A method and system for constructing a priori map of static space physics for embodied intelligence

Abstract

The application relates to a somatic intelligent static space physical priori map construction method and system, and relates to the technical field of somatic intelligent world model, which comprises the following steps: in response to a map construction signal, performing feature analysis on an imported 2D house plan to obtain module features, and performing feature matching one by one to obtain a three-dimensional physical prediction sandbox; performing a map scanning construction operation, and performing a somatic intelligent motion simulation exercise to obtain a pre-performance motion condition; when the pre-performance motion condition is a collision condition, the ground material features of the current position are looked up to obtain corresponding deceleration operations of the somatic intelligent robot, and the somatic intelligent robot is controlled to perform the deceleration operations; when the pre-performance motion condition is a smooth condition, the map scanning construction operation is continuously performed according to the pre-performance route, and the current position and the current orientation are continuously obtained. The application has the effect of preventing the somatic intelligent robot and the map module from being damaged due to collision by the pre-performance motion condition.

Description

Technical Field

[0001] This invention relates to the field of embodied intelligence world model technology, and in particular to a method and system for constructing static spatial physical prior maps of embodied intelligence. Background Technology

[0002] Various embodied intelligent robots need to frequently move and operate autonomously in complex and ever-changing environments, adapting to the long-term operational needs of multiple scenarios, and continuously traversing large areas of space. As the core carrier for intelligent agents' positioning, navigation, and interactive decision-making, the completeness of physical constraints and the stability of global positioning directly affect the accuracy of motion planning and the reliability of environmental interaction.

[0003] Currently, traditional 3D architectural drawings primarily serve two core needs: human visual expression and on-site construction design. Relying on standardized axonometric and perspective 3D drawing logic, they construct complete spatial forms. The drawings contain high-precision and detailed geometric information, accurately reproducing the length, width, height, corners, overlapping relationships, and spatial arrangement of various components such as building walls, beams, columns, doors, windows, stairs, and pipelines. The lines, layers, and annotation system strictly follow architectural drafting standards. They can not only intuitively present the overall appearance of the building and the atmosphere of the interior space through perspective renderings for communication of plans and reporting to clients, but also clearly dissect complex node structures using isometric 3D detailed drawings, making it easier for construction personnel to intuitively understand the three-dimensional overlapping logic that is difficult to express in two-dimensional plan drawings.

[0004] Regarding the aforementioned technologies, existing embodied intelligence, when entering new physical spaces, is forced to rely on sensors such as camera modules and lidar for extremely low-energy-efficiency on-site reasoning. When faced with highly reflective surfaces, highly transparent glass, carpets, and other acoustically sound-absorbing surfaces, it is prone to visual false positives, slipping, or overload collisions, and lacks common-sense priors for predicting motion in the physical environment. Summary of the Invention

[0005] To address the lack of common-sense priors for predicting motion in the physical environment, this invention provides a method and system for constructing static spatial physical prior maps of embodied intelligence.

[0006] In a first aspect, the present invention provides a method for constructing a static spatial physical prior map of embodied intelligence, employing the following technical solution:

[0007] A method for constructing a static spatial-physical prior map of embodied intelligence, comprising:

[0008] Step 100: In response to the map building signal, import the preset 2D floor plan;

[0009] Step 101: Perform feature analysis on the 2D floor plan to obtain module features that conform to preset text features or preset image fill features;

[0010] Step 102: After matching the module features one by one according to the preset standard map module features, a three-dimensional physical prediction sandbox is established through the successfully matched module features. The standard map module features have map object parameters, which include surface friction, stiffness, acoustics, load-bearing limit and mechanical parameters.

[0011] Step 103: Simulate the motion of the embodied intelligent robot in the 3D physics prediction sandbox to obtain the pre-execution route, and perform the preset map scanning and construction operation in the 3D physics prediction sandbox according to the pre-execution route;

[0012] Step 104: During the execution of the preset map scanning and construction operation, obtain the current location and current orientation;

[0013] Step 105: Input the current position and current orientation into the three-dimensional physics prediction sandbox to conduct embodied intelligence motion simulation exercise to obtain the pre-movement situation, which includes collision situation and unobstructed situation;

[0014] Step 106: When the pre-simulated motion is a collision, retrieve the ground material features of the current position in the 3D physics prediction sandbox;

[0015] Step 107: Based on the characteristics of the ground material, find the deceleration operation corresponding to the embodied intelligent robot, and control the embodied intelligent robot to execute the deceleration operation;

[0016] Step 108: When the pre-simulated movement is unobstructed, continue to perform map scanning and construction operations according to the pre-simulated route, and continue to obtain the current position and current orientation.

[0017] By adopting the above technical solution, map module features are extracted by importing the spatial map. These features include parameters of the map objects within the module, and a pre-planned route for the embodied intelligent robot is obtained based on these parameters. The pre-planned route allows for the prediction of motion, preventing collisions between the embodied intelligent robot and the map module that could damage either.

[0018] Optional, also includes:

[0019] Step 200: Based on the current location, the pre-simulation route and the preset verification distance, determine the map module directly in front, define the map module as the current map module, and retrieve the map object parameters of the current map module in the 3D physical prediction sandbox, and define the map object parameters as the current parameters.

[0020] Step 201: If the current parameters are the same as the actual parameters, continue to perform map scanning and construction operations according to the pre-simulated route, and continue to obtain the current position and current orientation;

[0021] Step 202: If the current map module does not exist, continue to perform map scanning and construction operations according to the pre-rehearsed route, and continue to obtain the current position and current orientation.

[0022] By adopting the above technical solution, the problem of map scanning and construction operation interruption caused by missing map modules is solved by comparing the current parameters with the actual parameters, ensuring the stable operation of map scanning and construction, and improving the integrity and realism of the 3D physical prediction sandbox.

[0023] Optional, also includes:

[0024] Step 300: If the current map module exists but no actual parameters are detected, search for the current map module to obtain its current appearance features;

[0025] Step 301: Rotate the camera in a preset rotation direction and acquire the captured image; the environmental sensor includes the camera.

[0026] Step 302: Analyze the captured images to obtain image features;

[0027] Step 303: When the image features and the current appearance features partially match or do not match, continue to rotate the camera until the image features and the current appearance features completely match;

[0028] Step 304: When the image features and the current appearance features are a perfect match, obtain the camera rotation angle and analyze the current distance between the current map module and the embodied intelligent robot based on the image features;

[0029] Step 305: Re-detect the actual parameters of the current map module in real time through the environmental sensor, and correct the actual parameters based on the camera rotation angle and the current distance to obtain the movement parameters;

[0030] Step 306: Update the movement parameters in the 3D physics prediction sandbox and update the simulation route.

[0031] By adopting the above technical solution, the accurate position of the current map is obtained by scanning with a rotating camera, which solves the problem that the deviation between the current map module and the actual map module causes the pre-simulated motion to be inconsistent with the actual motion, resulting in a collision between the embodied intelligent robot and the current map module, thus ensuring the safety of the embodied intelligent robot and the current map module.

[0032] Optionally, it also includes a method for searching the current map module when there is never a partial or complete match between the image features and the current appearance features, the method including:

[0033] Step 400: Define the current map module as a missing map module and the current location as a missing verification location;

[0034] Step 401: Based on the missing checkpoint and the rehearsal route, determine the rehearsal route after the missing checkpoint and define the rehearsal route as the remaining rehearsal route;

[0035] Step 402: Place the missing map module in the 3D physics prediction sandbox for simulation to determine the possible location of the missing map module on the remaining pre-simulation route;

[0036] Step 403: If the possible location does not exist, continue to perform map scanning and construction operations according to the remaining pre-simulation route;

[0037] Step 404: If possible locations exist, update the 3D physical prediction sandbox using all possible locations and missing map modules, update the remaining pre-simulation routes, and update the missing map modules according to possible locations until the image features and current appearance features partially or completely match.

[0038] Step 405: When the image features and the current appearance features partially or completely match, update the corresponding possible position to the actual position and delete all other possible positions from the 3D physical prediction sandbox.

[0039] By adopting the above technical solution, the safety of the current pre-rehearsal route is inferred by simulating the possible positions of the missing map module on the remaining pre-rehearsal route, avoiding collisions with map modules at possible positions, and finding the location of the missing animated module and updating the 3D physical prediction sandbox and prediction route, thus ensuring the safety of the embodied intelligent robot and the map module.

[0040] Optionally, if possible locations exist, updating the 3D physics prediction sandbox using all possible locations and missing map modules, updating the remaining simulation routes, and updating the missing map modules according to possible locations includes the following methods:

[0041] Step 500: Obtain the theoretical location of the missing map module on the remaining pre-rehearsal route;

[0042] Step 501: Based on the module's theoretical location, the remaining pre-simulation route is divided to obtain the blind zone route area before the module's theoretical location;

[0043] Step 502: If the possible location does not fall into the blind spot route area, then update the 3D physics prediction sandbox with all possible locations and missing map modules, update the remaining pre-simulation route and update the missing map modules according to the possible locations;

[0044] Step 503: If the possible location falls into the blind spot route area, determine the backtracking distance of the embodied intelligent robot based on the possible location and the verification distance, backtrack according to the backtracking distance, and then update the remaining rehearsal route based on the backtracking distance.

[0045] By adopting the above technical solution, the theoretical position of the missing map module on the remaining pre-simulation route is located and the blind zone route area is divided. Based on whether the possible position of the missing module is in the blind zone route area, the data update in the 3D physical prediction sandbox is executed. The embodied intelligent robot is controlled to backtrack according to the backtracking distance and correct the remaining pre-simulation route. This solution specifically solves the problem of missing map modules, eliminates the deviation problem caused by the blind zone route area, and ensures the stable operation of map scanning and construction and the accurate planning of the pre-simulation route.

[0046] Optionally, it also includes an optimized search method for the current map module when there is never a local or complete match between the image features and the current appearance features. This method:

[0047] Step 600: Analyze the 3D physics prediction sandbox based on the current location to obtain the occlusion module that can completely occlude the current appearance features and the occlusion location of the missing map module.

[0048] Step 601: Determine the visible locations of the missing map module on the remaining pre-playback route based on the occlusion positions of the occlusion module and the missing map module. The visible locations are several.

[0049] Step 602: Simulate the occlusion module in the 3D physics prediction sandbox based on the visible position of the embodied intelligent robot and the occlusion position of the missing map module;

[0050] Step 603: Filter out the visible locations where the additional occlusion module does not exist, and define the visible location as the verification location;

[0051] Step 604: Control the embodied intelligent robot to move to the verification position to obtain the verification image, and define the image features corresponding to the verification image as the verification image features;

[0052] Step 605: When the checked image features do not match the current appearance features, return to the missing check position and re-execute steps 401 to 404;

[0053] Step 606: When the image features and the current appearance features are partially or completely matched, the missing map module is updated by the occlusion location.

[0054] By employing the above technical solution, the occlusion points of the occlusion module and the missing map module are analyzed based on the 3D physical prediction sandbox. Several visible positions are deduced, and the embodied intelligent robot is moved to a visible position to re-capture and compare it with the current appearance features. If the match fails, the visible positions are re-selected for verification; if the match is successful, the missing map module is completed and updated according to the visible positions. This method can accurately avoid environmental interference to complete the verification and supplementation of the missing map module, improving the matching accuracy between the 3D physical prediction sandbox and spatial features.

[0055] Optionally, it also includes a method for searching the current map module if no image feature and the current appearance feature can be completely matched, but a partial match exists. This method includes:

[0056] Step 700: Obtain the shooting angle when locally matching image features and current appearance features;

[0057] Step 701: Analyze the 3D physical prediction sandbox based on the shooting angle and missing verification location to obtain a local occlusion module that can locally occlude the current appearance features;

[0058] Step 702: Control the embodied intelligent robot to move to the verification position corresponding to the partial occlusion module to perform verification and execute steps 604 to 606.

[0059] By adopting the above technical solution, and by combining the shooting angle and the verified location, the local occlusion module that causes partial occlusion of the current appearance features is analyzed in the 3D physical prediction sandbox. The embodied intelligent robot is then controlled to go to the corresponding location to re-shoot the image, analyze the features, and perform matching and correction. This effectively solves the problem of incomplete matching between the acquired image features and the current appearance features caused by partial environmental occlusion, and comprehensively corrects the missing map module information, further improving the accuracy and spatial matching precision of the 3D physical prediction sandbox.

[0060] Optionally, specific methods for finding the corresponding deceleration operation for the embodied intelligent robot based on ground material characteristics and controlling the embodied intelligent robot to execute the deceleration operation include:

[0061] Step 800: Obtain the collision route based on the current location and the verification distance;

[0062] Step 801: Find the coefficient of friction for each surface material characteristic one by one;

[0063] Step 802: Based on the friction coefficient and the verification distance, obtain the deceleration rate of the ground material characteristics on the collision route;

[0064] Step 803: Based on 3D physics prediction sandbox analysis, obtain the different ground material features that appear on the collision route;

[0065] Step 804: Determine the output torque based on the friction coefficient and deceleration rate;

[0066] Step 805: Control the embodied intelligent robot to decelerate according to the output torque on ground surfaces with different material characteristics.

[0067] By adopting the above technical solution, the collision route is predicted by combining the current position and the verification position of the embodied intelligent robot. Based on the torque adapted to the ground material characteristics of the route, the embodied intelligent robot is controlled to dynamically adjust the braking force according to the ground material characteristics to complete the graded deceleration. The deceleration rate can be precisely adjusted for different road conditions, avoiding the problem of slipping and falling or insufficient braking leading to collision when switching ground material characteristics due to a single deceleration rate. This improves the motion control stability and obstacle avoidance safety of the embodied intelligent robot during the movement process.

[0068] Optionally, an optimized method for controlling the embodied intelligent robot to decelerate according to the deceleration rate when the pre-simulated motion situation is a collision situation, the method includes:

[0069] Step 900: Obtain the second safe collision constraint value of the current map module through the 3D physics prediction sandbox;

[0070] Step 901: Compare the preset first safe collision constraint value and the second safe collision constraint value to obtain the maximum safe collision constraint value, wherein the maximum safe collision constraint value is the minimum of the first safe collision constraint value and the second safe collision constraint value;

[0071] Step 902: Update the deceleration rate based on the maximum safe collision constraint to obtain the actual required deceleration rate, wherein the actual required deceleration rate is the pressure value generated when the embodied intelligent robot moves forward at the actual required deceleration rate and collides with the current map module, which is less than the maximum safe collision constraint value.

[0072] Step 903: Control the embodied intelligent robot to decelerate according to the actual required deceleration rate.

[0073] By adopting the above technical solution, the maximum safe collision constraint value is obtained by comparing the safe collision constraint values ​​corresponding to the materials of the embodied intelligent robot and the map module, so as to ensure that the collision impact force does not exceed the actual required deceleration rate, and control the embodied intelligent robot to decelerate at this deceleration rate. The deceleration standard can be dynamically adjusted by combining the safety thresholds of both the embodied intelligent robot and the map module to prevent damage to the embodied intelligent robot or the map module when they collide.

[0074] Secondly, the present invention provides a static spatial physical prior map construction system for embodied intelligence, which adopts the following technical solution:

[0075] A static spatial-physical prior map construction system for embodied intelligence, comprising:

[0076] The acquisition module is used to acquire map building signals, spatial 2D floor plan drawings, module features, 3D physical prediction sandbox, current location, current orientation, current distance, camera rotation angle, environmental sensors, pre-simulated route and images;

[0077] A memory for storing the program for constructing a static spatial-physical prior map of embodied intelligence as described above;

[0078] The processor loads and executes programs from memory.

[0079] By adopting the above technical solution, the motion trajectory of the embodied intelligent robot can be predicted through three-dimensional physical prediction sandbox simulation, and the motion situation can be obtained through the pre-rehearsal route, thus preventing the embodied intelligent robot and the map module from colliding and causing damage to the embodied intelligent robot or the map module.

[0080] In summary, the present invention has at least one of the following beneficial technical effects:

[0081] 1. By importing the spatial map, map module features are extracted, and a pre-planned route for the embodied intelligent robot is obtained. This pre-planned route allows for the prediction of movement patterns, preventing collisions between the embodied intelligent robot and the map module that could damage either.

[0082] 2. By using a rotating camera to capture and scan the map, the accurate location of the current map can be obtained, thus solving the problem of discrepancies between the current map module and the actual map module, which leads to inconsistencies between the simulated motion and the actual motion, resulting in collisions between the embodied intelligent robot and the current map module.

[0083] 3. By combining the current position and the verification position of the embodied intelligent robot, the route of the collision is predicted. Based on the torque adapted to the ground material characteristics of the route, the deceleration rate can be precisely adjusted for different road conditions. This avoids the problem of slipping and falling or insufficient braking leading to collision when switching ground material characteristics due to a single deceleration rate. Attached Figure Description

[0084] Figure 1 This is a flowchart of a method for constructing static spatial physical prior maps of embodied intelligence.

[0085] Figure 2 It is a schematic diagram of the pre-rehearsed route obtained by simulating the motion of an embodied intelligent robot based on a 3D physics prediction sandbox.

[0086] Figure 3This is a schematic diagram illustrating the search for the current map module when there is no partial or complete match between the image features and the current appearance features.

[0087] Figure 4 This is a schematic diagram illustrating the method of updating the remaining rehearsal route by considering all possible locations and missing map modules, and updating the missing map modules according to the possible locations.

[0088] Figure 5 This is a schematic diagram of an optimized search method for the current map module when there is no partial or complete match between the image features and the current appearance features. Detailed Implementation

[0089] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0090] This invention discloses a method for constructing a static spatial physical prior map of embodied intelligence. (Refer to...) Figure 1 A method for constructing static spatial-physical prior maps of embodied intelligence includes:

[0091] Step 100: In response to the map building signal, import the preset 2D floor plan.

[0092] The map building signal is the signal to initiate the map scanning and building operation, generated by a user-issued command. The 2D floor plan is the original spatial layout data upon which the current 3D physics prediction sandbox will base its map scanning and building operation. It includes complete geometric information such as the planar 2D outline, wall locations and thicknesses, room partition boundaries, door and window opening locations and dimensions, and fixed furniture and obstacle boundaries. This data is pre-imported by the user into the 3D physics prediction sandbox's drawing management module. The import process includes drawing format parsing, coordinate system normalization, and scale calibration to ensure that the drawing data matches the scale of the actual physical space.

[0093] Step 101: Perform feature analysis on the 2D floor plan to obtain module features that conform to preset text features or preset image fill features.

[0094] Text features are generated by converting text describing modules in a 2D floor plan into 3D drawing signals that can be used for analysis within a 3D physics prediction sandbox. Image fill features are image identifiers using unique fill colors and textures to distinguish walls, floors, furniture, and passageways in different areas of the 2D floor plan, derived from the analysis of manually imported 2D floor plans. Module features are generated by extracting the recognition features of individual spatial units such as walls, floors, doors, windows, and furniture after splitting the 2D floor plan, also derived from the analysis of manually imported 2D floor plans. Module features are obtained by integrating the identified text and image fill features and grouping them according to the individual spatial units.

[0095] Step 102: After matching the module features one by one according to the preset standard map module features, a three-dimensional physical prediction sandbox is established through the successfully matched module features. The standard map module features have map object parameters, including surface friction, stiffness, acoustics, load-bearing limit and mechanical parameters.

[0096] The standard map module features are standard material features pre-existing in the module library within the 3D physics prediction sandbox. These features are filtered, stored, and entered into the material feature library through big data analysis. The 3D physics prediction sandbox is a virtual 3D simulation space with complete physical parameters, replicating the structure and physical properties of real space. It is used for pre-simulating the movement of embodied intelligent robots and is constructed by stitching together successfully matched map modules carrying spatial dimensions, structural information, and material features. A map module is a virtual module stored in the 3D physics prediction sandbox after matching with the standard map module features. It is used to build the virtual 3D simulation space. Map object parameters refer to the set of physical attribute data carried by each map object in the 3D physics prediction sandbox. These parameters are synchronously loaded into the 3D physics prediction sandbox after being modeled and matched with the features of each standard map module.

[0097] Step 103: Simulate the motion of the embodied intelligent robot in the 3D physics prediction sandbox to obtain a pre-simulated route, and perform a preset map scanning and construction operation in the 3D physics prediction sandbox according to the pre-simulated route.

[0098] An embodied intelligent robot is a robot with a physical entity that interacts physically with its environment through sensors and actuators; it serves as the platform for a 3D physics prediction sandbox. Motion simulation training involves simulating the physical interaction between the embodied intelligent robot and the virtual environment during its movement within the 3D physics prediction sandbox's virtual environment. The future trajectory is obtained by frame-by-frame integration using parameters matched within the 3D physics prediction sandbox. This frame-by-frame integration involves dividing future moments into countless tiny time segments and then, based on the current frame's position, velocity, forces, and dynamic equations, simultaneously performing collision detection and response correction. The aforementioned dynamic equations are... ,in The net external force acting on an object. For the mass of the object, The acceleration obtained by the object; the formula for calculating the velocity of the next frame by numerical integration is: ,in For the speed of the next frame, For the current frame rate, Acceleration for the current frame, The time step for predicting the sandbox in 3D physics. (Refer to...) Figure 2 Point A represents the location where the verification is performed according to the verification distance, module B is the current map module, and curve C is the pre-simulated route for the current movement. The arrow on curve C indicates the direction of movement of the embodied intelligent robot. The pre-simulated route is a movement route from the starting point to the target point that has undergone physical feasibility verification after motion simulation exercises. The path calculated by frame-by-frame integration in the motion simulation exercises is defined as the pre-simulated route. The map scanning and construction operation refers to the process of collecting object parameters in the real physical environment through environmental sensors in the 3D physics prediction sandbox. The real data of objects in the real physical environment obtained by the environmental sensors is integrated and compared with the data in the 3D physics prediction sandbox, and the 3D physics prediction sandbox is updated accordingly.

[0099] Step 104: During the execution of the preset map scanning and construction operation, obtain the current location and current orientation.

[0100] The current position is the real-time position of the embodied intelligent robot in real space, calculated in real time by the positioning system within the robot, which includes a combination of various positioning calculation units. The current orientation is the positive direction of the embodied intelligent robot in the current position coordinate system, and the deflection angle in the global reference coordinate system, ranging from 0° to 360°, obtained by the magnetometer within the embodied intelligent robot.

[0101] Step 105: Input the current position and current orientation into the 3D physics prediction sandbox for embodied intelligence motion simulation to obtain a preview of the motion, including collision and unobstructed conditions.

[0102] The pre-simulation motion result is the physical interaction result obtained after the embodied intelligent robot simulates movement along the pre-simulation route. It is output by integrating the state data recorded at each moment during the motion simulation exercise. The collision result is one outcome of the pre-simulation motion result, indicating the situation where the embodied intelligent robot collides with the map module during the virtual simulation motion. It is calculated by detecting the distance between the embodied intelligent robot and the map module. The unobstructed state is another outcome of the pre-simulation motion result, indicating that there are no map modules obstructing the path ahead of the embodied intelligent robot as it moves along the pre-simulation route during the virtual simulation motion. If no collision with a map module is detected throughout the pre-simulation, it is considered an unobstructed state.

[0103] Step 106: When the pre-simulated motion is a collision, retrieve the ground material features of the current position in the 3D physics prediction sandbox.

[0104] When the pre-simulated motion scenario indicates a collision, it means that the embodied intelligent robot will collide with the map module ahead while moving along the pre-simulated route at its current speed. This indicates that the embodied intelligent robot needs to take deceleration measures to prevent the collision. Ground material characteristics refer to the type of ground surface material and its related physical properties corresponding to the current location of the embodied intelligent robot in the 3D physics prediction sandbox, obtained through parameters stored in the 3D physics prediction sandbox.

[0105] Step 107: Based on the characteristics of the ground material, find the deceleration operation corresponding to the embodied intelligent robot, and control the embodied intelligent robot to perform the deceleration operation.

[0106] The deceleration operation is a safety braking or speed reduction control operation performed by the embodied intelligent robot to avoid collisions with the map module before a collision occurs. The deceleration rate is derived by working backward from the current speed, ground friction coefficient, and distance between the embodied intelligent robot and the map module, and then converted into an executable control signal. The formula for calculating the deceleration rate is as follows: ,in For deceleration rate, For the current speed, For the distance between the embodied intelligent robot and the map module, For the ground friction coefficient, This is the acceleration due to gravity.

[0107] Step 108: When the pre-simulated movement is unobstructed, continue to perform map scanning and construction operations according to the pre-simulated route, and continue to obtain the current position and current orientation.

[0108] When the pre-simulation movement is in a smooth state, it means that the embodied intelligent robot will not collide with the map module ahead when moving along the pre-simulation route at the current speed. This indicates that the embodied intelligent robot can continue to move at this speed without needing to decelerate.

[0109] This also includes:

[0110] Step 200: Based on the current location, the pre-simulation route, and the preset verification distance, determine the map module directly in front, define the map module as the current map module, and retrieve the map object parameters of the current map module in the 3D physical prediction sandbox, and define the map object parameters as the current parameters.

[0111] The verification distance is a fixed parameter set by the embodied intelligent robot to confirm the actual position of the map module ahead and to determine safety. The current map module is the one located directly in front of the embodied intelligent robot, determined by the current position, the pre-simulated route, and the verification distance. It is the first map module retrieved by querying the 3D physics prediction sandbox along the pre-simulated route according to the predicted distance range. The current parameters are the object parameters within the current map module retrieved from the 3D physics prediction sandbox, obtained by directly querying the physical parameters stored in the 3D physics prediction sandbox.

[0112] Step 201: If the current parameters are the same as the actual parameters, continue to perform map scanning and construction operations according to the pre-simulated route, and continue to obtain the current position and current orientation.

[0113] If the current parameters and the actual parameters are the same, it indicates that the current map module has not changed, and the data in the 3D physics prediction sandbox does not need to be updated. The actual parameters are the physical parameters of the real objects in the current map module, which are obtained in real time by the embodied intelligent robot through environmental sensors after entering the real physical environment. After the embodied intelligent robot moves to the verification position, it obtains the corresponding objects in the real physical environment through real-time measurement by environmental sensors.

[0114] Step 202: If the current map module does not exist, continue to perform map scanning and construction operations according to the pre-rehearsed route, and continue to obtain the current position and current orientation.

[0115] If the current map module is not present, it indicates that there is no map module on the current pre-simulation route that would obstruct the movement of the embodied intelligent robot. "Current map module not present" means that no map module was detected while moving along the embodied intelligent robot's current orientation and pre-simulation route in the 3D physics prediction sandbox; that is, there is no map module recorded in the 3D physics prediction sandbox in that direction.

[0116] This also includes:

[0117] Step 300: If the current map module exists but no actual parameters are detected, search for the current map module to obtain its current appearance features.

[0118] If the current map module exists but no actual parameters are detected, it indicates that the position parameters of the current map module have changed. The embodied intelligent robot, while at its verification position in the real physical environment, fails to detect the actual parameters of this map module through real-time environmental sensor monitoring. Current appearance features refer to the appearance information of the real object corresponding to the current map module, pre-stored in the 3D physical prediction sandbox, including shape, size, and surface color, obtained through parameters stored in the 3D physical prediction sandbox.

[0119] Step 301: Rotate the camera of the environmental sensor in the preset rotation direction and acquire the captured image.

[0120] The rotation direction refers to the direction setting for detecting actual parameters by changing the direction angle when the camera fails to detect them. The rotation direction includes clockwise rotation, counterclockwise rotation, and pitch rotation. It is set and stored in the 3D physical prediction sandbox based on factors such as the camera installation position and the azimuth angle deviation of the embodied intelligent robot corresponding to the current map module.

[0121] Step 302: Analyze the captured images to obtain image features.

[0122] Image features refer to the key visual information that can describe the content of an image, extracted through image processing.

[0123] Step 303: When the image features and the current appearance features partially match or do not match, continue to rotate the camera until the image features and the current appearance features completely match.

[0124] When the image features and the current appearance features partially match or do not match, it means that only some regions of the image features extracted from the captured image correspond to the current appearance features stored in the 3D physical prediction sandbox, or no effective features correspond at all. It is necessary to continue rotating the camera to find a position where the image features and the current appearance features completely match. A complete match means that all or most of the features extracted from the captured image correspond to the current appearance features stored in the 3D physical prediction sandbox.

[0125] Step 304: When the image features and the current appearance features are a perfect match, obtain the camera rotation angle and analyze the current distance between the current map module and the embodied intelligent robot based on the image features.

[0126] When the image features and the current appearance features perfectly match, it indicates that the embodied intelligent robot has completely captured the full appearance of the real object corresponding to the current map module and confirmed the object's identity. The camera rotation angle refers to the angle the camera has rotated in the direction of rotation from its initial position to the current feature-matched position. This angle value is read and recorded in real time by the environmental sensors within the embodied intelligent robot. The current distance refers to the straight-line distance between the embodied intelligent robot and the current map module in the current feature-matched state. This distance is also read and recorded in real time by the environmental sensors within the embodied intelligent robot.

[0127] Step 305: Re-detect the actual parameters of the current map module in real time through the environmental sensor, and correct the actual parameters based on the camera rotation angle and the current distance to obtain the movement parameters.

[0128] The movement parameters refer to the parameters that change as the embodied intelligent robot moves from a position where it partially matches or does not match the image features and current appearance features to a position where it completely matches. These parameters are obtained by reading and matching within a 3D physical prediction sandbox.

[0129] Step 306: Update the movement parameters in the 3D physics prediction sandbox and update the simulation route.

[0130] Updating the 3D physics prediction sandbox involves writing movement parameters into the sandbox, correcting any discrepancies between the map module and the actual physical environment, and assigning these movement parameters to the corresponding parameter fields of the current map module in the 3D physics prediction sandbox via a data write operation, overwriting the old values ​​with the new ones. Updating the pre-simulation route involves re-simulating the motion from the embodied intelligent robot's current position after the 3D physics prediction sandbox is updated, and outputting a new pre-simulation route that matches the updated actual physical environment.

[0131] This includes a method for searching the current map module when there is no local or complete match between the image features and the current appearance features. This method includes:

[0132] Step 400: Define the current map module as a missing map module and the current location as a missing verification location.

[0133] Reference Figure 3In this diagram, curve C represents the current pre-simulated route, dashed box D represents the missing map module, point E is the verification distance of the detected actual position of the current map module, curve F represents the remaining pre-simulated route, dashed circle G represents the range of the camera rotation to search for the missing map module, dashed box H represents the possible abandoned position, curve I represents the abandoned pre-simulated route updated based on the possible position H, point J represents the missing verification position, module K represents the actual position of the current map module, and the arrows on the curves indicate the direction of movement of the embodied intelligent robot. A missing map module is one that exists in the 3D physics prediction sandbox but cannot be detected by the embodied intelligent robot in the real physical environment after searching through environmental sensors, and still cannot achieve partial or complete matching of image features with the current appearance features after camera rotation search. The missing verification position refers to the current position of the embodied intelligent robot in the real physical space when a missing map module is detected; position data is recorded by environmental sensors and input into the 3D physics prediction sandbox.

[0134] Step 401: Based on the missing checkpoint and the rehearsal route, determine the rehearsal route after the missing checkpoint and define the rehearsal route as the remaining rehearsal route.

[0135] The remaining pre-simulation route is the route segment that has not yet been traversed on the original pre-simulation route, starting from the missing checkpoint. In the 3D physics prediction sandbox, the missing checkpoint is used as the split point. The route segments that have been traversed before the split point are deleted, and the route segments after the split point are retained to obtain the remaining pre-simulation route.

[0136] Step 402: Place the missing map module in the 3D physics prediction sandbox for simulation to determine the possible location of the missing map module on the remaining pre-simulation route.

[0137] The possible locations are the spatial positions in real physical space of all missing map modules obtained through physical simulation analysis within the remaining pre-simulation route range. By placing the missing map modules as objects to be matched into a 3D physical prediction sandbox, and scanning and arranging them along the remaining pre-simulation route, several possible locations that can meet the spatial constraints are obtained. The spatial constraints include not overlapping with other map modules and not exceeding the load-bearing support range.

[0138] Step 403: If the possible location does not exist, continue to perform map scanning and construction operations according to the remaining pre-simulation route.

[0139] If a possible location does not exist, it indicates that no missing map module location that meets the spatial constraints has been found within the remaining predicted route, meaning that the possible location is marked as non-existent.

[0140] Step 404: If possible locations exist, update the 3D physical prediction sandbox using all possible locations and missing map modules, update the remaining pre-simulation routes, and update the missing map modules according to possible locations until the image features and current appearance features partially or completely match.

[0141] If a possible location exists, it indicates that a missing map module location that satisfies the spatial constraints has been found within the remaining predicted route. Updating the missing map module according to the possible location refers to placing the missing map module at a possible location in the 3D physical prediction sandbox and then updating the location parameters of the 3D physical prediction sandbox.

[0142] Step 405: When the image features and the current appearance features partially or completely match, update the corresponding possible position to the actual position and delete all other possible positions from the 3D physical prediction sandbox.

[0143] When the image features and the current appearance features partially or completely match, it means that the image features extracted from the captured image correspond to some or all of the features of the missing map module stored in the 3D physical prediction sandbox. The actual location is the true and accurate coordinate position of the missing map module in the real physical space. When the image features corresponding to a possible location partially or completely match the current appearance features, the coordinate value of that possible location is confirmed as the actual location of the missing map module.

[0144] Among them, if possible locations exist, the method of updating the 3D physics prediction sandbox using all possible locations and missing map modules, updating the remaining pre-simulation routes, and updating the missing map modules according to possible locations includes:

[0145] Step 500: Obtain the theoretical location of the missing map module on the remaining pre-rehearsal route.

[0146] Reference Figure 4 In this diagram, curve C represents the current pre-simulated route, curve F represents the remaining pre-simulated route, dashed box L represents the blind zone area, dashed box M represents the possible location of the current map module, point N represents the location for verification based on the verification distance of the current map module's possible location, point J represents the missing verification location, and point S represents the location after reversing by the backtracking distance. The arrows on the curves indicate the forward direction of the embodied intelligent robot. The theoretical location of the module refers to the position where the missing map module should appear on the remaining pre-simulated route, which is directly obtained from the original location coordinates of the missing map module in the 3D physics prediction sandbox.

[0147] Step 501: Based on the module's theoretical location, the remaining pre-simulation route is split to obtain the blind zone route area before the module's theoretical location.

[0148] The blind spot route area refers to the shooting blind spot on the remaining pre-rehearsed route that the camera cannot detect when the embodied intelligent robot is checking the position and checking the current map module. The shooting range is determined by the rotation angle range of the camera, thus obtaining the blind spot route area.

[0149] Step 502: If the possible location does not fall into the blind spot route area, then update the 3D physics prediction sandbox using all possible locations and missing map modules, update the remaining pre-simulation routes, and update the missing map modules according to the possible locations.

[0150] If the possible location does not fall within the blind route area, it indicates that the possible location of the missing map module does not overlap with the blind route area. A parameter comparison between the possible location and the blind route area is used to determine their overlap. If the possible location does not overlap with the blind route area, it is determined that the possible location does not fall within the blind route area.

[0151] Step 503: If the possible location falls into the blind spot route area, determine the backtracking distance of the embodied intelligent robot based on the possible location and the verification distance, backtrack according to the backtracking distance, and then update the remaining rehearsal route based on the backtracking distance.

[0152] If a possible location falls within the blind spot route area, it indicates that the possible location of the missing map module overlaps with the blind spot route area. The overlap is determined by comparing parameters between the possible location and the blind spot route area. If they overlap, the possible location is considered to have fallen into the blind spot route area. The backtracking distance is the distance the embodied intelligent robot needs to backtrack to make the current blind spot route area observable when the possible location of the missing map module falls into the blind spot route area. This distance is determined by both the possible location and the verification distance, calculated as the route length from the current location to the possible location plus the verification distance.

[0153] This includes an optimized search method for the current map module when there is no local or complete match between the image features and the current appearance features. This method includes:

[0154] Step 600: Analyze the 3D physics prediction sandbox based on the current location to obtain the occlusion modules that can completely occlude the current appearance features and the occlusion locations of the missing map modules.

[0155] Reference Figure 5In this diagram, point A represents the location where the verification is performed according to the verification distance, module B is the current map module, curve C is the current pre-movement route, curve F is the remaining pre-movement route, module O is the occlusion module, module P is the additional occlusion module, point Q is the visible position, point R is the verification position, and the arrows on the curves indicate the direction of movement of the embodied intelligent robot. An occlusion module refers to a map module located between the current position and the missing map module in the 3D physics prediction sandbox, causing the image obtained from capturing the missing map module to occlude it. The occlusion module parameters are obtained by capturing and analyzing images using a camera. The occlusion position refers to the location where the missing map module is occluded by the occlusion module, preventing the embodied intelligent robot from observing the module's appearance from its current position. This position is obtained by reading the current coordinates of the missing map module and the coordinates of the occluded portion from the 3D physics prediction sandbox.

[0156] Step 601: Determine the visible location of the missing map module on the remaining pre-play route based on the occlusion positions of the occlusion module and the missing map module.

[0157] The visible location refers to the spatial location where the embodied intelligent robot can see the missing map module after moving on the remaining pre-play route. In the 3D physics prediction sandbox, the occluded location is the observation target. Images are taken and analyzed on the remaining pre-play route. If an observation point appears that is not blocked by any occluded module, then the location of that observation point is the visible location.

[0158] Step 602: Simulate in a 3D physics prediction sandbox based on the visible position of the embodied intelligent robot and the occlusion position of the missing map module to obtain additional occlusion modules.

[0159] Additional occlusion modules are other map modules that can obstruct the view of the missing map module after the embodied intelligent robot moves to a visible position and is analyzed in the 3D physics prediction sandbox. The scanning range is taken from the visible position to the occlusion position of the missing map module. If a map module is detected that intersects with the ray and is located between the viewpoint and the occlusion position, the module is marked as an additional occlusion module.

[0160] Step 603: Filter out the visible locations where the additional occlusion module does not exist, and define the visible location as the verification location.

[0161] Step 604: Control the embodied intelligent robot to move to the verification position to obtain the verification image, and define the image features corresponding to the verification image as the verification image features.

[0162] The verification image is captured by a camera after the embodied intelligent robot moves to the verification position. Verification image features refer to the key visual information extracted through image processing that describes the content of the image.

[0163] Step 605: When the checked image features do not match the current appearance features, return to the missing check position and re-execute steps 401 to 404.

[0164] When the verification image features do not match the current appearance features, it indicates that the matching result after comparing the verification image features extracted from the verification image with the current appearance features stored in the 3D physical prediction sandbox does not meet the standard of local matching or complete matching.

[0165] Step 606: When the image features and the current appearance features are partially or completely matched, the missing map module is updated by the occlusion location.

[0166] When the verification image features and the current appearance features partially or completely match, it indicates that the verification image features extracted from the verification image and the current appearance features stored in the 3D physical prediction sandbox have reached the standard of partial or complete matching.

[0167] This also includes a method for finding the current map module if no image feature and the current appearance feature can be completely matched, but a partial match exists. This method includes:

[0168] Step 700: Obtain the shooting angle when the image features and current appearance features are locally matched.

[0169] The shooting angle refers to the camera's posture angle relative to the robot when the image features and the current appearance features reach a "local matching" state. This angle is read and recorded in real time by the embodied intelligent robot's environmental sensors.

[0170] Step 701: Analyze the 3D physical prediction sandbox based on the shooting angle and missing check location to obtain a local occlusion module that can locally occlude the current appearance features.

[0171] The partial occlusion module refers to a map module in the 3D physics prediction sandbox that can partially occlude the current appearance features of missing map modules. It obtains the parameters of the partial occlusion module by capturing images with a camera and parsing the image information.

[0172] Step 702: Control the embodied intelligent robot to move to the verification position corresponding to the partial occlusion module to perform verification and execute steps 604 to 606.

[0173] The check position corresponding to the partial occlusion module refers to the new observation point that the embodied intelligent robot moves to in order to avoid the partial occlusion module. If the partial occlusion module is not detected at the observation point, the position is marked as the check position corresponding to the partial occlusion module.

[0174] The specific methods for finding the corresponding deceleration operation for the embodied intelligent robot based on ground material characteristics and controlling the embodied intelligent robot to execute the deceleration operation include:

[0175] Step 800: Obtain the collision route based on the current location and the verification distance.

[0176] The collision route refers to the segment of the path in which the embodied intelligent robot will continue to move forward and eventually collide with the map module ahead if it does not take deceleration measures while moving along the pre-rehearsed route. The collision route is a straight or curved path segment that extends forward along the pre-rehearsed route with the current position as the starting point and the verification distance as the length.

[0177] Step 801: Find the coefficient of friction of each ground material characteristic one by one according to the ground material characteristics.

[0178] Step 802: Based on the friction coefficient and the verification distance, obtain the deceleration rate of the ground material characteristics on the collision route.

[0179] Deceleration rate refers to the deceleration value that an embodied intelligent robot needs to adopt on specific ground material characteristics to avoid collisions with the map module in front. The formula for calculating the deceleration rate is: ,in For deceleration rate, For the current speed, For the distance between the embodied intelligent robot and the map module, For the ground friction coefficient, This is the acceleration due to gravity.

[0180] Step 803: Obtain the different ground material features that appear on the collision route based on 3D physical prediction sandbox analysis.

[0181] Step 804: Determine the output torque based on the friction coefficient and deceleration rate.

[0182] Output torque refers to the torque value that the drive motor or joint actuator of the embodied intelligent robot needs to output. The calculation formula is as follows: ,and ,in For braking force, For the quality of embodied intelligent robots, For deceleration rate, For the ground friction coefficient, This is the acceleration due to gravity.

[0183] Step 805: Control the embodied intelligent robot to decelerate according to the output torque on ground surfaces with different material characteristics.

[0184] Among them, the optimized method for controlling the embodied intelligent robot to decelerate according to the deceleration rate when the pre-simulated motion situation is a collision situation includes:

[0185] Step 900: Obtain the second safe collision constraint value of the current map module through the 3D physical prediction sandbox.

[0186] The second safe collision constraint value refers to the maximum allowable impact force or maximum allowable impact pressure value of the current map module, which can be obtained by directly querying the physical parameters stored in the 3D physical prediction sandbox.

[0187] Step 901: Compare the preset first safe collision constraint value and the second safe collision constraint value to obtain the maximum safe collision constraint value. The maximum safe collision constraint value is the minimum of the first safe collision constraint value and the second safe collision constraint value.

[0188] The first safe collision constraint value is the maximum permissible impact force or maximum permissible impact pressure value related to the embodied intelligent robot itself, which is directly written into the 3D physics prediction sandbox as a fixed parameter. The maximum safe collision constraint value refers to the maximum impact pressure value that the embodied intelligent robot can withstand on the contact surface when it collides with the current map module, and is obtained by comparing the smaller value between the first safe collision constraint value and the second safe collision constraint value.

[0189] Step 902: Update the deceleration rate based on the maximum safe collision constraint to obtain the actual required deceleration rate. The actual required deceleration rate is the pressure value generated when the embodied intelligent robot moves forward at the actual required deceleration rate and collides with the current map module, which is less than the maximum safe collision constraint value.

[0190] The actual required deceleration rate refers to the final deceleration rate obtained after correcting and updating the initially determined deceleration rate by introducing the maximum safe collision constraint value as an additional physical constraint. The calculation formula is as follows: ,in For the final deceleration rate, For the current speed, For the distance between the embodied intelligent robot and the map module, For the ground friction coefficient, For gravitational acceleration, For the maximum safe collision constraint value, For the quality of embodied intelligent robots, This refers to the effective contact area between the robot's front end and the object.

[0191] Step 903: Control the embodied intelligent robot to decelerate according to the actual required deceleration rate.

[0192] Deceleration operation refers to the specific sequence of control actions executed according to the actual required deceleration rate. The actual required deceleration rate is converted into an output torque that the actuator can execute, and the underlying motion control module executes the instruction.

[0193] Based on the same inventive concept, embodiments of the present invention provide a static spatial physical prior map construction system for embodied intelligence.

[0194] A static spatial-physical prior map construction system for embodied intelligence, comprising:

[0195] The acquisition module is used to acquire map building signals, spatial 2D floor plan drawings, module features, 3D physical prediction sandbox, current location, current orientation, current distance, camera rotation angle, environmental sensors, pre-simulated route and captured images;

[0196] A memory for storing a program for constructing a static spatial-physical prior map of embodied intelligence;

[0197] The processor loads and executes programs from memory.

[0198] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0199] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for constructing a static spatial physical prior map of embodied intelligence, characterized in that, include: Step 100: In response to the map building signal, import the preset 2D floor plan; Step 101: Perform feature analysis on the 2D floor plan to obtain module features that conform to preset text features or preset image fill features; Step 102: After matching the module features one by one according to the preset standard map module features, a three-dimensional physical prediction sandbox is established through the successfully matched module features. The standard map module features have map object parameters, which include surface friction, stiffness, acoustics, load-bearing limit and mechanical parameters. Step 103: Simulate the motion of the embodied intelligent robot in the 3D physics prediction sandbox to obtain the pre-execution route, and perform the preset map scanning and construction operation in the 3D physics prediction sandbox according to the pre-execution route; Step 104: During the execution of the preset map scanning and construction operation, obtain the current location and current orientation; Step 105: Input the current position and current orientation into the three-dimensional physics prediction sandbox to conduct embodied intelligence motion simulation exercise to obtain the pre-movement situation, which includes collision situation and unobstructed situation; Step 106: When the pre-simulated motion is a collision, retrieve the ground material features of the current position in the 3D physics prediction sandbox; Step 107: Based on the characteristics of the ground material, find the deceleration operation corresponding to the embodied intelligent robot, and control the embodied intelligent robot to execute the deceleration operation; Step 108: When the pre-simulated movement is unobstructed, continue to perform map scanning and construction operations according to the pre-simulated route, and continue to obtain the current position and current orientation.

2. The method for constructing a static spatial physical prior map of embodied intelligence according to claim 1, characterized in that, Also includes: Step 200: Based on the current location, the pre-simulation route and the preset verification distance, determine the map module directly in front, define the map module as the current map module, and retrieve the map object parameters of the current map module in the 3D physical prediction sandbox, and define the map object parameters as the current parameters. Step 201: If the current parameters are the same as the actual parameters, continue to perform map scanning and construction operations according to the pre-simulated route, and continue to obtain the current position and current orientation; Step 202: If the current map module does not exist, continue to perform map scanning and construction operations according to the pre-rehearsed route, and continue to obtain the current position and current orientation.

3. The method for constructing a static spatial physical prior map of embodied intelligence according to claim 2, characterized in that, Also includes: Step 300: If the current map module exists but no actual parameters are detected, search for the current map module to obtain its current appearance features; Step 301: Rotate the camera in a preset rotation direction and acquire the captured image; the environmental sensor includes the camera. Step 302: Analyze the captured images to obtain image features; Step 303: When the image features and the current appearance features partially match or do not match, continue to rotate the camera until the image features and the current appearance features completely match; Step 304: When the image features and the current appearance features are a perfect match, obtain the camera rotation angle and analyze the current distance between the current map module and the embodied intelligent robot based on the image features; Step 305: Re-detect the actual parameters of the current map module in real time through the environmental sensor, and correct the actual parameters based on the camera rotation angle and the current distance to obtain the movement parameters; Step 306: Update the movement parameters in the 3D physics prediction sandbox and update the simulation route.

4. The method for constructing a static spatial physical prior map of embodied intelligence according to claim 3, characterized in that, It also includes a method for searching the current map module when there is no partial or complete match between the image features and the current appearance features. This method includes: Step 400: Define the current map module as a missing map module and the current location as a missing verification location; Step 401: Based on the missing checkpoint and the rehearsal route, determine the rehearsal route after the missing checkpoint and define the rehearsal route as the remaining rehearsal route; Step 402: Place the missing map module in the 3D physics prediction sandbox for simulation to determine the possible location of the missing map module on the remaining pre-simulation route; Step 403: If the possible location does not exist, continue to perform map scanning and construction operations according to the remaining pre-simulation route; Step 404: If possible locations exist, update the 3D physical prediction sandbox using all possible locations and missing map modules, update the remaining pre-simulation routes, and update the missing map modules according to possible locations until the image features and current appearance features partially or completely match. Step 405: When the image features and the current appearance features partially or completely match, update the corresponding possible position to the actual position and delete all other possible positions from the 3D physical prediction sandbox.

5. The method for constructing a static spatial physical prior map of embodied intelligence according to claim 4, characterized in that, If possible locations exist, the method for updating the 3D physics prediction sandbox using all possible locations and missing map modules, updating the remaining simulation routes, and updating the missing map modules according to possible locations includes: Step 500: Obtain the theoretical location of the missing map module on the remaining pre-rehearsal route; Step 501: Based on the module's theoretical location, the remaining pre-simulation route is divided to obtain the blind zone route area before the module's theoretical location; Step 502: If the possible location does not fall into the blind spot route area, then update the 3D physics prediction sandbox with all possible locations and missing map modules, update the remaining pre-simulation route and update the missing map modules according to the possible locations; Step 503: If the possible location falls into the blind spot route area, determine the backtracking distance of the embodied intelligent robot based on the possible location and the verification distance, backtrack according to the backtracking distance, and then update the remaining rehearsal route based on the backtracking distance.

6. The method for constructing a static spatial physical prior map of embodied intelligence according to claim 4, characterized in that, It also includes an optimized search method for the current map module when there is no partial or complete match between the image features and the current appearance features. This method includes: Step 600: Analyze the 3D physics prediction sandbox based on the current location to obtain the occlusion module that can completely occlude the current appearance features and the occlusion location of the missing map module. Step 601: Determine the visible locations of the missing map module on the remaining pre-playback route based on the occlusion positions of the occlusion module and the missing map module. The visible locations are several. Step 602: Simulate the occlusion module in the 3D physics prediction sandbox based on the visible position of the embodied intelligent robot and the occlusion position of the missing map module; Step 603: Filter out the visible locations where the additional occlusion module does not exist, and define the visible location as the verification location; Step 604: Control the embodied intelligent robot to move to the verification position to obtain the verification image, and define the image features corresponding to the verification image as the verification image features; Step 605: When the checked image features do not match the current appearance features, return to the missing check position and re-execute steps 401 to 404; Step 606: When the image features and the current appearance features are partially or completely matched, the missing map module is updated by the occlusion location.

7. The method for constructing a static spatial physical prior map of embodied intelligence according to claim 6, characterized in that, It also includes a method for searching the current map module if no image feature and the current appearance feature can be completely matched, but a partial match exists. This method includes: Step 700: Obtain the shooting angle when locally matching image features and current appearance features; Step 701: Analyze the 3D physical prediction sandbox based on the shooting angle and missing verification location to obtain a local occlusion module that can locally occlude the current appearance features; Step 702: Control the embodied intelligent robot to move to the verification position corresponding to the partial occlusion module to perform verification and execute steps 604 to 606.

8. The method for constructing a static spatial physical prior map of embodied intelligence according to claim 2, characterized in that, The specific methods for finding the corresponding deceleration operation for the embodied intelligent robot based on ground material characteristics and controlling the embodied intelligent robot to execute the deceleration operation include: Step 800: Obtain the collision route based on the current location and the verification distance; Step 801: Find the coefficient of friction for each surface material characteristic one by one; Step 802: Based on the friction coefficient and the verification distance, obtain the deceleration rate of the ground material characteristics on the collision route; Step 803: Based on 3D physics prediction sandbox analysis, obtain the different ground material features that appear on the collision route; Step 804: Determine the output torque based on the friction coefficient and deceleration rate; Step 805: Control the embodied intelligent robot to decelerate according to the output torque on ground surfaces with different material characteristics.

9. A method for constructing a static spatial physical prior map of embodied intelligence according to claim 8, characterized in that, An optimized method for controlling an embodied intelligent robot to decelerate according to a deceleration rate when the pre-simulated motion scenario is a collision, the method includes: Step 900: Obtain the second safe collision constraint value of the current map module through the 3D physics prediction sandbox; Step 901: Compare the preset first safe collision constraint value and the second safe collision constraint value to obtain the maximum safe collision constraint value, wherein the maximum safe collision constraint value is the minimum of the first safe collision constraint value and the second safe collision constraint value; Step 902: Update the deceleration rate based on the maximum safe collision constraint to obtain the actual required deceleration rate, wherein the actual required deceleration rate is the pressure value generated when the embodied intelligent robot moves forward at the actual required deceleration rate and collides with the current map module, which is less than the maximum safe collision constraint value. Step 903: Control the embodied intelligent robot to decelerate according to the actual required deceleration rate.

10. A static spatial-physical prior map construction system for embodied intelligence, characterized in that, include: The acquisition module is used to acquire map building signals, spatial 2D floor plan drawings, module features, 3D physical prediction sandbox, current location, current orientation, current distance, camera rotation angle, environmental sensors, pre-simulated route and images; A memory for storing a program for constructing a static spatial physical prior map of embodied intelligence as described in any one of claims 1 to 9; The processor loads and executes programs from memory.

Images