Fall-preventing robot and fall-preventing method and device

Abstract

The application discloses a fall-preventing robot and a fall-preventing method and device, and relates to the field of intelligent robots, and aims to realize cliff identification and fall protection in all scenes. Whether a cliff exists in a first predetermined range and a second predetermined range around the robot body is detected, a fall risk is identified based on a detection result of a sensor module, and a motion state of the robot body is controlled based on the fall risk. The application synchronously detects the cliff through a near end and a far end, realizes cliff identification and fall protection in all scenes such as a high-speed moving state, an initial state, a state of being located at a cliff edge or a state of being close to the cliff edge, and has high reliability and safety, smooth system operation and high universality.

Description

Technical Field

[0001] This invention relates to the field of intelligent robot technology, and in particular to a fall-prevention robot, a fall-prevention method for robots, and a fall-prevention device for robots. Background Technology

[0002] Mobile robots (such as wheeled robots) are mobile platforms with the ability to adapt to complex environments and have been widely used in various scenarios such as home service, industrial inspection, and outdoor operations.

[0003] During robot navigation, it is inevitable to encounter special terrain such as cliffs (including steps). Among the known risk avoidance technologies, most fall prevention methods use infrared sensors to send coded signals in different directions. If no reflected signal is received within a predetermined time, the robot stops moving forward, thus preventing a fall. Alternatively, ultrasonic sensors can be used to obtain the distance between the robot and the obstacle, thereby determining whether the robot is within the defined fall range. If so, the robot stops moving forward to prevent a fall.

[0004] However, there is an extremely high risk of falling in the following scenarios, and the aforementioned known technologies cannot avoid this risk. Scenario 1: Robot start-up and stop phase: After the robot is moved by a human, its initial position may be close to the edge of a cliff, and the cliff position cannot be determined through historical navigation memory; Scenario 2: When traveling at high speed, it is necessary to identify distant cliffs in advance to allow for braking distance, and when traveling at low speed close to the edge, it is necessary to accurately detect nearby cliffs to avoid tipping over. Summary of the Invention

[0005] The purpose of this invention is to provide a fall-prevention robot, fall-prevention method, and device to address all or part of the problems mentioned above, so as to achieve cliff recognition and fall protection in all scenarios.

[0006] The technical solution adopted in this invention is as follows: A fall-resistant robot, comprising: The robot body, and the sensor module and controller mounted on the robot body; The sensor module detects at least whether there is a cliff within a first predetermined range and a second predetermined range around the robot body, wherein the radius of the first predetermined range is longer than that of the second predetermined range; The controller identifies the risk of fall based on the detection results of the sensor module; and controls the motion state of the robot body based on the risk of fall, wherein the motion state includes at least one of motion speed and motion direction.

[0007] Furthermore, the sensor module includes a first distance sensor and a second distance sensor; the first distance sensor determines whether there is a cliff within a first predetermined range around the robot body based on a first distance detected from an entity at a first predetermined distance from the robot body; the second distance sensor determines whether there is a cliff within a second predetermined range around the robot body based on a second distance detected from an entity at a second predetermined distance from the robot body; the first predetermined distance is longer than the second predetermined distance.

[0008] Furthermore, the controller identifies the risk of fall based on the detection results of the sensor module according to the following configuration: Based on the number and location of cliffs within a first and second predetermined range around the robot body detected by the sensor module, the fall risk level is identified.

[0009] Furthermore, the controller identifies the fall risk level based on the following configuration: If there are cliffs within both the first and second predetermined ranges around the robot body, it is determined to be an extremely high risk of falling. If there are no cliffs within the first and second predetermined ranges around the robot body, it is determined that there is no risk of falling. If a cliff exists only within a first predetermined range around the robot body, the fall risk level is determined based on the width or length of the cliff that is identified, and the fall risk level is positively correlated with the width of the cliff. If there are cliffs only within a second predetermined range around the robot body, the fall risk level is determined based on the location and number of cliffs.

[0010] Furthermore, the sensor module detects whether a cliff exists within a first predetermined range and a second predetermined range around the robot body according to the following configuration: The corresponding fall risk thresholds are determined based on the first distance and the second distance detected in history, respectively. If the first or second distance detected in real time changes abruptly relative to the corresponding fall risk threshold, it is identified as a candidate cliff. Determine the authenticity of the candidate cliff.

[0011] Furthermore, the sensor module determines the corresponding fall risk thresholds based on the historically detected first distance and second distance, respectively, according to the following configuration: Preprocess the first and second distances from historical detections; The ground model is fitted based on the preprocessed first distance and second distance respectively to obtain the corresponding multi-terrain distance curves; The first distance and the second distance under flat terrain are selected from the multi-terrain distance curves respectively, and are used as the first fall risk threshold corresponding to the first distance and the second fall risk threshold corresponding to the second distance.

[0012] Furthermore, the sensor module preprocesses the historically detected first distance and second distance according to the following configuration: Invalid ranging points in the first distance and second distance data detected in history are removed respectively; Then, isolated noise points in the first distance and the second distance detected in the history are filtered out respectively.

[0013] Furthermore, the sensor module determines the authenticity of the candidate cliff based on the following configuration: The detected length and / or detected width of the candidate cliff are identified. When the detected width and / or detected length of the candidate cliff reach a preset length, the candidate cliff is determined to be a real cliff.

[0014] In a second aspect, the present invention also proposes a method for preventing robots from falling, comprising: The system detects whether there is a cliff within a first predetermined range and a second predetermined range around the robot body, wherein the radius of the first predetermined range is longer than that of the second predetermined range. The risk of falling is identified based on the detection results of the sensor module; The movement state of the robot body is controlled based on the fall risk, and the movement state includes at least one of movement speed and movement direction.

[0015] In a third aspect, the present invention also proposes a robot fall prevention device, which includes a processor and a storage medium; the storage medium stores a computer program, and the processor runs the computer program to execute the above-described robot fall prevention method.

[0016] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This application incorporates sensor modules on the mobile robot body to detect the presence of cliffs (deep depressions) within a first and second predetermined range around the robot. The cliff detection within the first predetermined range is used to identify cliffs at a distance during normal operation, allowing sufficient braking distance. The cliff detection within the second predetermined range is used to determine whether the robot initially stands on the edge of a cliff or is running close to it, enabling precise close-range protection from the front or sides. Through simultaneous detection at both near and far points, cliff recognition and fall protection are achieved across the entire scene. Attached Figure Description

[0017] The present invention will be described by way of example and with reference to the accompanying drawings, wherein: Figure 1 This is a structural diagram of a fall-prevention robot in one embodiment.

[0018] Figure 2 This is a diagram showing the installation structure of a sensor module in one embodiment.

[0019] Figure 3 This is a diagram showing the structural relationship between the sensor module and the robot body.

[0020] Figure 4 This is a structural diagram of a fall-resistant robot using an arc-beam laser sensor module in one embodiment.

[0021] Figure 5 This is a structural diagram of a fall-resistant robot using a ray-shaped laser sensor module in one embodiment.

[0022] Figure 6 This is a structural diagram of a fall-resistant robot using a ray-shaped laser sensor module in another embodiment.

[0023] Figure 7 This is a flowchart illustrating the implementation of the cliff recognition method in one embodiment.

[0024] Figure 8 This is a schematic diagram illustrating the detected width of the cliff.

[0025] Figure 9 This is a schematic diagram showing the detected length of the cliff viewed from the side.

[0026] Figure 10 This is a schematic diagram showing the detected length of a cliff viewed from above.

[0027] Figure 11 This is a flowchart illustrating the implementation of a robot fall prevention method in one embodiment.

[0028] In the diagram, 1-robot body, 2-first distance sensor, 3-second distance sensor, 4-drive module, 5-candidate cliff, L1-first predetermined distance, L2-second predetermined distance, d1-first distance, d2-second distance, h1-installation height of the first distance sensor, h2-installation height of the second distance sensor. - The installation angle of the first distance sensor, - The installation angle of the second distance sensor, l - The length of the candidate cliffs that have been detected -Total length of candidate cliffs w - The width of the candidate cliff that has been detected - Total width of the candidate cliff, sq1 - Arc-shaped edge of the first predetermined region, sq2 - Arc-shaped edge of the detection area directly in front of the second predetermined region, sq3 - Arc-shaped edge of the detection area on the left side of the second predetermined region, sq4 - Arc-shaped edge of the detection area on the right side of the second predetermined region, 21 - First laser beam, 31 - Second laser beam, 32 - Third laser beam, 33 - Fourth laser beam. - Detection angle of the left detection area - Detection angle of the detection area directly in front - Detection angle of the right detection area. Detailed Implementation

[0029] All features disclosed in this specification, or all steps in all disclosed methods or processes, may be combined in any way, except for mutually exclusive features and / or steps.

[0030] Any feature disclosed in this specification (including any appended claims and abstract) may be replaced by other equivalent or similar features, unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is merely one example of a series of equivalent or similar features.

[0031] Known robot fall prevention technologies mostly employ single-point infrared sensors, ultrasonic sensors, or single LiDAR solutions, which have significant limitations: infrared / ultrasonic sensors have short detection ranges and weak anti-interference capabilities, failing to meet the early warning requirements of high-speed scenarios; single LiDAR, if focused on short-range detection, cannot cover long-range early warning, and if focused on long-range detection, it is insufficient in protecting against near cliffs and the start-up phase, resulting in the robot still facing the risk of falling in complex scenarios.

[0032] To address the aforementioned issues, this application proposes a fall-prevention robot that achieves cliff recognition and fall prevention protection across all scenarios through complementary detection at both near and far ends and a hierarchical control strategy.

[0033] like Figure 1 , Figure 2 As shown, the fall-prevention robot designed in this embodiment includes a robot body 1, and a sensor module and a controller mounted on the robot body 1. Furthermore, a drive module 4, such as wheels or tracks, is also provided at the bottom of the robot body 1 to enable movement. For example, the sensor module is mounted at the front end of the robot body 1 to obtain a wide field of view and prevent obstruction by the robot body; the controller is installed inside the robot body 1. The entire robot is powered by a power module inside the robot body 1.

[0034] like Figure 2As shown, the sensor module detects at least whether there is a cliff within a first predetermined range and a second predetermined range around the robot body 1; wherein the radius of the first predetermined range is longer than that of the second predetermined range. That is, the first predetermined range corresponds to the detection of a distant cliff, providing sufficient braking distance for the robot in scenarios where it moves normally; the second predetermined range corresponds to the detection of a near cliff, providing protection against the risk of fall when the robot starts, in scenarios where the robot is initially located on the edge of a cliff. In this way, by simultaneously detecting distant and near cliffs, cliff recognition is achieved in all scenarios, and combined with the controller's control of the robot body 1's motion state, fall protection is achieved in all scenarios.

[0035] Specifically, the controller identifies fall risk based on the detection results of the sensor module (including detection results of cliffs within a first predetermined range and detection results of cliffs within a second predetermined range). The controller then controls the motion state of the robot body 1 based on the fall risk. Fall risk is categorized into different levels, with different response measures for different levels of fall risk. That is, the controller switches or controls the robot body 1 to different motion states for different fall risk levels. The motion state includes at least one of motion speed and motion direction. The controller achieves the switching / maintenance of the robot body 1's motion state by controlling the drive module 4 in motion speed and / or motion direction.

[0036] It should be noted that the first and second predetermined ranges detected by the sensor module are fixed, meaning they remain relatively stationary with respect to the robot body 1. Movement of the robot body 1 will not change the sensor module's viewing angle. Figure 3 As shown, even if the robot body 1 moves forward, the detection angle of the sensor module will not change, thus ensuring the reliability of the robot's operation.

[0037] As an optional implementation method, such as Figure 2 As shown, the sensor module includes a first distance sensor 2 and a second distance sensor 3. The first distance sensor 2 determines whether a cliff exists within a first predetermined range around the robot body 1 based on a first distance d1 detected from an entity at a first predetermined distance from the robot body 1. The second distance sensor 3 determines whether a cliff exists within a second predetermined range around the robot body 1 based on a second distance d2 detected from an entity at a second predetermined distance from the robot body 1. The radius of the first predetermined range is determined by the first predetermined distance, and the radius of the second predetermined range is determined by the second predetermined distance. Corresponding to the relationship between the radii of the first and second predetermined ranges, the first predetermined distance L1 is longer than the second predetermined distance L2, and correspondingly, the first distance d1 will also be longer than the second distance d2.

[0038] like Figure 2As shown, assuming the first predetermined distance is denoted as L1 and the second predetermined distance is denoted as L2, based on the purpose of detecting the distant and near cliffs, the length of the first predetermined distance L1 needs to satisfy the following condition: ; in, This refers to the first safe distance set for the robot's normal movement, and it consists of three parts: ; in, The distance to be reacted represents the distance the robot moves within the time it takes for the first distance sensor 2 to detect the distant cliff (i.e., the existence of a cliff within a first predetermined range). Braking distance represents the distance the robot moves from the moment the controller begins braking until the robot comes to a complete stop. This is a safety distance set to account for potential movement errors of the robot. This safety distance can be set according to actual needs or omitted.

[0039] The reaction distance involved in the first safety distance It is determined by the robot's maximum speed. and reaction time A decision is expressed as: .

[0040] Braking distance Then it is determined by the robot's maximum speed and acceleration Make an estimate: .

[0041] For example, the inherent parameters of the robot are shown in Table 1.

[0042] Table 1 Robot Inherent Parameters

[0043] Based on the parameters shown in Table 1, the following can be calculated: Assuming the set safety distance is 0.1m, the radius of the first predetermined range must be at least 0.7m. That is, the first distance sensor 2 needs to detect the distance of an entity 10.7m away from the robot body in order to ensure that the robot can brake safely without falling after detecting the distant cliff.

[0044] Based on the determined first predetermined distance L1, the installation angle of the first distance sensor can be determined according to the installation height h1 of the first distance sensor. .

[0045] For example, assuming the first predetermined distance L1 is set to L1 = 0.7m, and the installation height h1 of the first distance sensor is 0.15m, then as follows: Figure 2 As shown, the installation angle of the first distance sensor for: .

[0046] Based on this, the first distance d1 detected by the first distance sensor 2 when the robot is on flat terrain can be further determined: .

[0047] Substituting the above parameters, the first distance d1 detected by the first distance sensor 2 under flat terrain is approximately 0.72m.

[0048] For the second distance sensor 3, since it is designed for a scenario where the robot is on the edge of a cliff before it starts, it does not need to consider reaction distance and braking distance; it only needs to detect whether there is a cliff beyond the reserved safety distance. Therefore, the second predetermined distance L2 is set to 0.1m, meaning the detection radius of the second predetermined range is 0.1m.

[0049] Typically, sensor modules use laser sensing to detect cliffs to ensure detection accuracy and effective detection distance. In a preferred embodiment, the detection laser beams of the first distance sensor 2 and the second distance sensor 3 do not interfere with each other. Therefore, the installation height h2 of the second distance sensor is no higher than the installation height h1 of the first distance sensor. Assuming the installation height h2 of the second distance sensor is the same as the installation height h1 of the first distance sensor, i.e., h2 = 0.15m.

[0050] Using the same method, the installation angle of the second distance sensor can be determined. The angle is approximately 33°. Furthermore, the second distance detected by the second distance sensor 3 in flat terrain can be calculated as d2 = 0.18m.

[0051] For other models of mobile robots, since they may have different maximum speeds and accelerations, as well as different robot body sizes, it may be necessary to set different first predetermined distances L1 (the second predetermined distance L2 can also be designed to other values) and installation heights (h1 and h2). Therefore, the first distance d1 and the second distance d2 for detection on flat terrain may be other values, but the calculation method can still be determined by the method described above.

[0052] In one alternative implementation, the sensor module uses a linear laser array to detect cliffs. That is, by emitting a linear laser array, the distance to the ground entity is calculated based on the reflected echo, thereby determining whether the entity is a cliff.

[0053] As an optional implementation method, such as Figure 4 As shown, the first distance sensor 2 and the second distance sensor 3 detect the cliff using an arc-shaped linear laser array. The first distance sensor 2 detects a fan-shaped area with a first predetermined range of 60° to 120° (calculated from 0° on the left side of the figure), and the radius of this fan-shaped area is the first predetermined distance L1. The second distance sensor 3 detects a left-side detection area with a second predetermined range of 0° to 30° (the detection angle of the left-side detection area is expressed as...). The detection area directly in front, ranging from 60° to 120° (the detection angle of the detection area directly in front is expressed as...). ) and the right-side detection zone within the range of 150°~180° (the detection angle of the right-side detection zone is expressed as All three detection zones are fan-shaped, with a radius of the second predetermined distance L2, and are all located in front of the robot body 1 (not all directly in front). It should be noted that the detection zones listed here are only examples. The shape, radius, number, and orientation of the first and second predetermined zones can be determined as needed. Preferably, both predetermined zones include at least the directly in front orientation.

[0054] for Figure 4 In the illustrated embodiment, the first distance sensor 2 emits a first linear laser beam 21 targeting an entity at a first predetermined distance L1 in front, which covers the arcuate edge sq1 of the first predetermined area. Similarly, the second distance sensor 3 targets an entity at a second predetermined distance L2 in front, and emits arcuate second linear laser beams 31, 32, and 33 respectively, to cover the arcuate edge sq2 of the detection area directly in front of the second predetermined area, the arcuate edge sq3 of the detection area on the left side of the second predetermined area, and the arcuate edge sq4 of the detection area on the right side of the second predetermined area.

[0055] A line-beam laser is actually a laser beam composed of multiple point lasers arranged sequentially. Each point laser can detect a first distance d1 / a second distance d2, and the set of first distances d1 detected by the first distance sensor 2 is represented as follows: D1 ( d 1 ,d 2 ,…, d i ,…,d n ),in d i Indicates the first i The first distance d1 for laser detection at each point, n The total number of point lasers detected by the first distance sensor 2; the set of second distances d2 detected by the second distance sensor 3 is represented as... D2 (d 1 ,d 2 ,…,d j ,…,d m ),in d j Indicates the first j The second distance d2 for point laser detection, m This represents the total number of laser points detected by the second distance sensor 3 in a single detection area. Since the data processing method is the same for all detection areas, this explanation uses data from one detection area (first distance d1 or second distance d2) as a representative example for both the first distance sensor 2 and the second distance sensor 3. When the cliff is sufficiently deep... d i and d j It could be empty, meaning there is no reflected laser beam. In other words, each point laser can detect the existence of a cliff individually. If the first distance d1 detected by any point laser changes abruptly upwards compared to the first distance d1 detected on flat terrain, then a cliff is suspected to exist at that location. If multiple consecutive point lasers all seem to detect a cliff, then the existence of a cliff can be confirmed. If only one or a few discrete point lasers seem to detect a cliff, it may simply be a small depression or hole, which will not affect the robot's normal passage.

[0056] It should be noted that, Figure 4 In the illustrated embodiment, each arc-shaped laser beam can be replaced with a straight laser beam, thus forming... Figure 4 The string of a medium-arc laser beam. Similarly, the abrupt change in the detected first distance d1 or second distance d2 can be used to determine whether a cliff has been detected. The only difference is that the first distance d1 or second distance d2 detected by the laser at different points on flat terrain is not entirely the same, requiring separate calculation / fitting.

[0057] In another alternative implementation, such as Figure 5 As shown, the first distance sensor 2 and the second distance sensor 3 can detect cliffs using multiple laser beams, respectively. This differs from... Figure 4 The arc-shaped laser beams shown indicate that the first distance sensor 2 and the second distance sensor 3 can respectively cover a first predetermined area and a second predetermined area using multiple ray-shaped laser beams. A ray-shaped laser beam is a linear laser beam arranged radially along a fan-shaped area. The length / authenticity of the detected cliff is determined by using multiple adjacent ray-shaped laser beams. For example, if multiple adjacent laser beams all detect a suspected cliff, it indicates that the cliff truly exists.

[0058] In one implementation using a ray-shaped laser beam to detect cliffs, the first distance sensor 2 and the second distance sensor 3 can respectively emit laser beams to cover the arc-shaped edge sq1 of the first predetermined area and the arc-shaped edge sq2 of the detection area directly in front of the second predetermined area, the arc-shaped edge sq3 of the left detection area of ​​the second predetermined area, and the arc-shaped edge sq4 of the right detection area of ​​the second predetermined area. In another feasible implementation, such as... Figure 6 As shown, for the area directly in front of the robot body 1, the same set of ray-shaped laser beams can simultaneously cover the arc-shaped edge sq1 of the first predetermined area and the arc-shaped edge sq2 of the front detection area of ​​the second predetermined area, while the covering laser beams of the arc-shaped edge sq3 of the left detection area of ​​the second predetermined area and the arc-shaped edge sq4 of the right detection area of ​​the second predetermined area remain unchanged.

[0059] In the embodiments of this application, Figure 4 Taking the implementation method as an example, the post-processing of the first distance d1 and the second distance d2 detected by the designed sensor module is explained. Figure 5 The first distance d1 and the second distance d2 detected in the embodiment shown can be post-processed in the same way.

[0060] For the first distance set D1 detected by the first distance sensor 2 and the second distance set D2 detected by the second distance sensor 3, in an optional implementation, such as Figure 7 As shown, the sensor module detects whether there is a cliff within a first predetermined range and a second predetermined range around the robot body 1 according to the following configuration: (1) Determine the corresponding fall risk thresholds based on the first distance d1 and the second distance d2 of the historical detection.

[0061] In one alternative implementation, this stage determines the fall risk threshold using the following configuration: 1.1) Preprocess the first distance d1 and the second distance d2 of the historical detection.

[0062] The preprocessing of historical exploration data can be achieved through the following methods: Invalid ranging points in the first distance d1 and the second distance d2 of the historical detection are removed respectively. Invalid ranging points are, for example, points in D1 and D2 that exceed the ranging range, have obvious data anomalies, or have insufficient signal strength (i.e., d i or d j ).

[0063] For example, the first and second predetermined areas are collectively referred to as the effective detection area. The distance value (first distance d1 for the first predetermined area, second distance d2 for the second predetermined area) and signal strength corresponding to each detection zone within the effective area are obtained. Ranging points with abnormal distance values ​​(distance values ​​that are too small or too large compared to the set normal distance range) and abnormal signal strength (signal strength lower than the set strength, possibly due to dust or light interference causing abnormal echoes) are filtered out. There are usually multiple such ranging points. The distance values ​​and signal strengths measured by lasers at multiple (e.g., 3-5) points adjacent to each ranging point are determined to be highly consistent, i.e., their differences are within a predetermined range. If they are highly consistent, the ranging point is determined to be a valid ranging point; otherwise, it is determined to be an invalid ranging point, and all ranging data (first distance d1 or second distance d2) from invalid ranging points are deleted.

[0064] After removing invalid ranging points, isolated noise points in the first distance d1 and the second distance d2 of the historical measurements are then filtered out. The method for filtering isolated noise points is the same for both the first distance d1 and the second distance d2, therefore a unified explanation is provided here.

[0065] Isolated noise points refer to situations where the ranging point itself is a valid point (with echo, within the range, and with acceptable signal strength), but due to environmental interference (such as small stones on the ground, laser reflection spot shift, etc.), the distance value of a single ranging point suddenly deviates from the normal range of most surrounding ranging points; that is, the distance value of a single ranging point undergoes a sudden change. These types of ranging points do not represent actual ground distance changes, and if left unaddressed, they can lead to misjudgments in cliff detection (for example, mistaking a small stone for a cliff).

[0066] In one alternative implementation, a median filtering algorithm is used to filter out isolated noise points in the ranging data of each detection area after invalid ranging points have been removed.

[0067] The median filtering algorithm works by replacing the distance value of an isolated noise point with the median of its neighborhood distances. Specifically, the isolated noise removal algorithm includes: Set the neighborhood window size, for example, the window contains 3 distance measurement points, i.e., window = "current distance measurement point - 1, current distance measurement point, current distance measurement point + 1", to avoid the ground trend distortion caused by an excessively large window. For distance measurement data that excludes invalid distance measurement points, traverse each distance measurement point according to the order of horizontal angles (e.g., from 60° to 120° within a first predetermined range). For each traversed distance measurement point, extract all valid distance values ​​within its neighborhood window and calculate the median. If the deviation between the distance value of the currently traversed distance point and the neighborhood median exceeds a threshold (e.g., 0.05m, i.e., ±5cm), then the currently traversed distance measurement point is determined to be an isolated noise point, and the distance value of the current distance measurement point is replaced with the neighborhood median; if the current distance measurement point is an edge point (without left or right neighbors, such as the first and last distance measurement points), then the original distance value is retained.

[0068] 1.2) Fit the ground model according to the first distance d1 and the second distance d2 of the preprocessing to obtain the corresponding multi-terrain distance curves.

[0069] For each detection area within the first and second predetermined ranges, a set of preprocessed ranging data will be obtained. This data will then be fitted to a ground model using algorithms such as RANSAC (Random Sample Consensus), yielding the desired multi-terrain distance curves for each detection area. The RANSAC algorithm is an interference-resistant algorithm capable of accurately identifying data patterns. Its core logic is to ignore a small number of residual interference points in a dataset (such as minor noise not removed after preprocessing or distance deviations caused by small stones on the ground) and find the pattern that best represents the overall trend of the data. The so-called "multi-terrain" refers to continuous flat land, slopes, depressions, and other uneven terrain; of course, it could also be entirely flat land.

[0070] 1.3) Select the first distance d1 and the second distance d2 under flat terrain from the multi-terrain distance curves respectively, and use them as the first fall risk threshold corresponding to the first distance d1 and the second fall risk threshold corresponding to the second distance d2.

[0071] For the fitted multi-terrain curves, a uniform pattern of distance values ​​under flat terrain is identified, namely the first distance d1 or the second distance d2 under flat terrain, forming a standard reference baseline. The found first distance d1 is used as the first fall risk threshold for the detection cliff in the corresponding detection area of ​​the first predetermined area, and the found second distance d2 is used as the second fall risk threshold for the detection cliff in the corresponding detection area of ​​the second predetermined area. Based on the previous example, the first fall risk threshold for the first predetermined area is approximately 0.72m, and the second fall risk threshold for the second predetermined area is approximately 0.18m, which is comparable to the theoretically calculated first distance d1 and second distance d2 for detecting flat terrain.

[0072] (2) Identify whether the first distance d1 or the second distance d2 detected in real time has changed abruptly relative to the corresponding fall risk threshold. If so, it is identified as a candidate cliff 5.

[0073] Specifically, determine whether the first distance d1 detected in real time satisfies the following condition: 2.1) Whether the first distance d1 is significantly increased relative to the first fall risk threshold, for example, exceeding the first fall risk threshold + safe range, or whether the distance measurement value disappears (i.e. the cliff is too deep and the echo disappears); if so, it is identified as candidate cliff 5. 2.2) Whether the first distance d1 is within the safe range relative to the first fall risk threshold; if so, it is judged as normal terrain. 2.3) Does the first distance d1 decrease relative to the first fall risk threshold? If so, it is determined that there is an obstacle.

[0074] Points 2.1) and 2.2) above are the criteria for determining whether a candidate cliff 5 exists. Point 2.3) indicates that this application can identify obstacles in addition to cliffs. The response measures for obstacles can be equivalent to those for cliffs. That is, in terms of controller response, obstacles can be regarded as cliffs and the corresponding motion state of the robot body 1 can be controlled.

[0075] (3) Determine the authenticity of candidate cliff 5.

[0076] Since the ground is scanned using a linear laser beam, which detects the existence of cliffs by using multiple laser beams to detect the presence of cliffs, small holes or similar features may be identified as candidate cliffs 5. However, such terrain features do not actually affect the robot's normal passage, so these candidate cliffs 5 can be ignored and there is no need to slow down or brake.

[0077] As an optional implementation method, such as Figure 8 , Figure 9 As shown, the detected width of the candidate cliff is identified. w (That is, the width continuously detected by the robot during its movement; the total width of the candidate cliff is expressed as...) , ) and / or the length of the candidate cliffs that have been detected l (That is, the length continuously detected by the robot during its movement; the total length of the candidate cliff is expressed as...) , The authenticity of candidate cliff 5 is determined by identifying whether its width, length, or both meet the specified conditions. Only when a candidate cliff is identified as true is the risk of fall further assessed. This is done by recognizing the width of the candidate cliff that has already been detected. w and / or the length of the candidate cliffs already detected lThe width of the candidate cliff has been detected w The preset width is reached, and / or the length of the candidate cliff has been detected. l When the preset length is reached, candidate cliff 5 is determined to be a real cliff.

[0078] Specifically, the robot uses multiple laser beams to detect the cliff in each detection area, such as... Figure 8 As shown, candidate cliff 5 is only determined to be a real cliff when it is detected by multiple consecutive ranging points; or, as... Figure 9 , Figure 10 As shown, candidate cliff 5 is only identified as a real cliff when it is detected in multiple consecutive frames (time frames) at the same location. That is, candidate cliff 5 is identified as a real cliff only if its width or length in the detection area is sufficient. For example, if there are K consecutive distance values ​​(e.g., ...) among M distance values ​​in the detection area... All of them are determined to be candidate cliff 5; or, for M distance values ​​in the detection area, K consecutive distance values ​​appear in multiple consecutive frames (e.g., T represents the number of time frames. If all values ​​in the time step (where i is the frame index) are identified as candidate cliff 5, then candidate cliff 5 is a real cliff; or, in the T frames of continuous detection, among the M distance values ​​in each frame's detection area, at least K consecutive distance values ​​(e.g., ...) are identified as candidate cliff 5. If all distance values ​​are identified as candidate cliff 5, and these distance values ​​are identified as candidate cliff 5 in consecutive T frames, then candidate cliff 5 is a real cliff. The controller needs to respond to the real cliff.

[0079] As mentioned earlier, the controller controls the motion state of the robot body 1 based on the identified fall risk level. In one optional implementation, the controller identifies the fall risk based on the detection results of the sensor module, according to the following configuration: Based on the number and location of cliffs within a first and second predetermined range around the robot body 1 detected by the sensor module, the fall risk level is identified.

[0080] Generally speaking, the more concentrated the cliffs are in front of the robot body 1, the higher the risk of falling. The more cliffs there are, the higher the risk of falling.

[0081] In one alternative implementation, the controller identifies the fall risk level based on the following configuration: (1) If there are cliffs in both the first and second predetermined ranges around the robot body 1, it is determined to be an extremely high risk of falling.

[0082] Taking the example where the first predetermined range includes one detection zone and the second predetermined range includes three detection zones, if a cliff is detected within the first predetermined range, regardless of which detection zone within the second predetermined range detects the cliff, it is considered an extremely high risk of fall. Under extremely high fall risk conditions, the controller typically needs to apply emergency braking to prevent a fall.

[0083] (2) If there are no cliffs within the first and second predetermined ranges around the robot body 1, it is determined that there is no risk of falling.

[0084] In contrast to (1), if no cliff is detected in either the first or the second predetermined range, it means that there is no cliff at the near end or the far end, and there is no risk of falling. In this case, it is sufficient to maintain the status quo (normal driving).

[0085] (3) If there is a cliff only within a first predetermined range around the robot body 1, the fall risk level is determined based on the width or length of the cliff that is identified. The fall risk level is positively correlated with the width or length of the cliff.

[0086] Since it was detected within a predetermined range, it constitutes a pre-judgment. When the cliff (referred to as the candidate cliff) was initially detected, the width of the candidate cliff was already detected. w Typically shorter, the length of the candidate cliffs that have been detected. l The value is usually 0, so immediate emergency braking is not required. However, if the robot continues to move towards the cliff, then the detected width of the identified candidate cliff will be less than the actual cliff width. w and the length of candidate cliffs already detected l If the pressure continues to increase, emergency braking should be considered to avoid plunging into the cliff.

[0087] Specifically, when a cliff is detected for the first time within the first predetermined range, i.e., in the first frame, it is determined to be of medium fall risk. At this time, the controller only needs to control the robot body 1 to decelerate and can adjust its direction to deviate from the cliff direction. If, after this, the width of the detected cliff increases to a certain extent, or if a cliff is still detected for multiple consecutive frames (e.g., T frames, where T is a positive integer greater than 1, such as 5 or other values), it indicates that the robot is moving towards the cliff. At this time, the controller upgrades the fall risk level, determining it to be of high or extremely high fall risk, and can control the robot body 1 to brake suddenly.

[0088] (4) If there are cliffs only within the second predetermined range around the robot body 1, the fall risk level is determined based on the location and number of cliffs.

[0089] As mentioned earlier, the more concentrated the cliffs are directly in front of the user, or the more cliffs there are, the higher the risk of falling. Taking the second predetermined range, which includes three detection zones—left, front, and right—as an example, if a cliff is detected only in the left or right detection zone, it is considered a medium risk of falling. The controller will then control the robot body 1 to decelerate and, in conjunction with directional control, shift towards the side where no cliff was detected. If a cliff is detected in both the left and right detection zones, or in the front detection zone, it is considered an extremely high risk of falling, and the controller will apply emergency braking to the robot body 1.

[0090] In other words, the motion states of robot body 1 controlled by the controller based on the fall risk level include: (1) For extremely high risk of falling, immediately apply emergency braking to robot body 1, i.e., brake at maximum acceleration. This is a hard protection measure that must be executed unconditionally. Alternatively, robot body 1 can be controlled to retreat a safe distance of 0.1m to 0.2m.

[0091] (2) For high risk of falling, immediately apply emergency braking to robot body 1. Optionally, robot body 1 can also be controlled to move backward appropriately.

[0092] (3) For moderate fall risk, control the robot body 1 to decelerate proportionally, with the acceleration remaining constant or gradually increasing. Path replanning can also be initiated to avoid cliffs.

[0093] (4) For cases where there is no risk of falling, control the robot body 1 to maintain its current motion state.

[0094] Furthermore, each of the aforementioned drop risk levels can be further subdivided into smaller drop risk sub-levels. For example, in the extremely high drop risk level, there are further sub-levels of high drop risk and even higher extremely high drop risk. The controller can control the same or different motion states for each drop risk level across its various drop risk sub-levels, depending on the specific circumstances. Additionally, the controller can issue different levels of alarms in response to different drop risk levels / sub-levels.

[0095] The fall-prevention robot designed in this application achieves the following technical effects through a complementary design of remote and near-end detection, precise parameter calculation, and a hierarchical control strategy: 1. Full-scene cliff protection coverage: Solves the cliff recognition problem in multiple scenarios such as startup (no memory dependency), high-speed driving (long-range early warning), and low-speed edge contact (close-range precise protection), eliminating detection blind spots; 2. High reliability and safety: The false detection rate is reduced through cross-verification at near and far ends, detection of continuous lateral areas, and multi-frame confirmation mechanisms; the safety distance calculation with redundant design ensures that the robot can still stop completely under extreme working conditions without the risk of falling. 3. Smooth motion is ensured: A graded control strategy (progressive deceleration and emergency braking) is adopted to avoid motion shock caused by a single braking mode and improve the stability of robot operation; 4. High adaptability: The sensor module installation parameters are based on a universal chassis structure design, and can be adapted to different specifications of wheeled robots by adjusting the installation height and tilt angle, resulting in high compatibility.

[0096] Based on the ideas of this application, an anti-fall method for robots is also proposed in the embodiments of this application, such as... Figure 11 As shown, it includes the following steps: S1. Detect whether there is a cliff within at least a first predetermined range and a second predetermined range around the robot body 1. The radius of the first predetermined range is longer than that of the second predetermined range.

[0097] S2. Identify drop risks based on sensor module detection results.

[0098] S3. The motion state of the robot body 1 based on fall risk control. The motion state includes at least one of motion speed and motion direction.

[0099] In the embodiments of the robot fall prevention method, the optional implementation methods for each step can be referred to in the design of the optional implementation methods for the sensor module or controller in the fall prevention robot above, and will not be repeated here.

[0100] Furthermore, based on the concept of this application, an embodiment of this application also provides a robot fall prevention device, which includes a processor and a storage medium. The storage medium stores a computer program, and the processor runs the computer program to execute the aforementioned robot fall prevention method.

[0101] This invention is not limited to the specific embodiments described above. The invention extends to any new feature or combination disclosed in this specification, as well as any new method or process step or combination disclosed herein.

Claims

1. A fall-prevention robot for cliff recognition and fall protection in all scenarios; characterized in that, include: The robot body (1), and the sensor module and controller disposed on the robot body (1); The sensor module simultaneously detects at least two predetermined ranges around the robot body (1) to determine whether there is a cliff. The radius of the first predetermined range is longer than that of the second predetermined range. The radius of the first predetermined range is determined by a first predetermined distance (L1), and the radius of the second predetermined range is determined by a second predetermined distance (L2). The first predetermined distance (L1) is above a first safety distance set for normal robot movement. The first safety distance consists of a reaction distance, a braking distance, and a safety distance. The second predetermined distance (L2) is a safety distance reserved to determine whether the robot is on the edge of a cliff before starting. The first predetermined distance (L1) is longer than the second predetermined distance (L2). The sensor module includes a first distance sensor (2) and a second distance sensor (3); the detection laser beams of the first distance sensor (2) and the second distance sensor (3) do not interfere with each other; The first distance sensor (2) determines whether there is a cliff within a first predetermined range around the robot body (1) based on the first distance (d1) of the detected entity at a first predetermined distance (L1) from the robot body (1); the second distance sensor (3) determines whether there is a cliff within a second predetermined range around the robot body (1) based on the second distance (d2) of the detected entity at a second predetermined distance (L2) from the robot body (1); the controller identifies the number and location of cliffs within the first and second predetermined ranges around the robot body (1) detected by the sensor module. Fall risk level: If there are cliffs in both the first and second predetermined ranges around the robot body (1), it is determined to be an extremely high fall risk; if there are no cliffs in either the first or second predetermined ranges around the robot body (1), it is determined to be no fall risk; if there are cliffs only in the first predetermined range around the robot body (1), the fall risk level is determined based on the width or length of the cliffs that are identified, and the fall risk level is positively correlated with the width of the cliffs; if there are cliffs only in the second predetermined range around the robot body (1), the fall risk level is determined based on the location and number of the cliffs. Based on different fall risk levels, the robot body (1) is switched or controlled to be in different motion states, the motion states including at least one of motion speed and motion direction.

2. The fall-resistant robot as described in claim 1, characterized in that, The sensor module detects whether there is a cliff within a first predetermined range and a second predetermined range around the robot body (1) according to the following configuration: The corresponding fall risk thresholds are determined based on the first distance (d1) and the second distance (d2) obtained from historical detection, respectively; If the first distance (d1) or the second distance (d2) detected in real time changes abruptly relative to the corresponding fall risk threshold, it is identified as a candidate cliff (5). Determine the authenticity of the candidate cliff (5).

3. The fall-resistant robot as described in claim 2, characterized in that, The sensor module determines the corresponding fall risk thresholds based on the first distance (d1) and the second distance (d2) detected historically, according to the following configuration: Preprocess the first distance (d1) and the second distance (d2) of the historical detection. The ground model is fitted based on the preprocessed first distance (d1) and second distance (d2) to obtain the corresponding multi-terrain distance curves; The first distance (d1) and the second distance (d2) under flat terrain are selected from the multi-terrain distance curves respectively, and are used as the first fall risk threshold corresponding to the first distance (d1) and the second fall risk threshold corresponding to the second distance (d2).

4. The fall-resistant robot as described in claim 3, characterized in that, The sensor module preprocesses the historically detected first distance (d1) and second distance (d2) according to the following configuration: Invalid ranging points in the first distance (d1) and the second distance (d2) from historical detections are removed respectively; Then, isolated noise points in the first distance (d1) and the second distance (d2) of the historical detection are filtered out respectively.

5. The fall-resistant robot as described in claim 2, characterized in that, The sensor module determines the authenticity of the candidate cliff (5) based on the following configuration: Identify the width of the candidate cliff that has been detected ( w ) and / or the length of the candidate cliffs that have been detected ( l The width of the candidate cliff that has been detected ( w ) and / or the length of the candidate cliffs that have been detected ( l When the preset length is reached, the candidate cliff (5) is determined to be a real cliff.

6. A robot fall prevention method for achieving cliff recognition and fall protection in all scenarios, characterized in that, include: The sensor module simultaneously detects whether there is a cliff within a first predetermined range and a second predetermined range around the robot body (1). The radius of the first predetermined range is longer than that of the second predetermined range. The radius of the first predetermined range is determined by a first predetermined distance (L1), and the radius of the second predetermined range is determined by a second predetermined distance (L2). The first predetermined distance (L1) is above a first safety distance set for normal robot movement. The first safety distance consists of three parts: reaction distance, braking distance, and safety distance. The second predetermined distance (L2) is a safety distance reserved to determine whether the robot is on the edge of a cliff before starting. The first predetermined distance (L1) is longer than the second predetermined distance (L2). The sensor module includes a first distance sensor (2) and a second distance sensor (3); the detection laser beams of the first distance sensor (2) and the second distance sensor (3) do not interfere with each other; The first distance sensor (2) determines whether there is a cliff within a first predetermined range around the robot body (1) based on the first distance (d1) of the entity at a first predetermined distance (L1) of the robot body (1); the second distance sensor (3) determines whether there is a cliff within a second predetermined range around the robot body (1) based on the second distance (d2) of the entity at a second predetermined distance (L2) of the robot body (1). Based on the number and location of cliffs within a first and second predetermined range around the robot body (1) detected by the sensor module, the fall risk level is identified: if cliffs exist within both the first and second predetermined ranges around the robot body (1), it is determined to be an extremely high fall risk; if no cliffs exist within either the first or second predetermined ranges around the robot body (1), it is determined to be a no fall risk; if only cliffs exist within the first predetermined range around the robot body (1), the fall risk level is determined based on the width or length of the cliffs identified, and the fall risk level is positively correlated with the width of the cliffs; if only cliffs exist within the second predetermined range around the robot body (1), the fall risk level is determined based on the location and number of cliffs. Based on different fall risk levels, the robot body (1) is switched or controlled to be in different motion states, the motion states including at least one of motion speed and motion direction.

7. A robot fall prevention device, characterized in that, It includes a processor and a storage medium; the storage medium stores a computer program, and the processor runs the computer program to perform the robot fall prevention method as described in claim 6.

Images