A method, system, and robot for recognizing robot body boundaries.

Abstract

This invention discloses a robot body boundary recognition method, system, and robot, relating to the interdisciplinary fields of embodied intelligence, biomimetic perception, robot autonomous control, and intelligent robot cognition. The method includes: constructing a robot body space mapping model, which at least defines the belonging space intervals of each component of the robot; driving the robot's movable parts to perform self-probing actions; during the self-probing actions, collecting motion signals from the movable parts and response signals from the touched object; and determining whether the touched object belongs to the robot body or the external environment based on whether the motion signal and response signal simultaneously match preset features within the belonging space interval. This invention can improve the robot's autonomy level and offers better scalability and long-term operational stability.

Description

Technical Field

[0001] This invention relates to the interdisciplinary fields of embodied intelligence, biomimetic perception, robot autonomous control, and intelligent robot cognition, specifically to a robot body boundary recognition method, system, and robot. Background Technology

[0002] Current robotics technologies largely employ a unidirectional "perception-execution" architecture. The perception, execution, and processing modules lack deep bidirectional interaction. The sensors used have low integration and lack a pre-defined 3D body coordinate system. They can only passively execute preset actions and identify external obstacles, unable to confirm the attribution of body parts and functional areas through autonomous actions and bidirectional signal verification. They also lack the core logic to distinguish between the robot's body and its external environment. Furthermore, traditional robots' body cognition relies on manual programming and requires low-level algorithms to define the robot's body range. They lack autonomous boundary confirmation capabilities, and reprogramming and calibration are necessary for changes in body posture or expansion of functional areas, resulting in poor adaptability. Simultaneously, current technologies do not combine autonomous boundary recognition with multimodal global perception, failing to achieve triple verification of action commands, execution feedback, and body perception. This leads to ambiguous boundary judgments between the robot and its environment, hindering stable body cognition and restricting the development of autonomous perception and adaptive control. Consequently, they cannot meet the body perception requirements of high-level bionic and intelligent robot systems. Summary of the Invention

[0003] This invention provides a robot body boundary recognition method, system, and robot, which solves the problems of existing technologies such as robot reliance on manual programming, disconnect between perception and execution, and lack of autonomous boundary confirmation logic.

[0004] This invention is achieved through the following technical solution:

[0005] In a first aspect, this application provides a robot body boundary recognition method, including the following:

[0006] Construct a body space mapping model for the robot, which at least defines the spatial range to which each component of the robot belongs;

[0007] Drive the robot's movable parts to perform self-probing actions;

[0008] During the self-detection process, motion signals of movable parts and response signals of the touched object are collected;

[0009] The system determines whether the touched object belongs to the robot itself or the external environment based on whether the motion signal and response signal simultaneously match the preset features within the assigned spatial range.

[0010] In some optional embodiments, simultaneously matching preset features within the assigned spatial interval includes the following:

[0011] The real-time position coordinates of the movable parts fall within the assigned spatial range;

[0012] Furthermore, the time of generation of the response signal coincides with the time of contact between the movable part and the object being touched within a preset time window.

[0013] In some optional embodiments, the ontology spatial mapping model is a three-dimensional ontology coordinate system, and the belonging spatial interval is the coordinate interval of the corresponding body part or functional area in the three-dimensional ontology coordinate system.

[0014] In some optional embodiments, the robot may also perform self-detection actions and judgment steps sequentially on all preset body parts or functional areas to complete the global body boundary confirmation and form or update the body boundary model; and dynamically adjust the belonging space interval or perception response threshold through parameter optimization algorithms.

[0015] In some optional embodiments, the parameter optimization algorithm is one of the following: a PID parameter self-updating algorithm, a Kalman filtering algorithm, or a gradient-based optimization algorithm.

[0016] In some optional embodiments, the motion signal includes spatial location coordinates, motion trajectory, and touch pressure signal, and the response signal includes touch response, target part coordinates, and sensing threshold trigger signal.

[0017] Secondly, this application provides a robot body boundary recognition system for implementing any of the robot body boundary recognition methods described in the first aspect, characterized in that it includes:

[0018] A storage unit is used to store a robot body spatial mapping model, which includes the spatial range of each robot component and its corresponding perception response features;

[0019] An execution unit is used to drive the robot's movable parts to perform self-probing actions;

[0020] A multimodal integrated sensing unit is used to collect motion signals of movable parts and response signals of the touched object during the self-probing process;

[0021] The processing unit is used to determine whether the touched object belongs to the robot body or the external environment based on whether the motion signal and the response signal simultaneously match the preset features within the assigned spatial interval.

[0022] In some optional embodiments, the multimodal sensing unit is configured to acquire one or more of visual perception signals, tactile perception signals, pressure perception signals, temperature perception signals, pose perception signals, acceleration perception signals, and spatial positioning perception signals.

[0023] In some optional embodiments, the multimodal integrated sensing unit and the processing unit communicate in real time through a fully bidirectional signal interaction link to form a closed-loop control loop of command, execution, sensing, and verification.

[0024] Thirdly, this application provides a robot, comprising:

[0025] One or more processors;

[0026] Memory;

[0027] And one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors to implement any of the robot body boundary recognition methods described in the first aspect.

[0028] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0029] 1. This invention constructs a ontological spatial mapping model and drives the robot to actively perform self-probing actions. It collects its own motion signals and the response signals of the touched object for dual matching and judgment, enabling the robot to possess autonomous cognitive abilities similar to the proprioception of a living organism. Compared to existing technologies that rely on manual pre-setting and passive perception, this invention eliminates the need for pre-programming and calibrating the ontological range, thus improving the robot's level of autonomy.

[0030] 2. This invention achieves dual locking of the body boundary from both spatial and temporal dimensions by simultaneously verifying two conditions: the real-time position coordinates falling within the assigned spatial interval and the response signal and contact action coinciding within a time window. This spatiotemporal dual verification mechanism effectively eliminates interference signals from the external environment, improves the accuracy and robustness of body boundary recognition, and solves the problem of fuzzy boundary judgment in traditional single-dimensional perception.

[0031] 3. This invention sequentially probes and determines all body parts and functional areas of the robot to form a complete body boundary model. Combined with parameter optimization algorithms, it dynamically adjusts the assigned spatial interval or perception response threshold, enabling the robot to continuously optimize its cognitive model. This allows it to adapt to changes in body posture, expansion of functional modules, or replacement of the carrier without manual recalibration. Compared to traditional technologies that require reprogramming for every change, this invention's adaptive capability enhances the system's scalability and long-term operational stability.

[0032] 4. This invention constructs a closed-loop control loop of command-execution-perception-verification through a fully bidirectional signal interaction link between the multimodal integrated sensing unit and the processing unit. This enables the robot to perceive its own state changes in real time while performing actions, and to send the perception results back for subsequent command adjustments, forming a self-verification and self-correction cognitive closed loop, providing fundamental technical support for the subsequent realization of embodied intelligence. Attached Figure Description

[0033] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0034] Figure 1 This is a schematic diagram of the robot body boundary recognition method provided in an embodiment of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0036] Firstly, such as Figure 1 As shown, this embodiment provides a robot body boundary recognition method. This method autonomously distinguishes its own components from external objects by having the robot actively perform touch actions and analyze the action feedback. It includes the following:

[0037] S1. Construct a ontology space mapping model that describes the robot's own structure. The ontology space mapping model records at least the spatial region where each part of the robot is currently located, i.e., the spatial interval to which it belongs.

[0038] S2. The robot drives its movable parts, such as robotic arms, fingers, and head, to actively touch the target object.

[0039] S3. During the touch process, the robot records two types of information: first, the motion signals such as the motion trajectory and position of the movable parts themselves; and second, the response signals generated by the touched object, such as pressure, vibration, and temperature changes.

[0040] S4. The robot compares the collected motion signals and response signals with the preset home space interval of the component: if the position of the movable component is within the home space interval and the response signal matches the touch action in time, it is determined that the touched object belongs to the robot itself; otherwise, it is determined to be the external environment.

[0041] In practical applications, the body space mapping model can be represented as a three-dimensional Cartesian coordinate system with the robot's own base as the origin. According to the mechanical design drawings, the spatial range of each component of the robot, such as the head, torso, left arm, right arm, left hand, and right hand, is pre-recorded in this three-dimensional Cartesian coordinate system in the form of coordinate intervals.

[0042] In the above embodiments, the motion signal includes at least the spatial position coordinates and movement trajectory of the movable part. The response signal includes at least the pressure value generated by touch or the tactile sensor signal.

[0043] For example, when the robot performs a self-detection action of touching its head with its left hand: the robot controls its left hand to move from its initial position towards the head's coordinate range. The controller receives position information from the left hand's joint encoder in real time. When the left hand enters the head's coordinate range, the system records its position coordinates. Simultaneously, the tactile sensors of the left hand and the pressure sensors of the robot's head monitor the contact signal in real time. When the left hand touches the head, the pressure sensor reading changes, and the system records the moment of this change. Subsequently, the central processing unit determines: whether the real-time coordinates of the left hand fall within the head's coordinate range; and whether the difference between the moment the left hand touches the head and the moment of the change in the head pressure sensor signal is within a preset time window. If both conditions are met, the head is determined to be part of the robot's body.

[0044] Unlike traditional methods that require manual teaching or preset body coordinates, this embodiment achieves automated and autonomous body boundary recognition by combining self-detection actions with dual matching judgments. This reduces the difficulty of robot deployment and maintenance, and eliminates the need for professional programming. It also improves the system's adaptability to changes in its own structure. For example, when the robot replaces its robotic arm with one of different lengths, the cognitive model can be automatically updated by re-executing this method.

[0045] In some optional embodiments, the specific implementation of simultaneously matching preset features within the assigned spatial interval is further described.

[0046] The matching of preset features within the assigned spatial interval includes two sub-conditions that must be met simultaneously: First, the spatial matching condition, meaning the real-time position coordinates of the robot's movable part when performing a touch action must fall within the preset assigned spatial interval corresponding to that part; second, the temporal matching condition, meaning the time difference between the abrupt change in the response signal generated by the touched object and the time of physical contact between the movable part and the touched object must be less than or equal to a preset allowable range, i.e., a time window. Only when both conditions are met simultaneously will the system determine the touched object as part of the body region.

[0047] In this embodiment, the assigned spatial interval can be a continuous closed region in three-dimensional space, such as a polygonal mesh or an ellipsoid. The determination of whether the real-time position coordinates fall within the assigned spatial interval can be achieved using the Point-in-Polygon (PIP) algorithm, or by calculating whether the Euclidean distance from the real-time coordinate point to the center point of the interval is less than a preset radius.

[0048] In this embodiment, the value of the preset time window is related to the robot's movement speed, the sensor's sampling frequency, and the signal transmission delay, and can be adjusted accordingly based on the actual situation.

[0049] This embodiment improves the accuracy and anti-interference capability of entity boundary recognition by introducing dual spatial and temporal dimensions of verification. Spatial matching ensures the attribution relationship in terms of physical location, while temporal matching ensures the attribution relationship in terms of causality, that is, the signal is triggered by its own action rather than by accidental external factors. Only responses that are correctly positioned and triggered by its own action are recognized as entities, effectively eliminating interference from environmental noise, sensor mis-triggers, accidental contact with external objects, etc., enabling the robot to stably maintain its cognitive boundaries in complex dynamic environments.

[0050] In some optional embodiments, the method further includes sequentially performing self-probing actions and judgment steps on all preset body parts or functional areas of the robot to complete the global body boundary confirmation and form or update the body boundary model; and dynamically adjusting the belonging space interval or perception response threshold through a parameter optimization algorithm. During a single body recognition process, the robot sequentially performs the self-probing actions and judgment steps described in Embodiment 1 on all body parts and functional areas that need to be confirmed according to a preset sequence or dynamically planned path, thereby completing a complete scan of its entire structure and generating or updating a comprehensive body boundary model. After or during the above traversal process, the system uses a parameter optimization algorithm to dynamically adjust the boundary parameters of the belonging space interval or the perception response threshold based on the deviation between the data actually collected during this traversal and the preset values ​​of the model, resulting in higher accuracy and stronger adaptability in the next global traversal.

[0051] For example, during the global ontology boundary confirmation process, the preset sequence or dynamically programmed path could be: right hand touches left shoulder -- left hand touches right shoulder -- right hand touches head -- left hand touches head -- right foot touches left knee -- left foot touches right knee, and so on, covering all joints, limbs, and torso positions. After each touch, the system records the judgment result and measured data. When all preset touch actions have been executed, the system completes the global traversal.

[0052] For example, as a specific implementation of the parameter optimization algorithm, it can adopt one of the following: PID parameter self-updating algorithm, Kalman filter algorithm, or gradient-based optimization algorithm.

[0053] When using the PID parameter self-updating algorithm, the system uses the deviation between the actual coordinates reached in each touch action and the preset coordinate interval boundary as the error signal e(t). The PID controller calculates the adjustment amount and fine-tunes the boundary of the coordinate interval so that subsequent actions can more accurately fall into the center of the interval.

[0054] When using the Kalman filter algorithm, the system uses the preset coordinate range as the state prediction value and the coordinate measured at the actual touch as the observation value. The two are fused by Kalman filtering to obtain the updated optimal estimated coordinate range, while smoothing out the random noise in a single measurement.

[0055] When using gradient-based optimization algorithms, the system constructs a loss function, for example, loss = Σ(actual touch position - preset interval boundary)^2. Then, by calculating the gradient of the loss function with respect to the interval boundary, the boundary value is gradually adjusted along the gradient descent direction to make subsequent touch positions closer to the center of the interval.

[0056] In this embodiment, through global traversal, the robot can fully understand its own structure and form a closed-loop cognitive system. Through parameter optimization algorithms, the robot can learn from each self-probing and continuously correct its own model, so that the body boundary cognition is not a one-time initialization operation, but a continuous optimization process that accompanies the robot throughout its entire life cycle, thereby improving the system's intelligence level and long-term stability.

[0057] In some optional embodiments, the motion signal may include one or more of spatial location coordinates, motion trajectory, and touch pressure signal; the response signal may include one or more of touch response, target part coordinates, and sensing threshold trigger signal. By acquiring sensing signals of multiple modalities, the system can verify the authenticity of touch events from multiple dimensions, further improving the reliability of body boundary recognition.

[0058] Spatial position coordinates can be obtained using a joint encoder.

[0059] An action trajectory can be described by recording a continuous sequence of position coordinates over a period of time.

[0060] For touch pressure signals, they can be provided by pressure sensors or force / torque sensors mounted on the moving parts.

[0061] Touch response specifically refers to any measurable physical change produced at the touched part, including but not limited to increased pressure, temperature change, vibration, sound, etc.

[0062] The coordinates of the target area can be transmitted back by the positioning sensor installed on the touched area itself.

[0063] The sensing threshold trigger signal is a digital pulse signal or interrupt signal generated when the sensor measurement value of the touched part exceeds a preset threshold. This signal has the characteristics of low latency and high reliability and is suitable for time window matching.

[0064] The fusion of multimodal signals in this embodiment enables the system to make judgments even when one type of sensor fails or is interfered with, relying on other modal signals to enhance the system's redundancy and robustness. Furthermore, cross-validation between different signal types can further eliminate false judgments caused by single-point sensor failures.

[0065] Secondly, this embodiment provides a robot body boundary recognition system for implementing any of the above methods. The system includes at least a storage unit, an execution unit, a multimodal integrated sensing unit, and a processing unit. The storage unit stores the aforementioned body space mapping model and related parameters; the execution unit drives the robot's movable parts to perform self-probing actions; the multimodal integrated sensing unit collects motion signals and response signals in real time; and the processing unit executes dual-matching judgment logic. In this system, the multimodal integrated sensing unit and the processing unit communicate in real time via a fully bidirectional signal interaction link, forming a closed-loop control circuit of command-execution-perception-verification.

[0066] In this embodiment, the storage unit can be an embedded Flash memory, an SD card, or an external cloud database. In addition to storing the ontological spatial mapping model, it can also store historical cognitive data, sensor calibration parameters, etc.

[0067] An execution unit typically includes a motion controller, a servo driver, and a motor. The execution unit receives self-detection commands from the processing unit, converts them into drive signals to control the motor's operation, and sends feedback data from the motor encoder back to the processing unit in real time.

[0068] The multimodal integrated sensing unit includes various sensors distributed throughout the robot, such as tactile sensor arrays, pressure sensors, cameras, temperature sensors, and accelerometers, to acquire visual, tactile, pressure, temperature, pose, acceleration, and spatial positioning signals. These sensors can provide either analog or digital outputs.

[0069] The processing unit can be an MCU, DSP, FPGA, or a CPU / GPU that runs the algorithm. The processing unit is responsible for executing the decision algorithm and writing the decision result back to the storage unit, or for updating the model.

[0070] Thirdly, this embodiment provides a robot that includes one or more processors, a memory, and program instructions stored in the memory and executable by the processor. When the processor executes these program instructions, it can implement the robot body boundary recognition method described in any of the above embodiments.

[0071] To better illustrate the present invention, the following is provided:

[0072] Application Example 1

[0073] This application example applies the invention to a bionic robot, enabling it to autonomously recognize the boundaries of its own face (such as the nose and cheeks).

[0074] First, a fully bidirectional interactive architecture was established between the central processing unit, the multimodal integrated sensing unit, and the execution unit. The multimodal integrated sensing unit, which integrates tactile, pressure, and pose sensors, was distributed across the robot's right hand, nose, and left cheek. The central processing unit, based on the robot's facial CAD model, pre-defined a three-dimensional body coordinate system and calibrated the coordinate ranges and corresponding tactile / pressure response thresholds for the "nose" and "left cheek."

[0075] Upon initiation of recognition, the central processing unit issues the instruction to "touch the nose with the right hand." The execution unit (right-hand mechanical structure) begins to move towards the "nose" coordinate region. During this movement, the multimodal sensing unit acquires the spatial position and trajectory of the right hand in real time. When the right hand touches the nose, the tactile / pressure sensors on the nose immediately generate feedback signals. All of the aforementioned signals are transmitted back in real time.

[0076] The central processing unit performs dual checks: 1) whether the real-time coordinates of the right hand fall within the coordinate range of the "nose"; 2) whether the time point when the right hand touches the nose highly coincides with the time point of the sudden change in the pressure signal of the nose. If both checks pass, the "nose" is determined to be the main body region. Subsequently, the system traverses all facial parts such as the "left cheek" in the same way to form a facial body boundary model and initiates parameter self-optimization to improve the subsequent recognition accuracy.

[0077] Application Example 2

[0078] This application example applies the invention to industrial robots to achieve high-precision and safe human-machine interaction.

[0079] Based on the three-module architecture, multimodal sensing units are deployed on the main and auxiliary robotic arms and grippers. The central processing unit presets and calibrates the coordinate system. When the self-check command of "main arm touches gripper" is executed, the system confirms that the "gripper" is part of the main body by verifying the linkage between coordinate matching and pressure signal.

[0080] When the robot is in operation, if an external person or object enters the robotic arm's range of motion, the central processing unit issues a "touch" command to detect it. Since the external object lacks a pre-defined coordinate range, and the touch pressure from the actuator does not trigger any pre-defined "body feedback signal," both coordinate consistency and signal linkage checks will fail. The system then determines the object as an "external area" and immediately triggers a safety stop or avoidance protocol, achieving autonomous and safe interaction without the need for external sensors (such as light gratings).

[0081] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for recognizing the boundary of a robot body, characterized in that, Includes the following: Construct a body space mapping model for the robot, which at least defines the spatial range to which each component of the robot belongs; Drive the robot's movable parts to perform self-probing actions; During the self-detection process, motion signals of movable parts and response signals of the touched object are collected; The system determines whether the touched object belongs to the robot itself or the external environment based on whether the motion signal and response signal simultaneously match the preset features within the assigned spatial range.

2. The robot body boundary recognition method according to claim 1, characterized in that, Simultaneously matching the preset features within the assigned spatial interval includes the following: The real-time position coordinates of the movable parts fall within the assigned spatial range; Furthermore, the time of generation of the response signal coincides with the time of contact between the movable part and the object being touched within a preset time window.

3. The robot body boundary recognition method according to claim 1, characterized in that, The ontological spatial mapping model is a three-dimensional ontological coordinate system, and the spatial interval to which it belongs is the coordinate interval of the corresponding body part or functional area in the three-dimensional ontological coordinate system.

4. The robot body boundary recognition method according to claim 1, characterized in that, It also includes performing self-detection actions and judgment steps sequentially on all preset body parts or functional areas of the robot to complete the confirmation of the whole body boundary and form or update the body boundary model; and dynamically adjusting the belonging space interval or perception response threshold through parameter optimization algorithm.

5. The robot body boundary recognition method according to claim 4, characterized in that, The parameter optimization algorithm is one of the following: PID parameter self-updating algorithm, Kalman filtering algorithm, or gradient-based optimization algorithm.

6. The robot body boundary recognition method according to claim 1, characterized in that, The motion signal includes spatial position coordinates, motion trajectory, and touch pressure signal, and the response signal includes touch response, target part coordinates, and sensing threshold trigger signal.

7. A robot body boundary recognition system, used to implement the robot body boundary recognition method according to any one of claims 1 to 6, characterized in that, include: A storage unit is used to store a robot body spatial mapping model, which includes the spatial range of each robot component and its corresponding perception response features; An execution unit is used to drive the robot's movable parts to perform self-probing actions; A multimodal integrated sensing unit is used to collect motion signals of movable parts and response signals of the touched object during the self-probing process; The processing unit is used to determine whether the touched object belongs to the robot body or the external environment based on whether the motion signal and the response signal simultaneously match the preset features within the assigned spatial interval.

8. The robot body boundary recognition system according to claim 7, characterized in that, The multimodal sensing unit is configured to acquire one or more of the following: visual perception signals, tactile perception signals, pressure perception signals, temperature perception signals, pose perception signals, acceleration perception signals, and spatial positioning perception signals.

9. The robot body boundary recognition system according to claim 7, characterized in that, The multimodal integrated sensing unit and the processing unit communicate in real time through a fully bidirectional signal interaction link, forming a closed-loop control loop of command, execution, sensing, and verification.

10. A robot, characterized in that, include: One or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors to implement the robot body boundary recognition method according to any one of claims 1 to 6.

Images