as the main area, find the branch areas connected to the central axis, and form several branch paths connected to the central axis; fit the shape of each branch path, and combine the conditions of A), B), or C) to obtain the final path.

7. The quadruped robot dog patrol control method according to claim 6, characterized in that, Extracting the central axis of an irregular region involves constructing a set of points equidistant from the boundary of the irregular region. Points near the entrance / exit are used as starting points. A curve is fitted to the starting point and non-current points to obtain the current curve. The part connected to the current curve is taken as a branch region. The set of points equidistant from the branch region boundary is then fitted to a branch curve intersecting the current curve. Several offset curves are symmetrically constructed on both sides of the current curve, with equal spacing between adjacent offset curves. The longest current curve is obtained, or the current curve with the fewest deviations in the number of intersection points between the offset curves on both sides of the current curve and different branch curves is selected as the central axis of the irregular region.

8. The quadruped robot dog patrol control method according to claim 4, characterized in that, Step S3 involves the quadruped robot dog verifying the identity of personnel within its field of vision by reading RFID tags on employee badges or using facial recognition. If the identity verification is successful, the robot dog ignores the personnel and marks the vicinity of their location as an obstacle. If the identity verification fails, the personnel are considered unverified, the quadruped robot dog interrupts the full-domain patrol mission, and executes the following removal gradient strategy: Non-verifying personnel should not stay for more than 30 seconds. The quadruped robot dog will activate its audible and visual warnings and maintain a safe distance from non-verifying personnel. If an unverified person stays for more than 60 seconds, the quadruped robot dog will activate the forced removal mode. The quadruped robot dog will activate audible and visual warnings and its built-in voice prompts the unverified person to move to the nearest entrance / exit to leave the protected area, and will report the intrusion to the remote control center. When an unverified person leaves their current location but does not move towards the nearest entrance or exit, the quadruped robot dog activates an audible and visual warning and its built-in voice prompts the unverified person to change direction and intercepts them in front of their current direction of movement. If the unverified person approaches the quadruped robot dog and is less than 2 / 3 of the safe distance, the quadruped robot dog retreats and maintains a safe distance from the unverified person. If a non-verification personnel leaves the protected area for more than 5 seconds and the distance from the boundary of the protected area continues to increase, the quadruped robot dog returns to the final path position where the global patrol mission was interrupted and continues to perform the global patrol mission.

9. The quadruped robot dog patrol control method according to claim 4, characterized in that, The temporary path mentioned in step S4 is that when the quadruped robot dog receives the assistance task request, it freezes the current state except for the tracking and expulsion operation, calculates the cost of reaching the target position from the current position, selects the path with the minimum cost as the temporary path, and performs gait adaptive adjustment; after completing the assistance task request, it returns along the temporary path.

10. A quadruped robot dog perception and following control system, used to implement the method described in any one of claims 1-9, characterized in that, include: The point cloud map annotation unit is configured on the quadruped robot dog to obtain the point cloud map of the fortified area, mark static obstacles and passage areas on the point cloud map, and write the positions of the obstacles and passage areas in the world coordinate system into the quadruped robot dog. The patrol route construction unit establishes patrol routes for all-area patrol missions, covering all accessible areas, based on the fortified area. The interrupt unit is used when the quadruped robot dog detects an uncertified person in its path while performing a full-domain patrol task. It then interrupts the full-domain patrol task, tracks and removes the uncertified person, and then returns to perform the full-domain patrol task. Alternatively, when the quadruped robot dog is in a state other than tracking and removing the uncertified person, it can also respond to the auxiliary task request of certified personnel, generate a temporary path and reach the target location. After completing the auxiliary task request, it returns to the initial state.

Images