Body-aware robot control method and system

Abstract

The present application relates to the technical field of robot control, in particular to a body robot control method and system, comprising: obtaining semantic map and task instruction information of a target body robot in a target scene, and extracting navigation strategy parameters and operation strategy parameters corresponding to each functional area from the semantic map; determining target moving speed of the target body robot in each functional area according to the task instruction information and the navigation strategy parameters. The present application extracts the navigation strategy parameters and the operation strategy parameters corresponding to each functional area from the semantic map, dynamically determines the target moving speed and target operation strength of each area in combination with the task instruction information, and performs spatio-temporal alignment and fusion of the two to generate a continuous behavior sequence, so that the robot can automatically adjust the behavior according to the area semantics, and the efficiency and safety of task execution are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, specifically to a method and system for controlling an embodied robot. Background Technology

[0002] Currently, traditional methods typically employ a globally fixed speed or simple reactive obstacle avoidance, failing to dynamically adjust movement speed based on semantic information of different functional areas within the scene. Robots are prone to collisions in narrow areas due to excessive speed, while in open areas, their efficiency may be low due to insufficient speed. Furthermore, they cannot dynamically update cost map weights using obstacle approach frequency, resulting in rigid obstacle avoidance strategies.

[0003] Furthermore, traditional operations often employ constant force or simple pose control, failing to adjust the force in real time based on the target object's material, deformation characteristics, and feedback from the end effector's force sensor. When handling fragile or elastic objects, excessive force can easily cause damage, while insufficient force can lead to grasping failure. Moreover, navigation and operational behavior are disconnected, failing to form a continuous sequence of actions aligned in time and space, resulting in poor overall task adaptability and safety. Summary of the Invention

[0004] To achieve the above objectives, a control method and system for embodied robots are provided.

[0005] The first aspect is the control method for embodied robots, including: Acquire semantic map and task instruction information of the target embodied robot in the target scene, and extract navigation strategy parameters and operation strategy parameters corresponding to each functional area from the semantic map; Based on the task instruction information and the navigation strategy parameters, the target movement speed of the target embodied robot in each of the functional areas is determined, and based on the operation strategy parameters and the target object position information obtained by the visual sensor of the target embodied robot, the target operation force of the target embodied robot in each of the functional areas is determined. The target movement speed and target operation force of each functional area are spatiotemporally aligned and fused to obtain the target behavior sequence of the target embodied robot in the target scene.

[0006] Preferably, the process of determining the navigation strategy parameters and the operation strategy parameters includes: The policy parameters that match the semantic labels of each functional region in the pre-trained policy library are used as the navigation policy parameters and the operation policy parameters; and / or, The strategy parameters in the pre-trained strategy library are adjusted based on the execution effect data of each functional area in the historical execution record of the target embodied robot to obtain the navigation strategy parameters and the operation strategy parameters.

[0007] Preferably, when the navigation mode is a semantic obstacle avoidance navigation mode, obtaining the target movement speed of the target embodied robot in the functional area where the navigation mode is semantic obstacle avoidance navigation, based on the task instruction information and the navigation strategy parameters, includes: The navigation strategy parameters and the obstacle semantic information of each functional area in the semantic map are input into the semantic cost map model with dynamically adjusted weights to obtain the initial speed adjustment amount of each functional area; Based on the traffic constraint relationship between the functional area in the semantic obstacle avoidance navigation mode and the corresponding functional area in the semantic map, and the initial speed adjustment, the target moving speed of the target embodied robot in the functional area in the semantic obstacle avoidance navigation mode is obtained.

[0008] Preferably, the method for dynamically adjusting the weights of the semantic cost map model includes: Monitor the approach frequency of the target android to obstacles in each of the aforementioned functional areas; The weights of the semantic cost map model are updated based on the proximity frequency.

[0009] Preferably, monitoring the approach frequency of the target android to obstacles in each of the functional areas includes: The obstacle distance change rate is determined based on the distance sensor data of the target android, and the movement speed change rate is determined based on the odometry data of the target android. The approach frequency is obtained by calculating the ratio of the rate of change of the obstacle distance to the rate of change of the moving speed.

[0010] Preferably, updating the weights of the semantic cost map model based on the proximity frequency includes: Obtain the mapping relationship between the weights of at least one semantic category in the semantic cost map model and the proximity frequency; The weights of the semantic cost map model are updated based on the mapping relationship.

[0011] Preferably, when the operation mode is a flexible operation mode, the step of obtaining the target operation force of the target robot in the functional area where the operation mode is flexible, based on the operation strategy parameters and the target object position information obtained by the target robot's visual sensor, includes: The real-time deformation of the contact point is determined based on the target object's position information and the force sensor data of the end effector of the target android. The operation deviation is determined based on the real-time deformation and the operation strategy parameters, and the operation deviation is input into the parameter adaptive impedance control model to obtain the initial force adjustment amount for each functional area; Based on the operational task constraints corresponding to each functional area and the initial force adjustment amount, the target operational force of the target embodied robot in the functional area where the operational mode is flexible is obtained.

[0012] Secondly, this application also provides a hymenoidery robot control system, including: The parameter extraction unit is configured to acquire the semantic map and task instruction information of the target embodied robot in the target scene, and extract the navigation strategy parameters and operation strategy parameters corresponding to each functional area from the semantic map. An automatic control unit is configured to determine the target movement speed of the target embodied robot in each of the functional areas based on the task instruction information and the navigation strategy parameters, and to determine the target operation force of the target embodied robot in each of the functional areas based on the operation strategy parameters and the target object position information obtained by the visual sensor of the target embodied robot. The spatiotemporal fusion unit is configured to spatiotemporally align and fuse the target movement speed and target operation force of each functional area to obtain the target behavior sequence of the target embodied robot in the target scene.

[0013] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the embodied robot control method.

[0014] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the embodied robot control method.

[0015] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention extracts navigation strategy parameters and operation strategy parameters corresponding to each functional area from the semantic map, dynamically determines the target movement speed and target operation force of each area by combining task instruction information, and aligns and fuses the two in time and space to generate a continuous behavior sequence, so that the robot can automatically adjust its behavior according to the semantics of the area, which significantly improves the efficiency and safety of task execution and avoids collisions or operation errors caused by blindly unifying speed or force. (2) In terms of navigation, the present invention adopts a semantic obstacle avoidance navigation mode and dynamically adjusts the weight of the semantic cost map model based on the proximity frequency, so that the robot can decelerate and detour around high-frequency obstacle areas in advance; in terms of operation, a flexible operation mode is adopted, and the operation force is adaptively adjusted through real-time deformation and impedance control to meet the operation constraints of different objects, effectively improving the adaptability and robustness of the robot in dynamic changing environments, while protecting fragile objects and its own structure and reducing the risk of damage. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart of the overall method in one embodiment of the present invention; Figure 2 This is a schematic diagram of the overall system architecture in one embodiment of the present invention; Figure 3 This is an internal structural diagram of a computer device according to one embodiment of the present invention; Figure 4 This is an internal structural diagram of a computer device according to another embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1 This invention provides a technical solution: a method for controlling an embodied robot, comprising: S1. Obtain the semantic map and task instruction information of the target embodied robot in the target scene, and extract the navigation strategy parameters and operation strategy parameters corresponding to each functional area from the semantic map. S2. Based on the task instruction information and navigation strategy parameters, determine the target movement speed of the target embodied robot in each functional area, and based on the operation strategy parameters and the target object position information obtained by the target embodied robot's vision sensor, determine the target operation force of the target embodied robot in each functional area. S3. After spatiotemporally aligning and fusing the target movement speed and target operation force of each functional area, the target behavior sequence of the target embodied robot in the target scene is obtained.

[0019] The embodied robot performs the task of "tidying up the coffee table in the living room" in a home environment. First, the robot builds a semantic map using LiDAR and visual SLAM, marking functional areas such as the sofa area, coffee table area, TV cabinet area, and aisle area. Navigation strategy parameters for each area are extracted from the map. For example, the coffee table area is a fine operation area with a maximum movement speed limit of 0.2 meters per second; the aisle area is a fast passage area with a maximum speed of 0.8 meters per second. Operation strategy parameters include: the coffee table area requires gentle handling with a maximum force of 5 Newtons; the TV cabinet area is a fragile item placement area with a maximum force of 3 Newtons.

[0020] The task instruction is "Place the cup on the coffee table onto the TV cabinet". Based on the instruction and navigation strategy parameters, the robot calculates its moving speed in the aisle area to be 0.6 meters per second, and decelerates to 0.1 meters per second when approaching the target object in the coffee table area. At the same time, the vision sensor locates the position of the cup, and according to the operation strategy parameters, it uses a force of 2 Newtons when grasping the cup in the coffee table area and a force of 1 Newton when moving it to the TV cabinet area.

[0021] The robot aligns the movement speed curves and operation force curves of each area along the time axis: for example, it moves through the aisle at 0.6 meters per second in 0-3 seconds, decelerates to 0.1 meters per second to approach the cup in 3-5 seconds, applies 2 Newtons of force to grab it at 5 seconds, then moves towards the TV cabinet at 0.4 meters per second, decelerates to 0.1 meters per second at 8 seconds, and applies 1 Newton of force to place it at 9 seconds; finally, a complete sequence of behaviors is formed, which the robot executes in sequence to complete the sorting task.

[0022] The process of determining navigation strategy parameters and operational strategy parameters includes: Use the policy parameters from the pre-trained policy library that match the semantic labels of each functional region as navigation policy parameters and operation policy parameters; and / or, Based on the execution effect data of each functional area in the historical execution record of the target embodied robot, the strategy parameters in the pre-trained strategy library are adjusted to obtain navigation strategy parameters and operation strategy parameters.

[0023] The embodied robot performs the task of "putting the bowls on the dining table into the dishwasher" in a home environment. The semantic map includes functional areas such as the dining table area, aisle area, and dishwasher area. First, the robot matches the policy parameters corresponding to the semantic labels of these areas from a pre-trained policy library. For example, the navigation policy parameter for the dining table area is a movement speed of 0.1 meters per second to avoid colliding with tableware, and the operation policy parameter is a grasping force of 2 Newtons. The navigation policy parameter for the dishwasher area is a movement speed of 0.05 meters per second, and the operation policy parameter is a pushing and pulling force of 8 Newtons. The robot directly uses these parameters.

[0024] In another scenario, the robot had multiple historical execution records within the same household. Previously, when grasping ceramic bowls in the dining area, a force of 2 Newtons caused the bowls to slip, and the execution data indicated that 3 Newtons were needed for a stable grasp. Therefore, the parameters in the pre-trained strategy library were adjusted based on historical data, updating the grasping force in the dining area's operation strategy parameters to 3 Newtons. Simultaneously, in the dishwasher area, due to track aging, a push-pull force of 8 Newtons frequently caused jamming; historical records showed that 10 Newtons were needed for smooth pushing. Therefore, the operation force in the dishwasher area was adjusted to 10 Newtons. Regarding navigation speed, a collision with a child occurred at 0.6 meters per second when passing through the aisle area; historical records suggested reducing the speed to 0.4 meters per second, and the navigation strategy parameters for the aisle area were adjusted accordingly.

[0025] The two methods can be combined: for the dining area, there are both pre-trained parameters and historical adjustments, with the historical adjustments taking precedence; for the storage area, which is being entered for the first time, there are no historical records, so the matching parameters from the pre-trained strategy library are used directly; finally, the navigation and operation strategy parameters for each functional area are determined.

[0026] When the navigation mode is semantic obstacle avoidance navigation mode, the target moving speed of the embodied robot in the functional area where the navigation mode is semantic obstacle avoidance navigation mode is obtained based on the task instruction information and navigation strategy parameters, including: The navigation strategy parameters and obstacle semantic information of each functional area in the semantic map are input into the semantic cost map model with dynamically adjusted weights to obtain the initial speed adjustment amount of each functional area. Based on the functional areas in the semantic obstacle avoidance navigation mode, the traffic constraints corresponding to each functional area in the semantic map, and the initial speed adjustment, the target embodied robot's target movement speed in the functional areas in the semantic obstacle avoidance navigation mode is obtained.

[0027] The embodied robot performs the task of "walking from the sofa area to the coffee table area" in the living room. The navigation mode is set to semantic obstacle avoidance navigation mode. The semantic map marks functional areas such as the sofa area, coffee table area, carpet area, floor lamp area, etc. Each area is accompanied by obstacle semantic information. For example, the carpet area is a low friction and easy slippery area, and the floor lamp area is a fragile and high-risk area. The robot acquires navigation strategy parameters, such as a maximum speed of 0.8 meters per second and a safe distance threshold of 0.5 meters. These parameters and the semantic map are then input into a semantic cost map model with dynamically adjusted weights. The model automatically adjusts the cost weights based on obstacle semantics: the sliding risk weight for the carpet area is set to 2.0, the collision cost weight for the floor lamp area is set to 3.0, and the weight for the ordinary floor area is 1.0. After calculation, the model outputs the initial speed adjustment for each area: the speed in the carpet area is reduced by 30%, the speed in the floor lamp area is reduced by 50%, and the speed in the ordinary area remains unchanged. The robot is currently located in the sofa area and needs to pass through a short section of carpet area and floor lamp area to reach the coffee table area. The passage constraint between the areas is: it must pass through the carpet area to reach the coffee table area, but it can detour to avoid the floor lamp area. Based on this constraint and the initial speed adjustment, the final target moving speed is calculated. The robot decides to detour around the floor lamp area, so it is only affected by the carpet area: the initial speed of 0.8 m / s is reduced by 30% to 0.56 m / s. Considering the task instruction of "quick arrival", the model allows it to increase to 0.65 m / s in non-high-risk areas. The final target moving speed is set at 0.65 m / s, and the robot can safely reach the coffee table area along the detour path at this speed. If it cannot detour, it must be reduced to 0.56 m / s, and obstacle avoidance monitoring is added.

[0028] Methods for dynamically adjusting weights in semantic cost map models include: Monitor the frequency of approach between the target embodied robot and obstacles in each functional area; The weights of the semantic cost map model are updated based on proximity frequency.

[0029] The embodied robot performs handling tasks in a warehouse environment. The semantic map includes shelving areas, aisle areas, and charging station areas. It monitors the robot's approach frequency to obstacles in each functional area. For example, in the shelving area, the robot approaches the shelf edge an average of 10 times per hour; in the aisle area, it approaches the wall an average of 2 times per hour; and in the charging station area, it approaches the charging station an average of 1 time per hour. Initially, the semantic cost map model sets the collision cost weight for the shelving area to 1.0, the aisle area to 0.5, and the charging station area to 0.2. Based on the approach frequency, the weights are dynamically adjusted: a high approach frequency in the shelving area means frequent obstacle avoidance, so the weight is increased to 2.0, making the model penalize approaching obstacles more severely in this area, causing the robot to slow down or detour in advance; a low approach frequency in the aisle area reduces the weight to 0.3, allowing the robot to pass through at a higher speed; and an extremely low approach frequency in the charging station area reduces the weight to 0.1, allowing the robot to ignore minor approaches. The adjusted weights are updated in real-time in the semantic cost map model for subsequent speed adjustments, enabling the robot's obstacle avoidance behavior to adapt to changes in the actual operating environment.

[0030] Monitoring the approach frequency of the target embodied robot to obstacles in various functional areas, including: The rate of change of obstacle distance is determined based on the distance sensor data of the target embodied robot, and the rate of change of movement speed is determined based on the odometry data of the target embodied robot. The approach frequency is obtained by calculating the ratio of the rate of change of obstacle distance to the rate of change of moving speed.

[0031] The embodied robot moves through the warehouse shelving area, with distance sensors continuously measuring the distance to the edge of the shelving. Within 0.2 seconds, the distance decreases from 0.8 meters to 0.6 meters, and the obstacle distance change rate is calculated to be a decrease of 1.0 meter per second. Simultaneously, the odometry records that the robot's moving speed increases from 0.3 meters per second to 0.5 meters per second, and the moving speed change rate is an increase of 1.0 meters per second. The ratio of the obstacle distance change rate to the moving speed change rate is calculated, i.e., -1.0 divided by +1.0 equals -1.0, and the absolute value is 1.0, yielding an approach frequency of 1.0 times per second. This value represents the rate at which the robot approaches obstacles per second. Based on this, the weights of the semantic cost map model are dynamically adjusted, for example, by increasing the collision cost weight in areas with high approach frequencies, allowing the robot to decelerate and avoid obstacles in advance.

[0032] The weights of the semantic cost map model are updated based on proximity frequency, including: Obtain the mapping relationship between the weights and proximity frequencies of at least one semantic category in the semantic cost map model; Update the weights of the semantic cost map model based on the mapping relationship.

[0033] The embodied robot operates in an office environment. The semantic cost map model contains three semantic categories: table, chair, and whiteboard. A mapping relationship between the weight and proximity frequency of each category is pre-established. For example, the weight is 0.5 when the proximity frequency is less than 0.2 times per second, 1.0 when it is between 0.2 and 0.5 times per second, and 2.0 when it is greater than 0.5 times per second. The robot monitors the proximity frequency to the table in real time. It is 0.6 times per second, which is greater than 0.5, so the weight of the table is updated to 2.0. The proximity frequency to the chair is 0.3 times per second, so the weight is updated to 1.0. The proximity frequency to the whiteboard is 0.1 times per second, so the weight is updated to 0.5. The updated semantic cost map model is used for subsequent speed adjustment, so that the robot can avoid the table with high-frequency proximity in advance and pass through the whiteboard with low-frequency proximity quickly.

[0034] When the operation mode is flexible operation mode, based on the operation strategy parameters and the target object position information obtained by the target embodied robot's vision sensors, the target operation force of the target embodied robot in the functional area where the operation mode is flexible is obtained, including: The real-time deformation of the contact point is determined based on the target object's position information and the force sensor data of the end effector of the target embodied robot; The operational deviation is determined based on the real-time deformation and operational strategy parameters, and the operational deviation is input into the parameter adaptive impedance control model to obtain the initial force adjustment amount for each functional area. Based on the operational task constraints and initial force adjustment amount corresponding to each functional area, the target operational force of the target embodied robot in the functional area where the operation mode is flexible is obtained.

[0035] An embodied robot performs the task of "grabbing a fragile beaker" on a laboratory workbench, operating in a flexible mode. The robot's vision sensor acquires the beaker's position coordinates, while the force sensor of the end effector detects the contact force in real time. When the robotic gripper first contacts the beaker, the force sensor measures a contact force of 0.5 N, while the preset expected contact force is 0.3 N, resulting in a compressive deformation of 0.2 mm in the beaker wall, i.e., the real-time deformation. According to the operating strategy parameters, the allowable deformation limit for this beaker is 0.1 mm. The current deformation exceeds the limit, and the operating deviation is calculated as deformation. The difference is 0.1 mm. This deviation is input into an adaptive impedance control model. The model adjusts the impedance parameters adaptively based on the deviation. For example, if the stiffness is reduced from 1000 N / m to 600 N / m, the initial force adjustment is reduced by 0.2 N. Next, the operational task constraints are considered: the beaker is thin-walled glass and must be grasped smoothly to prevent slippage. Combining the constraints, the target operational force is finally set to 0.4 N. The robot grasps the beaker with this force without crushing it or slipping it. If it were in a metal part area, the task constraints would be different, and the target operational force would be increased accordingly.

[0036] Example 2, please refer to Figure 2 This invention provides a technical solution: a hymenoid robot control system, applicable to the aforementioned hymenoid robot control method, comprising: The parameter extraction unit is configured to acquire the semantic map and task instruction information of the target embodied robot in the target scene, and extract the navigation strategy parameters and operation strategy parameters corresponding to each functional area from the semantic map. The automatic control unit is configured to determine the target movement speed of the target embodied robot in each functional area based on task instruction information and navigation strategy parameters, and to determine the target operation force of the target embodied robot in each functional area based on operation strategy parameters and target object position information obtained by the target embodied robot's vision sensor. The spatiotemporal fusion unit is configured to spatiotemporally align and fuse the target movement speed and target operation force of each functional area to obtain the target behavior sequence of the target embodied robot in the target scene.

[0037] Example 3, please refer to Figure 3This invention provides a technical solution: a computer device, which can be a server. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network. When the computer program is executed by the processor, it implements a holographic robot control method.

[0038] Example 4, please refer to Figure 4 This invention provides a technical solution: a computer device, which can be a terminal. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a robot control method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0039] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for controlling an embodied robot, characterized in that, include: Acquire semantic map and task instruction information of the target embodied robot in the target scene, and extract navigation strategy parameters and operation strategy parameters corresponding to each functional area from the semantic map; Based on the task instruction information and the navigation strategy parameters, the target movement speed of the target embodied robot in each of the functional areas is determined, and based on the operation strategy parameters and the target object position information obtained by the visual sensor of the target embodied robot, the target operation force of the target embodied robot in each of the functional areas is determined. The target movement speed and target operation force of each functional area are spatiotemporally aligned and fused to obtain the target behavior sequence of the target embodied robot in the target scene.

2. The embodied robot control method according to claim 1, characterized in that, The process of determining the navigation strategy parameters and the operation strategy parameters includes: The policy parameters that match the semantic labels of each functional region in the pre-trained policy library are used as the navigation policy parameters and the operation policy parameters; and / or, The strategy parameters in the pre-trained strategy library are adjusted based on the execution effect data of each functional area in the historical execution record of the target embodied robot to obtain the navigation strategy parameters and the operation strategy parameters.

3. The embodied robot control method according to claim 2, characterized in that, When the navigation mode is semantic obstacle avoidance navigation mode, obtaining the target movement speed of the target embodied robot in the functional area where the navigation mode is semantic obstacle avoidance navigation mode, based on the task instruction information and the navigation strategy parameters, includes: The navigation strategy parameters and the obstacle semantic information of each functional area in the semantic map are input into the semantic cost map model with dynamically adjusted weights to obtain the initial speed adjustment amount of each functional area; Based on the traffic constraint relationship between the functional area in the semantic obstacle avoidance navigation mode and the corresponding functional area in the semantic map, and the initial speed adjustment, the target moving speed of the target embodied robot in the functional area in the semantic obstacle avoidance navigation mode is obtained.

4. The embodied robot control method according to claim 3, characterized in that, The method for dynamically adjusting the weights of the semantic cost map model includes: Monitor the approach frequency of the target android to obstacles in each of the aforementioned functional areas; The weights of the semantic cost map model are updated based on the proximity frequency.

5. The embodied robot control method according to claim 4, characterized in that, Monitoring the approach frequency of the target android to obstacles in each of the aforementioned functional areas includes: The obstacle distance change rate is determined based on the distance sensor data of the target android, and the movement speed change rate is determined based on the odometry data of the target android. The approach frequency is obtained by calculating the ratio of the rate of change of the obstacle distance to the rate of change of the moving speed.

6. The embodied robot control method according to claim 5, characterized in that, Updating the weights of the semantic cost map model based on the proximity frequency includes: Obtain the mapping relationship between the weights of at least one semantic category in the semantic cost map model and the proximity frequency; The weights of the semantic cost map model are updated based on the mapping relationship.

7. The embodied robot control method according to claim 6, characterized in that, When the operation mode is flexible operation mode, the step of obtaining the target operation force of the target robot in the functional area where the operation mode is flexible operation mode, based on the operation strategy parameters and the target object position information obtained by the target embodied robot's vision sensor, includes: The real-time deformation of the contact point is determined based on the target object's position information and the force sensor data of the end effector of the target android. The operation deviation is determined based on the real-time deformation and the operation strategy parameters, and the operation deviation is input into the parameter adaptive impedance control model to obtain the initial force adjustment amount for each functional area; Based on the operational task constraints corresponding to each functional area and the initial force adjustment amount, the target operational force of the target embodied robot in the functional area where the operational mode is flexible is obtained.

8. A hymen robot control system, applicable to the hymen robot control method according to any one of claims 1-7, characterized in that, include: The parameter extraction unit is configured to acquire the semantic map and task instruction information of the target embodied robot in the target scene, and extract the navigation strategy parameters and operation strategy parameters corresponding to each functional area from the semantic map. An automatic control unit is configured to determine the target movement speed of the target embodied robot in each of the functional areas based on the task instruction information and the navigation strategy parameters, and to determine the target operation force of the target embodied robot in each of the functional areas based on the operation strategy parameters and the target object position information obtained by the visual sensor of the target embodied robot. The spatiotemporal fusion unit is configured to spatiotemporally align and fuse the target movement speed and target operation force of each functional area to obtain the target behavior sequence of the target embodied robot in the target scene.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images