Robot precision landing planning method and device based on prior terrain model

Abstract

The embodiment of the application provides a robot precise landing planning method and device based on a priori terrain model, which carries out digital modeling on a preset terrain in a simulation environment, constructs a digital map containing terrain geometry and semantic information, and clearly calibrates the center coordinates of each ideal landing area as a target landing point. In the training process, the global coordinates of the robot foot end are tracked in real time, and the three-dimensional deviation of the robot foot end from the nearest target landing point is determined through map mapping, and then a reward function with the deviation size as the evaluation index is designed to drive the robot to learn a high-precision landing strategy. The present application introduces the priori terrain target point as a guide, so that the reward signal has a clear geometric meaning and can directly and efficiently drive the strategy to converge to the precise landing behavior, greatly improving the motion accuracy and reliability on the structured terrain.

Description

Technical Field

[0001] This application relates to the field of data processing, specifically to a method and apparatus for precise robot landing planning based on a priori terrain model. Background Technology

[0002] In the field of robot control technology, motion control of legged and wheeled robots, especially adaptive traversal in complex and unknown terrain, has always been a key research focus and challenge. In recent years, reinforcement learning-based methods have shown great potential in this field. By enabling robots to learn autonomously in simulated environments, their strategies can handle high-dimensional state inputs and output complex motion commands.

[0003] However, existing reinforcement learning methods still have significant limitations. First, regarding the coupling of perception and action, many methods rely on simplified terrain representations or ontological information, lacking precise geometric perception of the surrounding environment, which prevents robots from achieving accurate adaptive landing. Second, in terms of adaptability to complex scenes, when faced with discrete landing points (such as bouncy stakes) or extremely complex environments with gaps and ditches, existing strategies often struggle to generate safe and efficient motion sequences, easily leading to collisions, instability, or planning failures. This is mainly because sparse reward signals are difficult to capture detailed terrain geometry features, and end-to-end strategies lack explicit constraints on landing point safety and long-range planning capabilities.

[0004] This invention aims to overcome the limitations of existing legged / wheeled robots in adapting to complex terrain, especially to achieve precise, stable, and adaptive traversal capabilities at discrete footholds (such as bouncy stakes). Summary of the Invention

[0005] To address the problems in the existing technology, this application provides a method and apparatus for precise robot landing planning based on a priori terrain model, which guides the robot to plan precise landing points, enabling the robot to achieve precise, stable, and adaptive traversal capabilities on discrete landing points.

[0006] To solve at least one of the above problems, this application provides the following technical solution:

[0007] Firstly, this application provides a method for precise robot landing planning based on a priori terrain model, including:

[0008] In the simulation environment, the preset terrain is gridded to construct a digital map containing terrain geometry and attributes. The digital map contains several terrain units. The digital map defines the ideal landing area of ​​each terrain unit, the target landing point within each ideal landing area, and the three-dimensional coordinates of each target landing point in a predefined world coordinate system.

[0009] The robot foot is tracked in real time in the three-dimensional coordinates of the world coordinate system and mapped onto the digital map to determine whether the robot foot lands in the ideal landing area at the end of the swing phase; if the robot foot lands in the ideal landing area, the relative positional relationship between the robot foot and the target landing point in the ideal landing area is calculated.

[0010] The robot's foot placement is set to deviate further from the target placement point, resulting in a lower reward or a heavier penalty. A robot placement reward function is then set based on this setting to guide the robot to optimize its ability to place its foot at the target placement point.

[0011] Furthermore, the step of performing gridding processing on the preset terrain to construct a digital map containing terrain geometry and attributes includes:

[0012] In a simulation environment, a 3D model of a preset terrain is obtained; the 3D model contains geometric information and preset semantic tags.

[0013] Define a fixed world coordinate system as a reference for terrain and robot pose, and determine a reference horizontal plane based on the overall orientation of the terrain;

[0014] The three-dimensional model of the preset terrain is projected into two dimensions on the reference horizontal plane to determine the two-dimensional envelope boundary of the terrain; a grid is generated within the two-dimensional envelope boundary, and each grid cell is identified by a two-dimensional index.

[0015] The grid cells are assigned attribute values, including geometric attribute assignment, semantic attribute assignment, and derived attribute assignment; the semantic attributes include the coordinates of the target landing point; the derived attributes include accessibility and normal vector.

[0016] The generated data, which includes geometry and attributes, is structured and stored to obtain the digital map.

[0017] Furthermore,

[0018] The method for assigning geometric attributes includes: projecting the mesh cell vertically upward onto the three-dimensional model of the terrain, finding the intersection with the surface of the three-dimensional model, and setting the surface height of the mesh cell based on the intersection point;

[0019] The method for assigning the semantic attributes includes: pre-defining an ideal landing area based on the semantic labels of the terrain; specially marking the grid cells within the ideal landing area; and storing the three-dimensional coordinates of the target landing point in the world coordinate system in the attributes of all associated grid cells within the ideal landing area.

[0020] The method for assigning accessibility includes: assigning a accessibility score to a grid cell based on its attribute characteristics, including slope, height difference, surface material, and coefficient of friction.

[0021] The method for assigning the normal vector includes: calculating the average normal vector of the surface where the grid cell is located as the normal vector of the grid cell.

[0022] Furthermore, methods for obtaining target landing points within each ideal landing area include:

[0023] A specific point on the horizontal plane of the grid cell is selected as the target landing point, and the specific point includes the geometric center point and the centroid point;

[0024] Alternatively, multiple candidate target landing points can be distributed within the ideal landing area, where the target landing point is the point closest to the foot landing point when the foot lands.

[0025] Methods for obtaining the three-dimensional coordinates of each target's landing point in a predefined world coordinate system include:

[0026] The position of the target's landing point on the horizontal plane of the grid cell is taken as its x and y coordinates in the world coordinate system, and the height of the grid cell surface corresponding to the target's landing point is taken as its z coordinate in the world coordinate system.

[0027] Furthermore, the step of real-time tracking the three-dimensional coordinates of the robot's foot in the world coordinate system and mapping them onto the digital map to determine whether the robot's foot lands within the ideal landing area at the end of the swing phase includes:

[0028] The robot's foot's three-dimensional coordinates in the world coordinate system are calculated in real time using the robot's forward kinematics chain and joint angle sensor data.

[0029] The three-dimensional coordinates of the robot's foot are mapped to a digital map to obtain the grid cell index corresponding to the horizontal coordinate of the foot in the digital map when the foot is at the end of the swing phase.

[0030] Use the grid cell index to query the semantic tag of the grid cell to determine whether the robot's foot lands within the ideal landing area.

[0031] Furthermore, the step of calculating the relative positional relationship between the robot's foot and the target landing point within the ideal landing area if the robot's foot lands within that area includes:

[0032] Obtain the three-dimensional coordinates of the target landing point in the world coordinate system within the ideal landing area of ​​the robot; the three-dimensional coordinates of the target landing point are stored in the attributes of each grid cell associated with the ideal landing area;

[0033] Based on the three-dimensional coordinates of the robot's foot landing point in the world coordinate system and the three-dimensional coordinates of the target landing point in the world coordinate system, the three-dimensional deviation vector between the two is calculated, which is the relative positional relationship between the robot's foot and the target landing point in the ideal landing area.

[0034] Furthermore, the step of setting the robot's foot landing point to deviate further from the target landing point, resulting in a lower reward or a heavier penalty, and setting the robot's landing reward function according to this setting, includes:

[0035] Calculate the deviation vector between the robot's foot landing point and the target landing point, and input it into the deviation evaluation function, which outputs a reward or penalty according to the settings.

[0036] The deviation evaluation functions include the squared difference function, the L1 norm penalty term, the saturated continuous reward function, the separated weighted and nonlinear processing function, and the piecewise function based on contact point classification.

[0037] Secondly, this application provides a robot precise landing planning device based on a priori terrain model, comprising:

[0038] The map building module is used to perform gridding on the preset terrain in a simulation environment to build a digital map containing terrain geometry and attributes. The digital map contains several terrain units. The digital map defines the ideal landing area of ​​each terrain unit, the target landing point in each ideal landing area, and the three-dimensional coordinates of each target landing point in a predefined world coordinate system.

[0039] The deviation calculation module is used to track the three-dimensional coordinates of the robot's foot in the world coordinate system in real time and map them onto the digital map to determine whether the robot's foot lands in the ideal landing area at the end of the swing phase; if the robot's foot lands in the ideal landing area, the module calculates the relative positional relationship between the robot's foot and the target landing point in the ideal landing area.

[0040] The function setting module is used to set the reward or penalty for the robot's foot landing point being further away from the target landing point. Based on this setting, the robot landing reward function is set to guide the robot to optimize its ability to land at the target landing point.

[0041] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the robot precise landing planning method based on a priori terrain model.

[0042] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the robot precise landing planning method based on a priori terrain model.

[0043] Fifthly, this application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the robot precise landing planning method based on a priori terrain model.

[0044] As described above, this application provides a method and apparatus for precise robot landing planning based on a priori terrain model. By digitally modeling a preset terrain in a simulation environment, a digital map containing terrain geometric and semantic information is constructed, and the center coordinates of each ideal landing area are explicitly marked as the target landing point. During training, the global coordinates of the robot's foot are tracked in real time, and the three-dimensional deviation between the deviation and the nearest target landing point is determined through map mapping. A reward function with the magnitude of the deviation as an evaluation index is then designed to drive the robot to learn a high-precision landing strategy. This invention introduces priori terrain target points as guidance, giving the reward signal a clear geometric meaning. This allows for direct and efficient convergence of the strategy to precise landing behavior, significantly improving the motion accuracy and reliability on structured terrain. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a flowchart illustrating the robot's precise landing planning method based on a priori terrain model in an embodiment of this application.

[0047] Figure 2 This is a structural diagram of the robot precise landing planning device based on a priori terrain model in an embodiment of this application.

[0048] Figure 3 This is a schematic diagram of the structure of the electronic device in the embodiments of this application.

[0049] Figure label:

[0050] Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver storage unit 9144, antenna 9111, speaker 9131, microphone 9132. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0052] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0053] In view of the problems existing in the prior art, this application provides a method and apparatus for precise robot landing planning based on a prior terrain model. By digitally modeling a preset terrain in a simulation environment, a digital map containing terrain geometric and semantic information is constructed, and the center coordinates of each ideal landing area are explicitly marked as the target landing point. During training, the global coordinates of the robot's foot are tracked in real time, and the three-dimensional deviation between the deviation and the nearest target landing point is determined through map mapping. A reward function with the magnitude of the deviation as an evaluation index is then designed to drive the robot to learn a high-precision landing strategy. This invention introduces prior terrain target points as guidance, giving the reward signal a clear geometric meaning. This allows for direct and efficient driving of the strategy to converge to precise landing behavior, significantly improving the motion accuracy and reliability on structured terrain.

[0054] To guide a robot in precise landing point planning, enabling it to achieve accurate, stable, and adaptive traversal capabilities at discrete landing points, this application provides an embodiment of a robot precise landing point planning method based on a priori terrain model. The core idea is to utilize precise prior knowledge from a simulation environment to guide the robot in accurate landing point planning, applicable to structured terrain with clearly defined objectives. See also... Figure 1 The robot precise landing planning method based on a priori terrain model specifically includes the following:

[0055] Step S101: In the simulation environment, the preset terrain is meshed to construct a digital map containing terrain geometry and attributes. The digital map contains several terrain units. The digital map defines the ideal landing area of ​​each terrain unit, the target landing point within each ideal landing area, and the three-dimensional coordinates of each target landing point in a predefined world coordinate system.

[0056] This step provides a specific method for digital environmental modeling: During the simulation training phase, the pre-set complex terrain (such as steps and stumps) is meshed to construct a digital map containing terrain geometry and attributes. This map clearly marks the ideal foothold area on each terrain unit (such as the center of the step tread or the center of the stump top), and obtains the coordinates of each ideal foothold point. .

[0057] Optionally, in this embodiment, to achieve precise guidance of the preset structured terrain, the following steps are used for digital modeling of the environment:

[0058] Step 1: Topographic Geometry and Semantic Analysis

[0059] In the simulation environment, a complete 3D model of the preset terrain is first obtained. This model contains accurate geometric information (vertices, triangular patches) and preset semantic labels (such as "step tread", "pile top", "slope surface", "no-foot zone").

[0060] Step 2: Establish a unified world coordinate system and grid reference surface

[0061] 1. Define a fixed world coordinate system as the reference benchmark for all terrain and robot poses.

[0062] 2. Determine a reference horizontal plane (usually the Z=0 plane) based on the overall orientation of the terrain (such as a horizontal surface or a main slope). All height information is relative to this reference plane.

[0063] Step 3: Two-dimensional mesh projection and partitioning

[0064] 1. Project the entire terrain onto the reference horizontal plane (XY plane) in two dimensions to determine the two-dimensional envelope boundary of the terrain.

[0065] 2. Within this two-dimensional boundary, a regular grid is created. The grid size (resolution) is a configurable key parameter that determines the level of detail in the digital map. For example, a square grid with sides ranging from 1cm to 5cm can be used.

[0066] 3. Each grid cell is identified by a unique two-dimensional index (i, j), corresponding to its center horizontal coordinate (x, j) in the world coordinate system. i , y j).

[0067] Step 4: Assigning Grid Cell Attributes (The Core of Map Construction)

[0068] The following key attributes are calculated and assigned to each grid cell (i, j) to form a multi-dimensional attribute digital map:

[0069] 1. Geometric property - Surface height Z(i, j):

[0070] Project the grid cells vertically upwards (Z-axis direction) onto the 3D terrain model and find their intersection with the model surface.

[0071] Take the highest Z-coordinate value among the intersection points (or take the intersection point of a specific surface according to semantics) as the surface height of the mesh cell.

[0072] If there is no intersection (such as in the air or in a ravine), it is marked as "invalid" or assigned a special height value (such as negative infinity).

[0073] 2. Semantic Attributes - Ideal Foot Marker and Coordinate T n :

[0074] Core step: Based on terrain semantics, predefine the "ideal landing area". For example:

[0075] For steps: the central area of ​​each step tread.

[0076] For plum blossom piles: the circular area at the top of each pile.

[0077] In digital maps, all grid cells located within these ideal landing areas are specially marked (e.g., flags). ideal =1).

[0078] At the same time, store the coordinates T of the "target landing point" in this ideal region. n =(x ideal , y ideal , z ideal ). (x ideal ,y ideal , z ideal The center point or centroid of the region on the horizontal plane is usually taken as z. ideal This represents the surface height Z corresponding to that point. An ideal region may cover multiple grids, but its T... n The coordinates are unique.

[0079] 3. Derived attributes:

[0080] Accessibility: Each grid is assigned a accessibility score based on factors such as slope, height difference, and surface material (friction coefficient).

[0081] Normal vector: Calculates the average normal vector of the ground surface where the grid cell is located, used to determine the foot posture when landing.

[0082] Step 5: Data Structure of Digital Maps

[0083] The final generated digital map is a structured dataset, which can be represented as:

[0084] A two-dimensional array (or lookup table) where each element corresponds to a grid cell, storing its (Z, flag) values. ideal ,T n Attributes such as , ...).

[0085] Or an equivalent multi-layer raster map, where each layer stores an attribute (such as a height layer, a semantic tag layer, and a target coordinate layer).

[0086] Optionally, the definition of target landing points within the ideal landing area is flexible and multi-layered, depending on the implementation method and terrain constraints. This embodiment provides two examples of target landing point selection:

[0087] 1. Basic Implementation Example: Single-Point Guidance Model

[0088] Scenario: Suitable for specific scenarios with extreme requirements for foot accuracy, such as precision assembly or stepping on small target points.

[0089] Definition: At this point, an ideal landing region is associated with only one unique ideal landing point (usually the geometric or dynamic center of the region, such as T_center).

[0090] Objective: To achieve the highest precision "point-to-point" control as the basic form of the solution.

[0091] 2. Preferred Implementation: Multi-point Ensemble Guidance Model

[0092] Scenario: Applicable to scenarios where the area has a certain size and multiple locations within the area are mechanically equivalent, such as a step tread.

[0093] Definition: At this point, an ideal landing region is associated with a set of ideal landing points {T1, T2, ..., T} consisting of multiple discrete points. k These points are evenly or regularly distributed within the area.

[0094] Mapping logic: The deviation of the foot position F from the region is defined as the deviation of F from the nearest point in the set (i.e., the minimum distance). This is equivalent to guiding the foot to fall towards any predetermined "good point" within the region.

[0095] Objective: To provide a degree of flexibility while ensuring accuracy, and to avoid excessive constraints on a single center point.

[0096] Step S102: Track the three-dimensional coordinates of the robot's foot in the world coordinate system in real time and map them onto the digital map to determine whether the robot's foot lands in the ideal landing area at the end of the swing phase; if the robot's foot lands in the ideal landing area, calculate the relative positional relationship between the robot's foot and the target landing point in the ideal landing area.

[0097] This step provides a specific method for mapping the relative position of the robot's foot to the terrain: The position of the robot's foot in the global coordinate system is tracked in real time and mapped onto this digital map, thereby obtaining the relative positional relationship between the foot and the nearest ideal landing area. The position of the robot's foot / wheel bottom in the world coordinate system is F. i = (x_foot_i, y_foot_i, z_foot_i). Where x_foot_i, y_foot_i, z_foot_i represent the positions of the bottom of the i-th foot / wheel in the global coordinate system.

[0098] In this embodiment, the core of this step lies in "real-time positioning" and "map query". Optionally, the process is as follows:

[0099] Step 1: Obtain the real-time pose of the foot in the global coordinate system

[0100] 1. Using the robot's forward kinematics chain and real-time joint angle sensor data, calculate the three-dimensional coordinates F of each foot (or wheel contact point) i in the world coordinate system at the current time t. i (t) = [x i (t), y i (t),z i (t)]. This is the standard result for the robot's internal state perception.

[0101] Step 2: Map the foot position to the digital map grid

[0102] 1. Obtain the horizontal coordinate component (x) of the foot. i (t), y i (t)).

[0103] 2. Based on the origin coordinates and grid resolution (e.g., side lengths dx, dy) of the digital map, calculate the grid cell index (m, n) corresponding to the horizontal coordinate:

[0104]

[0105] Where dx and dy are the resolutions of the elevation map grid in the x and y directions (e.g., 0.02 m / pixel), x map-origin ,y map-originIt is the coordinate of the origin of the elevation map in the world coordinate system.

[0106] `floor()` is a floor function. This step discretizes the continuous foot positions and associates them with a unique map grid cell.

[0107] Step 3: Query the coordinates of the target's landing point

[0108] 1. Using the calculated grid index (m,n), query the digital map that was pre-stored in step one (building the map).

[0109] 2. Key judgment: Check the semantic tag flag_ideal(m, n) of this grid cell:

[0110] If flag_ideal == 1 (meaning the ideal region): then directly read the coordinates T of the target landing point associated with that unit. target =T(m, n) = [x T , y T , z T Note: An ideal region corresponds to a unique T, whose coordinates are typically stored in the properties of all associated mesh cells within that region.

[0111] If flag_ideal == 0 (non-ideal region): the "precise foot landing reward" of this invention will not be activated. The robot's foot may be in a transitional region or non-target terrain, in which case other basic motion strategies can be executed.

[0112] Step 4: Calculate the relative position deviation vector

[0113] 1. Once the target point T associated with the current foot i is determined, the three-dimensional deviation vector between the two can be calculated.

[0114] d i (t)=F i (t)-T=[Δx_i(t),Δy_i(t),Δz_i(t)]

[0115] This vector d i (t) represents the real-time deviation between the current position of the foot and the ideal landing point in three-dimensional space.

[0116] Step S103: Set the reward to be lower or the penalty to be heavier as the landing point of the robot's foot deviates further from the target landing point, and set the robot landing reward function according to the setting to guide the robot to optimize its ability to land at the target landing point.

[0117] This step provides a specific method for designing a reward function for precise foot placement: the reward function is designed such that the further the robot's foot deviates from the ideal center at the end of the swing phase, the lower the reward. This mechanism acts like "precision guidance," directly driving the reinforcement learning agent to learn to achieve precise foot placement with zero or minimal deviation on specific terrain. A basic implementation example: a simplified form of the penalty term is...

[0118] penalty=(T n -F i ) 2

[0119] The aforementioned difference of squares form is merely one exemplary implementation. This approach encompasses, but is not limited to, this specific form; any implementation using T... n T i Functions whose independent variables are negatively correlated (for rewards) or positively correlated (for penalties) with the magnitude (or components) of the deviation vector are all within the scope of protection of this invention. This includes, but is not limited to, more complex function forms such as those using the L1 norm, continuous functions with saturation characteristics, or weighted separation of horizontal and vertical deviations.

[0120] Optionally, in this embodiment, applying bias (timing and logic) to the reward function includes:

[0121] 1. Application Timing: This deviation calculation and reward application only applies to the foot at the end of the swing phase (i.e., the moment the foot is about to touch the ground). It can be determined by foot speed, foot height off the ground, or gait phase signals.

[0122] 2. Reward / Penalty Generation: The calculated deviation vector d i (t) (or its modulus, components) is input into a deviation evaluation function R. precision (d i In this function, the reward (positive value) or penalty (negative value) is output as part of the overall reward function for reinforcement learning.

[0123] Optionally, in this embodiment, the formulas for several bias evaluation functions other than the squared difference (L2 norm) are as follows:

[0124] The following lists several function forms with different characteristics. They all use the deviation vector d=[Δx,Δy,Δz] as the independent variable, and the function value monotonically decreases (for rewards) or increases (for punishments) as the deviation increases. Δx, Δy, Δz: represent the positional deviations in the X (front / back), Y (left / right), and Z (up / down) directions in the world coordinate system or body coordinate system, respectively.

[0125] 1. L1 norm (sum of absolute values) penalty term

[0126] formula: , where k is a positive scaling factor.

[0127] Feature: Less sensitive to outliers than the squared difference, and tends to produce sparse solutions during optimization (i.e., it may precisely align with a certain dimension while ignoring other dimensions). The calculation is simpler.

[0128] 2. Saturated continuous reward function

[0129] Formula: or

[0130] where ||d|| is the norm of the deviation vector (L2 norm), C is the maximum reward value, and α, β are decay coefficients.

[0131] Feature: When the deviation is extremely small, the reward approaches the maximum value C; when the deviation increases, the reward smoothly decays to zero. This form is more in line with the characteristics of the reward signal in reinforcement learning, provides a smoother gradient, and is beneficial for training.

[0132] 3. Separated weighted and non - linear processing function

[0133] Formula:

[0134] where w x , w y , w z are the weights in each direction and can be adjusted according to the importance of the terrain (e.g., when going up and down stairs, the weight of w x can be increased).

[0135] f(·) and g(·) can be different functions, for example:

[0136] f(Δx) = |Δx| (using L1 in the horizontal direction)

[0137] g(Δz) = (Δz)^(2) (using L(2) in the vertical direction, more sensitive to height errors)

[0138] or f(Δ) = max(0, |Δ| - threshold) (only punishing deviations exceeding the threshold).

[0139] Feature: The highest flexibility, which can precisely express different tolerances for deviations in different directions and the non - linear response to deviations.

[0140] 4. Piecewise function based on contact point classification

[0141] Logic: First, determine whether the foot - end landing point is within the ideal area (i.e., ||d_{horizontal}|| < radius, where radius is the radius of the ideal area).

[0142] If within the area: Reward (Primarily penalizing height error). Where A: Base reward value. This reward is received as long as the player lands within the target area. B: Height deviation penalty coefficient. Here, the main penalty is for vertical height error (|Δz|), since the horizontal position is already considered "acceptable." This encourages the agent to land steadily after finding the correct position.

[0143] If outside the designated area: Reward (Primarily penalizes horizontal deviation, with heavier penalties). Where C: Horizontal deviation penalty coefficient. In this case, a negative reward (penalty) is given, and the penalty amount is proportional to the distance of the horizontal deviation. Furthermore, C is usually set relatively large, meaning that the penalty for landing outside the safe zone is very severe, prioritizing ensuring that the agent does not step into dangerous areas.

[0144] Features: Closer to physical intuition, clearly distinguishing the cost difference between "stepping in the right spot but stepping off" and "completely missing the mark".

[0145] As described above, the robot precise landing planning method based on a priori terrain model provided in this application can construct a digital map containing terrain geometry and semantic information by digitally modeling a preset terrain in a simulation environment, and clearly mark the center coordinates of each ideal landing area as the target landing point. During training, the global coordinates of the robot's foot are tracked in real time, and the three-dimensional deviation between the robot's foot and the nearest target landing point is determined through map mapping. A reward function with the magnitude of the deviation as an evaluation index is then designed to drive the robot to learn a high-precision landing strategy. This invention introduces a priori terrain target point as guidance, giving the reward signal a clear geometric meaning, which can directly and efficiently drive the strategy to converge to precise landing behavior, significantly improving the motion accuracy and reliability on structured terrain.

[0146] To guide a robot in precise landing point planning, enabling it to achieve accurate, stable, and adaptive traversal across discrete landing points, this application provides an embodiment of a robot precise landing planning device based on a priori terrain model, which implements all or part of the aforementioned robot precise landing planning method based on a priori terrain model. See [link to embodiment]. Figure 2 The robot precise landing planning device based on the prior terrain model specifically includes the following components:

[0147] The map building module 10 is used to perform gridding processing on the preset terrain in a simulation environment to build a digital map containing terrain geometry and attributes. The digital map contains several terrain units. The digital map defines the ideal landing area of ​​each terrain unit, the target landing point in each ideal landing area, and the three-dimensional coordinates of each target landing point in a predefined world coordinate system.

[0148] The deviation calculation module 20 is used to track the three-dimensional coordinates of the robot foot in the world coordinate system in real time and map them onto the digital map to determine whether the robot foot lands in the ideal landing area at the end of the swing phase; if the robot foot lands in the ideal landing area, the relative positional relationship between the robot foot and the target landing point in the ideal landing area is calculated.

[0149] The function setting module 30 is used to set the reward to be lower or the penalty to be heavier as the landing point of the robot's foot deviates further from the target landing point. Based on this setting, the robot landing reward function is set to guide the robot to optimize its ability to land at the target landing point.

[0150] As described above, the robot precise landing planning device based on a priori terrain model provided in this application can construct a digital map containing terrain geometry and semantic information by digitally modeling a preset terrain in a simulation environment, and clearly mark the center coordinates of each ideal landing area as the target landing point. During training, the global coordinates of the robot's foot are tracked in real time, and the three-dimensional deviation between the robot's foot and the nearest target landing point is determined through map mapping. A reward function with the magnitude of the deviation as an evaluation index is then designed to drive the robot to learn a high-precision landing strategy. This invention introduces priori terrain target points as guidance, giving the reward signal a clear geometric meaning, which can directly and efficiently drive the strategy to converge to precise landing behavior, significantly improving the motion accuracy and reliability on structured terrain.

[0151] From a hardware perspective, in order to guide the robot to perform precise landing point planning and enable the robot to achieve accurate, stable, and adaptive traversal capabilities at discrete landing points, this application provides an embodiment of an electronic device for implementing all or part of the aforementioned robot precise landing planning method based on a priori terrain model. The electronic device specifically includes the following components:

[0152] The system comprises a processor, memory, a communication interface, and a bus; wherein the processor, memory, and communication interface communicate with each other via the bus; the communication interface is used to realize information transmission between the robot precise landing planning device based on a priori terrain model and core business systems, user terminals, and related databases and other related devices; the logic controller can be a desktop computer, tablet computer, or mobile terminal, etc., and this embodiment is not limited to these. In this embodiment, the logic controller can be implemented with reference to the embodiments of the robot precise landing planning method based on a priori terrain model and the embodiments of the robot precise landing planning device based on a priori terrain model, the contents of which are incorporated herein, and repeated details will not be described again.

[0153] It is understood that the user terminal may include smartphones, tablet computers, network set-top boxes, portable computers, desktop computers, personal digital assistants (PDAs), in-vehicle devices, smart wearable devices, etc. Among these, the smart wearable devices may include smart glasses, smartwatches, smart bracelets, etc.

[0154] In practical applications, the robot's precise landing planning method based on a priori terrain models can be partially executed on the electronic device side as described above, or all operations can be completed in the client device. The choice can be made based on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations on this. If all operations are completed in the client device, the client device may further include a processor.

[0155] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission with the server. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0156] Figure 3 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 3 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 3 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0157] In one embodiment, the robot's precise landing planning method based on a prior terrain model can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following control:

[0158] Step S101: In the simulation environment, the preset terrain is meshed to construct a digital map containing terrain geometry and attributes. The digital map contains several terrain units. The digital map defines the ideal landing area of ​​each terrain unit, the target landing point within each ideal landing area, and the three-dimensional coordinates of each target landing point in a predefined world coordinate system.

[0159] Step S102: Track the three-dimensional coordinates of the robot foot in the world coordinate system in real time and map them onto the digital map to determine whether the robot foot lands in the ideal landing area at the end of the swing phase; if the robot foot lands in the ideal landing area, calculate the relative positional relationship between the robot foot and the target landing point in the ideal landing area.

[0160] Step S103: Set the reward to be lower or the penalty to be heavier as the landing point of the robot's foot deviates further from the target landing point, and set the robot landing reward function according to the setting to guide the robot to optimize its ability to land at the target landing point.

[0161] As described above, the electronic device provided in this application provides a digital map containing terrain geometry and semantic information by digitally modeling a preset terrain in a simulation environment, and explicitly marking the center coordinates of each ideal landing area as the target landing point. During training, the global coordinates of the robot's foot are tracked in real time, and the three-dimensional deviation between the robot's foot and the nearest target landing point is determined through map mapping. A reward function with the magnitude of the deviation as an evaluation index is then designed to drive the robot to learn a high-precision landing strategy. This invention introduces prior terrain target points as guidance, giving the reward signal a clear geometric meaning. This allows the strategy to converge to precise landing behavior directly and efficiently, significantly improving the motion accuracy and reliability on structured terrain.

[0162] In another embodiment, the robot precise landing planning device based on the prior terrain model can be configured separately from the central processing unit 9100. For example, the robot precise landing planning device based on the prior terrain model can be configured as a chip connected to the central processing unit 9100, and the robot precise landing planning method function based on the prior terrain model can be realized through the control of the central processing unit.

[0163] like Figure 3As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily need to include these components. Figure 3 All components shown; in addition, the electronic device 9600 may also include Figure 3 For components not shown, please refer to existing technologies.

[0164] like Figure 3 As shown, the central processing unit 9100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device, which receives inputs and controls the operation of various components of the electronic device 9600.

[0165] The memory 9140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 9100 may execute the program stored in the memory 9140 to perform information storage or processing, etc.

[0166] Input unit 9120 provides input to central processing unit 9100. Input unit 9120 may be, for example, a keypad or touch input device. Power supply 9170 provides power to electronic device 9600. Display 9160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0167] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 via the central processing unit 9100.

[0168] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device for communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0169] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processing unit 9100 to provide input signals and receive output signals, which is the same as in a conventional mobile communication terminal.

[0170] Based on different communication technologies, multiple communication modules 9110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module 9110 (transmitter / receiver) is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing typical telecommunications functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 9130 is coupled to a central processing unit 9100, enabling on-device recording via the microphone 9132 and on-device playback of stored audio via the speaker 9131.

[0171] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the robot precise landing planning method based on a priori terrain model, where the execution subject is a server or client, as described in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the robot precise landing planning method based on a priori terrain model, where the execution subject is a server or client, as described in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0172] Step S101: In the simulation environment, the preset terrain is meshed to construct a digital map containing terrain geometry and attributes. The digital map contains several terrain units. The digital map defines the ideal landing area of ​​each terrain unit, the target landing point within each ideal landing area, and the three-dimensional coordinates of each target landing point in a predefined world coordinate system.

[0173] Step S102: Track the three-dimensional coordinates of the robot foot in the world coordinate system in real time and map them onto the digital map to determine whether the robot foot lands in the ideal landing area at the end of the swing phase; if the robot foot lands in the ideal landing area, calculate the relative positional relationship between the robot foot and the target landing point in the ideal landing area.

[0174] Step S103: Set the reward to be lower or the penalty to be heavier as the landing point of the robot's foot deviates further from the target landing point, and set the robot landing reward function according to the setting to guide the robot to optimize its ability to land at the target landing point.

[0175] As described above, the computer-readable storage medium provided in this application provides a digital map containing terrain geometry and semantic information by digitally modeling a preset terrain in a simulation environment, and explicitly marking the center coordinates of each ideal landing area as the target landing point. During training, the global coordinates of the robot's foot are tracked in real time, and the three-dimensional deviation between the robot's foot and the nearest target landing point is determined through map mapping. A reward function with the magnitude of the deviation as an evaluation index is then designed to drive the robot to learn a high-precision landing strategy. This invention introduces prior terrain target points as guidance, giving the reward signal a clear geometric meaning, which can directly and efficiently drive the strategy to converge to precise landing behavior, significantly improving the motion accuracy and reliability on structured terrain.

[0176] Embodiments of this application also provide a computer program product capable of implementing all steps in the robot precise landing planning method based on a priori terrain model, where the execution subject is a server or client, as described in the above embodiments. When executed by a processor, this computer program / instruction implements the steps of the robot precise landing planning method based on a priori terrain model. For example, the computer program / instruction implements the following steps:

[0177] Step S101: In the simulation environment, the preset terrain is meshed to construct a digital map containing terrain geometry and attributes. The digital map contains several terrain units. The digital map defines the ideal landing area of ​​each terrain unit, the target landing point within each ideal landing area, and the three-dimensional coordinates of each target landing point in a predefined world coordinate system.

[0178] Step S102: Track the three-dimensional coordinates of the robot foot in the world coordinate system in real time and map them onto the digital map to determine whether the robot foot lands in the ideal landing area at the end of the swing phase; if the robot foot lands in the ideal landing area, calculate the relative positional relationship between the robot foot and the target landing point in the ideal landing area.

[0179] Step S103: Set the reward to be lower or the penalty to be heavier as the landing point of the robot's foot deviates further from the target landing point, and set the robot landing reward function according to the setting to guide the robot to optimize its ability to land at the target landing point.

[0180] As described above, the computer program product provided in this application, through digital modeling of a preset terrain in a simulation environment, constructs a digital map containing terrain geometry and semantic information, and explicitly marks the center coordinates of each ideal landing area as the target landing point. During training, the global coordinates of the robot's foot are tracked in real time, and its three-dimensional deviation from the nearest target landing point is determined through map mapping. A reward function with the magnitude of the deviation as an evaluation index is then designed to drive the robot to learn a high-precision landing strategy. This invention introduces prior terrain target points as guidance, giving the reward signal a clear geometric meaning, enabling it to directly and efficiently drive the strategy to converge to precise landing behavior, significantly improving the motion accuracy and reliability on structured terrain.

[0181] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0182] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0183] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0184] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0185] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for precise robot landing planning based on a priori terrain model, characterized in that, The method includes: In the simulation environment, the preset terrain is gridded to construct a digital map containing terrain geometry and attributes. The digital map contains several terrain units. The digital map defines the ideal landing area of ​​each terrain unit, the target landing point within each ideal landing area, and the three-dimensional coordinates of each target landing point in a predefined world coordinate system. The robot foot is tracked in real time in the three-dimensional coordinates of the world coordinate system and mapped onto the digital map to determine whether the robot foot lands in the ideal landing area at the end of the swing phase; if the robot foot lands in the ideal landing area, the relative positional relationship between the robot foot and the target landing point in the ideal landing area is calculated. The robot's foot placement is set to deviate further from the target placement point, resulting in a lower reward or a heavier penalty. A robot placement reward function is then set based on this setting to guide the robot to optimize its ability to place its foot at the target placement point.

2. The robot precise landing planning method based on a priori terrain model according to claim 1, characterized in that, The step of performing gridding processing on the preset terrain to construct a digital map containing terrain geometry and attributes includes: In a simulation environment, a 3D model of a preset terrain is obtained; the 3D model contains geometric information and preset semantic tags. Define a fixed world coordinate system as a reference for terrain and robot pose, and determine a reference horizontal plane based on the overall orientation of the terrain; The three-dimensional model of the preset terrain is projected in two dimensions onto the reference horizontal plane to determine the two-dimensional envelope boundary of the terrain; a grid is generated within the two-dimensional envelope boundary, and each grid cell is identified by a two-dimensional index. The grid cells are assigned attribute values, including geometric attribute assignment, semantic attribute assignment, and derived attribute assignment; the semantic attributes include the coordinates of the target landing point; the derived attributes include accessibility and normal vector. The generated data, which includes geometry and attributes, is structured and stored to obtain the digital map.

3. The robot precise landing planning method based on a priori terrain model according to claim 2, characterized in that: The method for assigning geometric attributes includes: projecting the mesh cell vertically upward onto the three-dimensional model of the terrain, finding the intersection with the surface of the three-dimensional model, and setting the surface height of the mesh cell based on the intersection point; The method for assigning the semantic attributes includes: pre-defining an ideal landing area based on the semantic labels of the terrain; specially marking the grid cells within the ideal landing area; and storing the three-dimensional coordinates of the target landing point in the world coordinate system in the attributes of all associated grid cells within the ideal landing area. The method for assigning accessibility includes: assigning a accessibility score to a grid cell based on its attribute characteristics, including slope, height difference, surface material, and coefficient of friction. The method for assigning the normal vector includes: calculating the average normal vector of the surface where the grid cell is located as the normal vector of the grid cell.

4. The robot precise landing planning method based on a priori terrain model according to claim 1, characterized in that, Methods for obtaining target landing points within each ideal landing area include: A specific point on the horizontal plane of the grid cell is selected as the target landing point, and the specific point includes the geometric center point and the centroid point; Alternatively, multiple candidate target landing points can be distributed within the ideal landing area, where the target landing point is the point closest to the foot landing point when the foot lands. Methods for obtaining the three-dimensional coordinates of each target's landing point in a predefined world coordinate system include: The position of the target's landing point on the horizontal plane of the grid cell is taken as its x and y coordinates in the world coordinate system, and the height of the grid cell surface corresponding to the target's landing point is taken as its z coordinate in the world coordinate system.

5. The robot precise landing planning method based on a priori terrain model according to claim 1, characterized in that, The step of tracking the three-dimensional coordinates of the robot's foot in the world coordinate system in real time, mapping them onto the digital map, and determining whether the robot's foot lands within the ideal landing area at the end of the swing phase includes: The robot's foot is used to calculate the three-dimensional coordinates in the world coordinate system in real time using the robot's forward kinematics chain and joint angle sensor data. The three-dimensional coordinates of the robot's foot are mapped to a digital map to obtain the grid cell index corresponding to the horizontal coordinate of the foot in the digital map when the foot is at the end of the swing phase. Use the grid cell index to query the semantic tag of the grid cell to determine whether the robot's foot lands within the ideal landing area.

6. The robot precise landing planning method based on a priori terrain model according to claim 1, characterized in that, The step of calculating the relative positional relationship between the robot's foot and the target landing point within the ideal landing area if the robot's foot lands within that area includes: Obtain the three-dimensional coordinates of the target landing point in the world coordinate system within the ideal landing area of ​​the robot; the three-dimensional coordinates of the target landing point are stored in the attributes of each grid cell associated with the ideal landing area; Based on the three-dimensional coordinates of the robot's foot landing point in the world coordinate system and the three-dimensional coordinates of the target landing point in the world coordinate system, the three-dimensional deviation vector between the two is calculated, which is the relative positional relationship between the robot's foot and the target landing point in the ideal landing area.

7. The robot precise landing planning method based on a priori terrain model according to claim 1, characterized in that, The step of setting a robot landing reward function based on the premise that the further the robot's foot landing point deviates from the target landing point, the lower the reward or the heavier the penalty, includes: Calculate the deviation vector between the robot's foot landing point and the target landing point, and input it into the deviation evaluation function, which outputs a reward or penalty according to the settings. The deviation evaluation functions include the squared difference function, the L1 norm penalty term, the saturated continuous reward function, the separated weighted and nonlinear processing function, and the piecewise function based on contact point classification.

8. A robot precise landing planning device based on a priori terrain model, characterized in that, The device includes: The map building module is used to perform gridding on the preset terrain in a simulation environment to build a digital map containing terrain geometry and attributes. The digital map contains several terrain units. The digital map defines the ideal landing area of ​​each terrain unit, the target landing point in each ideal landing area, and the three-dimensional coordinates of each target landing point in a predefined world coordinate system. The deviation calculation module is used to track the three-dimensional coordinates of the robot's foot in the world coordinate system in real time and map them onto the digital map to determine whether the robot's foot lands in the ideal landing area at the end of the swing phase; if the robot's foot lands in the ideal landing area, the module calculates the relative positional relationship between the robot's foot and the target landing point in the ideal landing area. The function setting module is used to set the reward or penalty for the robot's foot landing point being further away from the target landing point. Based on this setting, the robot landing reward function is set to guide the robot to optimize its ability to land at the target landing point.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the robot precise landing planning method based on a priori terrain model as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the robot precise landing planning method based on a priori terrain model as described in any one of claims 1 to 7.

Images