A wireless modular embodied intelligent robot control method and system

Abstract

The application discloses a wireless modular embodied intelligent robot control method and system, and belongs to the technical field of wireless module robot control. Specifically disclosed is a wireless modular control architecture composed of multiple modules; after each module is powered on, structured capability description information is broadcast through a local wireless network, and based on physical connection perception, adjacent records are bidirectionally exchanged, the whole machine topology is autonomously discovered by a coordination module, and a kinematics and dynamics model is generated; a three-layer control architecture is adopted, a standard time frame is broadcast by the coordination module, and trajectory parameters or action primitives with future time stamps are issued, and each module executes cross-module cooperative action based on local clock alignment. The wireless and modular recombination of the robot structure is realized, the plug-and-play and automatic reconstruction of the system model are realized when modules are added, deleted or replaced, the anti-interference ability and reliability of the system are improved, and the adaptability and safety of the robot under complex working conditions are enhanced.

Description

Technical Field

[0001] This invention belongs to the field of wireless modular robot control technology, specifically relating to a wireless modular embodied intelligent robot control method and system. Background Technology

[0002] Currently, robots need to handle a wide variety of complex tasks. Modular design offers significant advantages over non-modular robots, such as easier independent manufacturing, testing, maintenance, and repair. It also allows for flexible adjustments to robot functions and configurations to meet specific needs, making them more adaptable and scalable. However, existing modular robot solutions still have many limitations in terms of module autonomy, interconnection methods, and overall robot modeling.

[0003] For example, the invention with publication number CN119283055B discloses the use of modular robot joints and standardized mechanical interfaces to achieve modular reconfiguration of the robot's structure; and modular reconfiguration of circuit topology and communication architecture is achieved through joint communication circuits with universal interfaces. However, it still falls under the category of centralized control and wired bus interconnection, and its modular intelligence, interconnection method, and overall machine modeling capabilities have certain limitations.

[0004] Therefore, a wireless modular embodied intelligent robot control method and system are proposed to solve the technical problems of existing technologies that neglect the embodied intelligence level of the modules themselves, wireless internal interconnection, and the ability to automatically model the whole machine based on physical connection perception. Summary of the Invention

[0005] This invention addresses the problems of low module intelligence, reliance on wired interconnection leading to reliability and form limitations, and lack of automatic whole-machine modeling capabilities based on physical connections in existing modular robot systems. It proposes a wireless modular embodied intelligent robot control method and system.

[0006] A further objective of this invention is to establish a topology autonomous discovery mechanism based on module self-description and physical connection awareness, enabling each module to automatically broadcast its structured capability description information to the network after power-on, and to identify the actual physical connection relationship with other modules through a connection detection unit, so that the system can automatically identify the complete topology of the current system without human intervention.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a wireless modular embodied intelligent robot control method, comprising: S1. Define each robot functional unit that can independently complete a certain function and participate in collaboration as an autonomous module; combine multiple autonomous modules to form a robot; S2. The autonomous modules power on and perform self-tests in sequence, establish a local wireless network, load the locally stored module description information, and join the wireless network. S3. Each module broadcasts its capability description information to the coordination module; S4. The coordination module performs self-discovery of the whole machine topology and builds a body model, and performs overall calibration and synchronization. S5. The whole machine accepts task commands or external control.

[0008] Preferably, each module performs a self-test upon power-up, checking the status of local hardware, sensors, actuators, power control units, and stored data, and loading pre-stored module description information. It joins the robot's internal local area network via a preset wireless protocol and completes authentication using keys, certificates, tokens, or whitelist mechanisms to prevent unauthorized nodes from joining the system. Preferably, each module broadcasts its capability description information to the coordinating module, including at least a module identity field, interface field, pose description field, and the initial value of the current connection status.

[0009] As a preferred approach, the two modules that have completed identity authentication and detected the connection exchange each other's identity information, interface information, and interface pose information to establish a bidirectional, consistent adjacency record.

[0010] Each adjacency record includes at least: local module ID, local interface number, peer module ID, peer interface number, connection direction, locking status, and relative installation constraints between interfaces.

[0011] Preferably, the coordination module receives connection records reported by all modules and constructs an initial connection graph with modules as nodes and connection relationships as edges. The coordination module determines and identifies the root node of the whole machine. Starting from the root node, the coordination module traverses the initial connection graph to generate parent-child hierarchical relationships between modules, forming a topology tree or a constrained topology graph, and determines the parent and child modules of each module in the topology structure.

[0012] Preferably, for each identified set of connection modules, the coordination module determines the pose relationship between the sub-module coordinate system and the parent module coordinate system based on the actual physical interface information, interface coordinate system definition, interface orientation, mechanical installation constraints, and connection direction, and updates the transformation relationship between the sub-module's own coordinate system and the parent coordinate system. The system automatically generates a corresponding homogeneous transformation matrix based on this pose relationship to describe the rigid body transformation between the sub-module coordinate system and the parent module coordinate system.

[0013] For joint modules, the joint motion transformation relationship can be further established based on the homogeneous transformation matrix by combining the joint type, joint axis direction, zero offset, and range of motion.

[0014] The coordination module assigns logical roles to each module based on module type, spatial location, connection level, kinematic parameters, and template matching results. Based on topological relationships, parent-child coordinate transformation relationships, and the kinematic fields of each module, the coordination module automatically identifies the series chains, parallel branches, end effector chains, and the distribution of degrees of freedom of the entire machine, thereby generating the overall kinematic model, dynamic model, and power supply path diagram.

[0015] After completing kinematic recognition, the system can automatically generate a structured description file corresponding to the current robot form. This description file characterizes the module hierarchy, connection relationships, joint attributes, link parameters, coordinate transformation relationships, and inertial parameters. The description file can adopt a unified robot description file format, preferably a machine-readable format that can characterize the robot's topology, joint relationships, geometric parameters, inertial parameters, and coordinate transformation relationships, such as URDF-type format or other equivalent structured description formats. The coordination module distributes the generated topology results, role information, parent-child coordinate transformation relationships, control parameters, clock synchronization parameters, and structured description results to each module for local control, cooperative control, or subsequent reconfiguration.

[0016] Preferably, during the whole-machine topology self-discovery, if two modules declare inconsistent connections for the same relationship, the connection is marked as a connection to be confirmed; if the connection direction declared by a module conflicts with the pose constraint, it is marked as an abnormal connection; if the module type does not match the current topology position, it is marked as a role conflict module; if the constructed topology graph contains closed loops that do not conform to the preset structure, closed loop pruning, connection disabling, or special mode processing is performed; if key nodes are missing, the system enters a degraded control mode or a safe shutdown mode; if topology changes cause the original control model to fail, the system suspends high-dynamic actions and retains only the safe action mode.

[0017] As a preferred method, the following clock synchronization mechanism is used during overall calibration: the coordination module periodically broadcasts a standard time frame; each module receives the time frame, measures its local reception time, and corrects its local time reference; the coordination module issues action parameters with future trigger timestamps; each module begins to execute the relevant action when the corresponding timestamp arrives; if the clock deviation is detected to exceed the threshold, the system restricts high-dynamic collaborative actions or triggers a resynchronization process.

[0018] Preferably, the coordination module does not directly send the control value at each servo sampling moment, but instead sends the target trajectory parameters, action segment parameters, or abstract action primitives within a future time window; each module performs interpolation, expansion, and closed-loop execution on the received target trajectory based on the local controller, local model, and current state; high-frequency control variables are actively generated locally by the module; cross-module collaborative actions adopt a future timestamp aligned execution mechanism; in the event of short-term packet loss, the module uses the most recent valid target and maintains local security control; when the wireless interruption time exceeds a set threshold, the module switches to a preset security mode.

[0019] As a preferred approach, communication messages are categorized into different priorities, such as emergency stop, control, status, synchronization, and maintenance. Emergency stop and fault messages receive the highest priority and can preempt the current transmission queue. Control messages are scheduled periodically. Status messages employ a combination of periodic reporting and event triggering. Maintenance messages are transmitted in idle network windows or low-priority channels. Key modules are allocated guaranteed bandwidth, fixed time slots, or priority transmission rights. Communication messages are categorized as follows: discovery messages, control messages, status messages, synchronization messages, security messages, and maintenance messages.

[0020] As a preferred approach, discovery messages include module online broadcasts, capability description broadcasts, interface connection status declarations, module offline notifications, and hot-plug event notifications; control messages include control mode switching commands, target position / speed / torque / impedance ranges, target action segment parameters, action primitive call commands, role configuration commands, and parameter update commands; status messages include joint status, voltage, current, and temperature status, sensor summaries, health status, self-test results, and execution progress; synchronization messages include clock synchronization frames, cycle start markers, future trigger timestamps, and synchronization action trigger messages; safety messages include emergency stop broadcasts, disconnection alarms, power supply anomaly alarms, over-temperature alarms, mechanical tripping alarms, degraded operation commands, and safety lock commands; and maintenance messages include log summaries, fault codes, firmware upgrade fragments, maintenance authorization information, and calibration information synchronization.

[0021] As a preferred feature, the coordination function possesses fault-tolerant processing capabilities, including: short-term packet loss handling, continuous disconnection handling, critical module disconnection handling, link quality degradation handling, and security broadcast redundancy. Specifically, when no new control messages are received for a short period, the module continues to execute the most recent effective control objective and maintains stability through local limiting, soft start, or damping control; when the disconnection duration exceeds the first threshold, the module enters a restricted operation mode; when it exceeds the second threshold, the module enters a safety hold, drive disconnection, return to center, brake, or low-power waiting mode; when the coordination module disconnects, the coordination function can be taken over by a preset backup module, or each module can switch to a degraded autonomous mode; when the latency, packet loss rate, jitter, or bit error rate exceeds the set threshold, the system automatically reduces the severity of actions, reduces the frequency of status reporting, suspends non-critical communication, or switches to a low-bandwidth secure communication mode; for high-security messages such as emergency stops, redundancy methods such as repeated sending, asynchronous repeated broadcasting, sending through different channels, or multi-node forwarding can be used.

[0022] To realize the aforementioned wireless modular embodied intelligent robot control method, a wireless modular embodied intelligent robot control system is also proposed, comprising: a power module, which may include multiple modules, providing energy and participating in intelligent control; joint modules, constituting the robot's various degrees of freedom, such as hip, knee, ankle, shoulder, elbow, and wrist; a dexterous hand module, enabling multi-finger, compliant grasping and fine manipulation; a perception module, including a visual head module, an auditory module, and a tactile module; a computing and coordination module, responsible for global task planning and module coordination control; and a communication module, constructing an internal wireless network and interacting with external networks.

[0023] Preferably, the modules are interconnected in the following ways: wireless communication connection for data exchange of control commands, status information, and model parameters; contact connection for power supply and necessary high-speed or safe signal transmission, with ring electrodes, contact springs, and slip rings at joint connections, and mechanical quick-locking latches and electrical contact pieces at the connections between the torso and limbs; the communication module includes a module wireless communication unit, a network coordination unit, a clock synchronization unit, a link monitoring unit, a security authentication unit, and a cooperative interface unit.

[0024] Preferably, the control system adopts a three-layer control architecture: the first layer is a high-frequency closed-loop control layer within each module, where the internal controller of each module independently executes high-frequency servo control of position, speed, current, torque, impedance, or damping, with a control frequency of 1kHz to 5kHz; the second layer is a wireless coordination control layer between modules, where the coordination module exchanges target actions, status summaries, and synchronization commands with each module via wireless communication, with a preferred update frequency of 50Hz to 200Hz; the third layer is a task planning layer, where the computing and coordination module, the local high-performance computing platform, or external computing resources perform task understanding, action planning, and strategy switching, and issue task-level targets to the coordination control layer at a lower frequency, with a preferred frequency of 10Hz to 50Hz.

[0025] The beneficial effects of this invention are as follows: by completing the high-frequency servo closed loop locally in the module, the wireless network only undertakes the mid- and low-frequency coordination commands, replacing the long-distance flexible wire harness that runs through the whole machine, simplifying the structure and reducing costs; by automatically broadcasting structured capability description information and autonomously detecting physical connection relationships after the module is powered on, the system can achieve plug-and-play when adding, deleting or replacing modules; through a hierarchical control architecture and priority scheduling based on message type, high-frequency servo, mid-frequency coordination and low-frequency planning are run in a time-sharing and hierarchical manner, improving the coordination efficiency in complex communication environments. Attached Figure Description

[0026] Figure 1 This is a block diagram of the internal structure of the self-developed module of this invention.

[0027] Figure 2 This is a flowchart of the robot startup and module self-discovery process of the present invention.

[0028] Figure 3 This is a schematic diagram of the module connection topology according to an embodiment of the present invention.

[0029] Figure 4 This is a schematic diagram of the module body model according to an embodiment of the present invention.

[0030] Figure 5 This is a timing diagram for the system control and communication of the present invention.

[0031] Figure 6 This is a schematic diagram of the combined structure of the motor module and the power supply module in an embodiment of the present invention.

[0032] Figure 7 This is a schematic diagram of the overall structure of the humanoid robot system according to an embodiment of the present invention.

[0033] In the accompanying drawings of this invention, the following annotations are used uniformly: 100—Overall embodied intelligent humanoid robot; 110—Head sensing module; 120—Tortoise module; 130—Left arm module; 140—Right arm module; 150—Left leg module; 160—Right leg module; 171—Left dexterous hand module; 172—Right dexterous hand module; 181—First power supply module; 182—Second power supply module; 190—Coordination module; 191—Communication module; 200—Autonomous module (general); 201~209—Examples of joint / motor modules; 300—Mechanical interface unit; 310—Electrical / energy interface unit (contact interface); 320—Wireless communication unit; 330—Computation and AI / AGI unit; 340—Action and sensing unit (motor / sensor); 350—Safety and fault management unit. 370—Power management submodule, 380—Battery unit, 400—Inter-module wireless communication network, 501—Schematic curve of servo control cycle within module, 502—Schematic curve of coordination control cycle between modules, 503—Schematic curve of task / planning control cycle, 504—Module online / offline event signal, 261—Module mechanical locking structure (clamp / lock), 262—Contact contact area for module docking. Detailed Implementation

[0034] Example. This example describes the specific implementation of a wireless modular embodied intelligent robot control method and system on a humanoid robot.

[0035] refer to Figure 1 , Figure 6 and Figure 7 The embodied intelligent humanoid robot in this embodiment is composed of multiple functionally independent autonomous modules. These modules are physically connected through mechanical interface units and electrical / energy interface units to form a complete body. Specifically, it includes: Head perception module: integrates visual and auditory sensors and a local AI processing unit.

[0036] Trunk Module: As the core of the structure, it integrates a computing and coordination module (brain module) and a communication / gateway module.

[0037] Left and right arm modules: including shoulder, elbow, and wrist joint modules.

[0038] Left and right leg modules: including joint modules such as hip, knee, and ankle, such as right hip joint and left hip joint.

[0039] Left and right dexterous hand modules: The dexterous hand has a tactile / force sensor array on its palm, and the multiple finger modules are composed of multiple finger joint modules. They complete grasping and manipulation behaviors through local control and AI unit.

[0040] Power modules: The first power module (main power) and the second power module (auxiliary power) contain battery units and a power management module to provide energy to the system.

[0041] Joint Modules: Each joint is a complete autonomous module.

[0042] Take the motor-power supply combination module as an example. One end of the module is a mechanical module with a mechanical locking structure, and the other end is a matching locking groove. After docking, a stable physical connection is formed. At the same time, the contact areas of the module docking automatically fit together to establish an electrical / energy interface.

[0043] refer to Figure 2 The flowchart, system startup, and workflow are as follows.

[0044] First, the user triggers startup. The first power module (main power) and the second power module (auxiliary power) are powered on. Their internal communication / gateway modules act as network coordination nodes, immediately establishing a high-bandwidth, low-latency internal wireless network. Subsequently, the various autonomous modules distributed throughout the robot (such as joint modules and dexterous hand modules) obtain power and start up through contact interfaces. Each module's computing and storage unit executes a self-test program, checking the hardware status of motors, sensors, storage units, etc., and loading the locally pre-stored module description file from the storage unit. After the self-test passes, the wireless communication unit within the module searches for the wireless network and completes authentication through a pre-set whitelist and certificate mechanism, securely joining the network. If the self-test fails, the module reports a fault and enters safe mode.

[0045] All modules that successfully join the network immediately broadcast their structured capability descriptions to the entire network. Taking the right hip joint module as an example, its broadcast message includes: Module Identity field: ID="207", Type="Hip Joint_Yaw", Role="Pending". Interface fields: Interface A (Proximal / Input): Physically connected to "Torso Module 120", Interface Type="Ring Electrode Quick-Lock Interface", Interface Pose={Position(0.15, -0.10, 0.00), Orientation(0, 0, 0)}; Interface B (Remote / Output): Physically connected to "Right Leg Module 208", Interface Type="Ring Electrode Quick-Lock Interface", Interface Pose={Position(0.00, 0.00, -0.15), Orientation(0, -90°, 0)}. Kinematic fields: Joint Axis="Z-axis", Range of Motion="-30° to +90°". Current connection status initial values: Interface A = "Connected", Interface B = "Connected".

[0046] Meanwhile, the module detects physical connection events in real time through changes in the pin levels of the interface or a digital handshake protocol. Once a connection is detected, two adjacent modules (such as the torso module and the right hip joint) exchange their identities, interface numbers, and pose information to establish a bidirectional, consistent adjacency record, such as: {Local ID: 120, Local Interface: B3, Peer ID: 207, Peer Interface: A, Connection Direction: "120 Output to 207 Input", Locking Status: "Locked"}.

[0047] All modules report their local adjacency records to the computation and coordination module. The coordination module then executes the following algorithm.

[0048] Connection relationship summarization and topology construction. Receive all adjacency records and construct an initial connection graph with modules as nodes and connections as edges, such as... Figure 3 As shown in the “Module Connection Topology”, the trunk module (including the coordination module) is identified as the root node of the whole machine.

[0049] Conflict detection and handling. During the construction process, if the right arm module declares that it is connected to the left arm module, but the left arm has no such record, the connection is marked as "pending confirmation" and a prompt is issued. If a circular connection (such as torso-leg-foot-torso) is found, non-load-bearing redundant connections are automatically trimmed according to structural constraints.

[0050] Topology hierarchy construction. Starting from the root node, traverse the tree to generate a parent-child relationship tree, establishing the parent and child modules for each module.

[0051] Body shape and kinematic model generation. Combining the parent-child coordinate transformation relationship (i.e., the interface pose information of each joint), the homogeneous transformation matrix of each child module in the parent module's coordinate system is automatically calculated, and finally concatenated to generate a model as follows: Figure 4 The complete overall body topology model is shown. Based on this body model and the kinematic fields of each module, the system automatically derives the kinematic / dynamic model of the whole machine, identifies the left and right arms as series chains and the two legs as parallel branches, and automatically generates a structured description file in URDF format.

[0052] Model and parameter distribution. The coordination module distributes the calculated logical roles of each module, parent-child transformation matrices, unified clock synchronization parameters, and URDF files to all modules.

[0053] After the entire system has completed modeling and calibration, it enters the operational state, employing a three-layer control architecture and a precise clock synchronization mechanism. The communication / gateway module periodically (e.g., every 10 milliseconds) broadcasts a standard time frame with a high-precision timestamp to the entire network. Upon receiving this frame, each joint module determines its local reception time, compensates for transmission delay using an algorithm, and aligns its local clock with the global reference.

[0054] When the computing and coordination module issues dynamic action commands, it includes a "future trigger timestamp." For example, it might send a command stating, "At global time T=100.000 seconds, all right arm joints begin executing trajectory A." Upon receiving this command, each module stores it in its local buffer. When the clock reaches the precise moment T=100.000 seconds, the local controller independently and synchronously initiates the action.

[0055] The three-layer control architecture is as follows: Task planning layer (10-50Hz): When the robot receives the instruction to "take a step forward", the calculation and coordination module executes task planning, generates gait parameters and centroid trajectory, and sends them to the coordination control layer at a low frequency.

[0056] Inter-module coordination control layer (50-200Hz): The coordination control layer decomposes the task planning into the target position, velocity envelope curve or motion primitive (such as "impedance control mode") of each joint within a future time window (such as 500 milliseconds), and sends it to each joint module in non-real-time and unequal intervals through a wireless network 400.

[0057] High-frequency closed-loop layer within the module (1-5kHz): Taking the joint module as an example, after receiving instructions from the upper layer, its internal microcontroller (MCU) executes closed-loop servo control of the position at a frequency above 1kHz, based on the dense setpoints obtained by spline interpolation of the target trajectory, and combined with real-time feedback from the local motor encoder. This architecture ensures that joint control remains completely unaffected even when there are brief fluctuations in the wireless network.

[0058] If the right knee module does not receive new coordination commands for 50 consecutive milliseconds during movement, it is considered a short-term packet loss. Its local controller will use the most recent target speed and maintain safe operation with a preset damping control strategy. If the loss of connection exceeds 500 milliseconds, the module will automatically switch to the preset "support hold" mode, lock the current angle, activate the brake, and other modules will also enter a deceleration safety mode and report a "right knee module loss of connection alarm" to the main controller.

[0059] When the robot's left arm module physically disengages from the body due to an external impact, a mechanical disengagement alarm event is triggered by a high-frequency interrupt signal. The module immediately performs a safety lock autonomously and broadcasts a high-priority emergency stop message to all modules in the body within 1 millisecond. All joints enter torque release or brake state within a few milliseconds to ensure the safety of the robot and the environment.

[0060] In another preferred embodiment, after the robot is assembled with new modules and powered on, the system enters the overall topology self-discovery phase. During this process, the computing and coordination module encounters and intelligently handles various anomalies when summarizing adjacency records.

[0061] Handling Inconsistencies and Abnormal Connections: If the left arm module claims its first interface is connected to the left dexterous hand module, but the left dexterous hand module claims it is connected to the right arm module, the system will immediately mark this connection as pending confirmation, flash the LED indicators on both modules, and simultaneously prompt the user via voice to check the physical connection. Similarly, if the left leg module claims its connection direction conflicts with built-in pose constraints (e.g., the interface should face downwards, but the actual data shows upwards), the system will mark it as an abnormal connection and refuse to perform kinematic modeling on this connection link to prevent potential hazards.

[0062] Role Conflicts and Topology Adjustment. Suppose a user mistakenly installs a wrist joint module that only supports rotation in the knee joint position, which requires high torque output. When building the topology tree, the system detects that the module type does not match the current topology location and immediately marks it as a role conflict module, sending a detailed fault report to the main control terminal. If a physical closed loop that does not conform to the preset tree or chain structure is unexpectedly formed in the topology diagram (e.g., due to the installation of an additional rigid reinforcement rod), the system will perform closed-loop pruning, disabling one of the connections at the software level, logically disconnecting it to ensure the correct definition of the kinematic model, and prompting the user.

[0063] Critical Missing Components and Degradation. When the robot is preparing to perform combat training, the system detects a critical missing component—the right ankle joint module responsible for stability. To prevent tipping, the system forcibly enters a degraded control mode, limiting the robot's maximum movement speed to 0.1 m / s and locking all jumping and single-leg standing motion primitives. Similarly, if the left arm accidentally detaches during task execution, causing a drastic change in the topology, the previously generated overall motion control model instantly fails. The system will pause all high-dynamic movements within milliseconds, retaining only the "slowly crouch and hold" safety motion mode to prevent the robot from losing control and being damaged.

[0064] In a preferred embodiment, the system employs a high-precision coordinated clock synchronization mechanism.

[0065] To achieve precise coordination among dozens of joints throughout the body, the system employs a sophisticated clock synchronization mechanism. The computing and coordination module utilizes the communication / gateway module to periodically broadcast a "standard time frame" with a precise timestamp to the wireless network every 10 milliseconds (100Hz). Upon receiving this frame at the physical layer, the wireless communication unit of each joint module (taking the right shoulder joint as an example) records the local reception time. Then, combining this with known instruction processing and transmission delays, the local clock reference is fine-tuned in software to ensure that the "time sense" of all modules remains strictly consistent.

[0066] When the robot needs to perform a rapid dance rotation, the coordination module does not issue immediate position commands. Instead, it generates a sequence of motion parameters containing future timestamps, such as, "{Action segment: Raise right arm, start timestamp: 10:00:00.500, end timestamp: 10:00:01.200, initial posture: (0,0,0), target posture: (90°, 45°, 0)}". The shoulder, elbow, and wrist joint modules related to the right arm will be independently triggered by the local controller when the exact future time of 10:00:00.500 arrives. During execution, if the system detects that the clock deviation of a certain joint module suddenly exceeds 50 microseconds, to avoid the movement of that joint generating antagonistic torques with other joints, the system will immediately restrict all highly dynamic coordinated movements, retaining only quasi-static balance adjustments, and triggering a resynchronization process for that node.

[0067] For example, for a task of "grabbing a water cup," the task planning layer will send an abstract action primitive to the coordination module within an execution window of approximately 0.5 seconds, such as "Execute <resistance grasping mode> on the right dexterity hand module, with a target force of 2 Newtons." The coordination module will then distribute this primitive and the corresponding force control parameters via a wireless network.

[0068] The local control and AI unit within the right dexterous hand module operates entirely on its own after receiving commands. Based on built-in models and feedback from current tactile sensors, it independently performs high-frequency position and force hybrid closed-loop control on each finger joint, using spline interpolation to generate thousands of control setpoints per second from the received target trajectory. Cross-module collaboration, such as arm movement and finger grasping, is accomplished through a future timestamp-aligned execution mechanism: the arm stops and locks at "T=20.000", and the fingers begin pinching at "T=20.050".

[0069] In this autonomous operation mode, occasional packet loss of tens of milliseconds in the communication network will not affect the grasping process. The knuckle module will use the most recent valid target command and maintain local damping control to ensure safety. Only when the wireless interruption time exceeds the preset 200 millisecond threshold will the module determine it as an abnormal interruption and actively switch to the preset safety mode (such as slowly opening the finger to release the object).

[0070] In a preferred embodiment, a priority-based communication and messaging system is employed.

[0071] In order to ensure the real-time performance of critical commands within the limited wireless bandwidth, the system strictly divides communication messages into different priorities and defines six specific message types.

[0072] Safety and emergency stop categories: These have the highest priority and support preemptive sending queues. When the tactile sensor on the dexterous hand detects abnormally high temperatures, it immediately generates an "over-temperature alarm" and broadcasts it. This message is given the highest priority by the network and is instantly delivered to the coordination module, triggering cooling protection.

[0073] Control and synchronization types: employing periodic scheduling or high real-time channels. The coordination module precisely sends commands such as "future trigger timestamp" (synchronization type) and "target impedance range" (control type) in fixed time slots within each 200Hz control cycle.

[0074] Status-based reporting: A combination of periodic reporting and event triggering is used. In a healthy state, the joint module only reports a summary status at 10Hz; once its temperature exceeds the threshold, a high-priority detailed health status report is immediately triggered.

[0075] Maintenance and discovery modules are transmitted in idle network windows or low-priority channels. When the robot is in standby mode, each module uses background bandwidth to upload fault codes and log summaries (maintenance). When a user hot-plugs a new dexterous hand module into the robot, the module sends a "module online broadcast" (discovery), and the system immediately detects this and begins the self-discovery process.

[0076] For critical modules such as the torso module, the network allocates fixed time slots and guaranteed bandwidth to ensure that control commands can be delivered on time without interruption, even when the network is congested.

[0077] The system's coordination function provides multi-layered fault tolerance. In the event of short-term packet loss or continuous disconnection, as mentioned above, packet loss within 50 milliseconds is maintained locally by the module. If the disconnection exceeds the first threshold of 100 milliseconds, the module enters a restricted operation mode, actively reducing the movement speed and torque limit; if it exceeds the second threshold of 1 second, it enters a safety holding state, with the motor drive disconnected or the brake engaged.

[0078] If the coordination module loses connection, and the master computing and coordination module fails unexpectedly, the preset backup module (such as the network coordination node in the communication module) will detect and take over the coordination function within 10 milliseconds. If the backup also fails, each module will immediately switch to a pre-saved degraded autonomous mode. For example, the two-leg module switches to static standing damping control, and the two-arm module stays in its current position, waiting for instructions to restore.

[0079] When the link monitoring unit detects that the network packet loss rate and latency exceed the set threshold, the system will adaptively perform degraded actions, such as switching from running to walking, automatically reducing the reporting frequency of status messages from 50Hz to 5Hz, and suspending the uploading of non-critical logs.

[0080] For critical safety commands such as "emergency stop," the system employs a multi-redundancy transmission strategy. When the emergency stop button on any module is pressed, the emergency stop broadcast is not only transmitted through the main wireless channel but also asynchronously repeated through an independent low-power secondary channel, and forwarded by adjacent modules within the network. This ensures that the command can reach all execution modules with minimal latency under any extreme network conditions. The above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of the invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A control method for a wireless modular embodied intelligent robot, characterized in that, include: S1. Define each robot functional unit that can independently complete a certain function and participate in collaboration as an autonomous module; Combining multiple autonomous modules to form a robot; S2. The autonomous modules power on and perform self-tests in sequence, establish a local wireless network, load the locally stored module description information, and join the wireless network. S3. Each module broadcasts its capability description information to the coordination module; S4. The coordination module performs self-discovery of the whole machine topology and builds a body model, and performs overall calibration and synchronization. S5. The whole machine accepts task commands or external control.

2. The wireless modular embodied intelligent robot control system according to claim 1, characterized in that, Each main module performs a self-test after being powered on, joins the robot's internal local area network through a preset wireless protocol, and completes identity authentication; Each module broadcasts its capability description information to the coordinating module, including the module identity field, interface field, pose description field, and the initial value of the current connection state.

3. The wireless modular embodied intelligent robot control system according to claim 1 or 2, characterized in that, After identity authentication is completed, the two modules that have detected the connection exchange each other's identity information, interface information and interface pose information, and establish a bidirectional consistent adjacency record; the coordination module receives the connection records reported by all modules and constructs an initial connection graph with modules as nodes and connection relationships as edges; The coordination module determines and identifies the root node of the whole machine. Starting from the root node, the coordination module traverses the initial connection graph to generate the parent-child hierarchical relationship between modules, forming a topology tree or a constrained topology graph, and determines the parent and child modules of each module in the topology structure.

4. The wireless modular embodied intelligent robot control system according to claim 3, characterized in that, For each identified set of connection modules, the coordination module determines the pose relationship between the sub-module coordinate system and the parent module coordinate system based on the actual physical interface information, interface coordinate system definition, interface orientation, mechanical installation constraints, and connection direction, and updates the transformation relationship between the sub-module's own coordinate system and the parent coordinate system. Based on the topological relationship, parent-child coordinate transformation relationship, and kinematic fields of each module, the coordination module automatically identifies the series chain, parallel branch chain, end effector chain, and degree of freedom distribution relationship of the whole machine, and generates the whole machine kinematic model, dynamic model, and power supply path diagram.

5. The wireless modular embodied intelligent robot control system according to claim 1, characterized in that, When performing topology self-discovery, if two modules declare the same connection relationship inconsistently, the connection is marked as a connection to be confirmed; if the connection direction declared by a module conflicts with the pose constraint, it is marked as an abnormal connection. If the module type does not match the current topology location, it is marked as a role conflict module; If the constructed topology graph contains closed loops that do not conform to the preset structure, then closed loop pruning, connection disabling, or special mode processing will be performed. If a critical node is missing, the system enters a degraded control mode or a safe shutdown mode; if a topology change causes the original control model to fail, the system suspends high-dynamic actions and retains only the safe action mode.

6. The wireless modular embodied intelligent robot control system according to claim 1, characterized in that, The following clock synchronization mechanism is used during overall calibration: the coordination module periodically broadcasts standard time frames; each module receives the time frame, measures its local reception time, and corrects its local time reference; the coordination module issues action parameters with future trigger timestamps; each master module begins to execute the relevant actions when the corresponding timestamp arrives. If a clock deviation exceeding a threshold is detected, the system restricts highly dynamic collaborative actions or triggers a resynchronization process.

7. The wireless modular embodied intelligent robot control system according to claim 1, characterized in that, The coordination module does not directly send the control value at each servo sampling moment, but instead sends the target trajectory parameters, action segment parameters, or abstract action primitives within a future time window; each master module performs interpolation, expansion, and closed-loop execution on the received target trajectory based on the local controller, local model, and current state; high-frequency control variables are actively generated locally by the autonomous module. Cross-module collaborative actions adopt a future timestamp aligned execution mechanism; the coordination function has fault tolerance capabilities.

8. A wireless modular embodied intelligent robot control system, used to implement the wireless modular embodied intelligent robot control method according to any one of claims 1-7, characterized in that, It consists of multiple autonomous modules, including: a power module, which may include multiple modules to provide energy and participate in intelligent control; an articulation module that constitutes the robot's various degrees of freedom; a dexterous hand module that enables multi-finger, compliant grasping and fine manipulation; a perception module, including a vision head module, an auditory module, and a tactile module; a computing and coordination module that is responsible for global task planning and module coordination control; and a communication module that builds an internal wireless network and interacts with external networks.

9. The wireless modular embodied intelligent robot control system according to claim 8, characterized in that, The modules are interconnected in the following ways: wireless communication connection for data exchange of control commands, status information and model parameters; contact connection for power supply and necessary high-speed or safety signal transmission, with ring electrodes, contact springs and slip rings at joint connections, and mechanical quick-locking and electrical contact pieces at the connections between the torso and limbs. The communication module includes a module wireless communication unit, a network coordination unit, a clock synchronization unit, a link monitoring unit, a security authentication unit, and a cooperative interface unit.

10. The wireless modular embodied intelligent robot control system according to claim 8, characterized in that, The control system adopts a three-layer control architecture: the first layer is the intra-module high-frequency closed-loop control layer, in which the internal controller of each module independently executes high-frequency servo control of position, speed, current, torque, impedance or damping; the second layer is the inter-module wireless coordination control layer, in which the coordination module exchanges target actions, status summaries and synchronization commands with each module through wireless communication; the third layer is the task planning layer, which sends task-level targets to the coordination control layer at a lower frequency.

Images