A chip structure of an embodied robot power system

Abstract

The present invention discloses a chip structure of an embodied robot power system, which relates to the field of integrated circuit technology, including a power system master control module, and a plurality of sub-power modules connected to the power system master control module through a sub-power module interface; the sub-power module includes a motor and a motor drive chip; the power system master control module can process the information of all sub-power modules in parallel, and control each sub-power module in parallel. The present invention reduces the time of information transmission and processing, so that the robot can respond to external environmental changes and internal state changes more quickly, and the parallel control mechanism enables each sub-power module to receive control signals and perform actions at the same time, reducing the delay of action execution and improving the fluency and coordination of actions.

Description

Technical Field

[0001] The present invention relates to the field of integrated circuit technology, and in particular to a chip structure of an embodied robot power system. Background Art

[0002] Based on the large language model and combined with the visual model, many embodied AI models have been developed, which can operate objects according to instructions and real-time observed images. The cloud AI model is responsible for planning and macro-control, and the AI ​​model is responsible for real-time processing of the current tasks of the embodied robot and issuing instructions to control the power system in real time. These small models have the function of recognizing the corresponding language and operation of the dialogue or task output. The AI ​​model acts as the brain, interacting with the real world through the robot body, and completing various human-like actions.

[0003] The existing chip structure of embodied robot power system is as follows Figure 1As shown in the figure, the central processor is used as the intelligent control center. The action control is obtained by processing the instructions, camera and sensor information. After calculation, the control signal output of each joint is obtained, including the speed, angle and other parameters of each motor rotation. Then the MCU (microprocessor) of each active joint sub-power system (1,2...n) receives the control signal and decodes it. The MCU controls the motor operation according to the initial state of the motor and the real-time encoding information. The MCU continuously outputs the motor drive signal according to the received instructions to control the motor operation state until the set value is reached. The coordinated operation of these motors is reflected in the action of the embodied robot. Since it is difficult for each sub-power system to communicate with each other, each sub-power system is directly controlled by the MCU of each sub-power system according to the received instructions, which requires each motor to operate with high precision. If one motor operates abnormally, it will cause abnormal action, which will seriously affect the center of gravity and the stability of the robot. During the actual action of the robot, the central processor needs to gradually correct the motor rotation parameters through feedback signals such as sensors and camera visual information, so as to correct the action. For example, when the robot walks, the MCU will control the motor to make the thigh step out and the arm swing according to the pre-calculated parameters output by the central processor. However, when the feet touch the uneven ground and other abnormal situations, the pre-set parameters will cause the body's center of gravity to be unstable. Then the central processor needs to quickly adjust the motors of the thighs and soles of the feet after receiving the data of the center of gravity sensor such as the gyroscope, so that the body's center of gravity can be restored to balance. However, in the prior art system, the adjustment from the center of gravity offset to the output motor needs to go through multiple steps such as sensor detection output, central processor reception and processing, MCU reception and processing, and motor drive. The processing path is long and the response time is also long. As a result, the robot cannot adjust the center of gravity quickly, resulting in obvious center of gravity instability, which seriously affects the reliability of the robot. At the same time, in the prior art solution, each motor operates independently, and the operation status of each motor needs to be fed back to the MCU by the driver chip and the motor, and then fed back to the central processor by the MCU for coordination. It also leads to the problem of slow response speed of motor coordinated operation. Therefore, the robot using the prior art solution has a long response time for action adjustment and the motor does not operate in a coordinated manner, which makes the robot appear clumsy and prone to obvious problems such as excessive center of gravity offset, stiff and uncoordinated movements. Summary of the invention

[0004] The present invention provides a chip structure of an embodied robot power system, which can solve the above-mentioned problems.

[0005] In order to solve the above problems, the technical solution adopted by the present invention is as follows:

[0006] The present invention provides a chip structure of a power system of an embodied robot, comprising a power system master control module, and a plurality of sub-power modules connected to the power system master control module via a sub-power module interface; the sub-power modules comprise motors and motor drive chips; the power system master control module can process information of all sub-power modules in parallel, and control each sub-power module in parallel.

[0007] As a further description of the above technical solution, the power system master control module includes a sensor signal receiving and processing module, a coded signal receiving and processing module, a parallel instruction decoding module, a parallel processor, an interface module, and a real-time read-write control module.

[0008] As a further description of the above technical solution, the parallel instruction decoding module is used to process the instructions output by the artificial intelligence model in real time and in parallel, and respectively calculate the instruction parameters in the processing instructions, including action category, coordinate information, force information, target volume and angular direction information of limbs, select the corresponding motor information according to the action category, determine the rotation angle of the corresponding motor according to the coordinate information, obtain the motor control information through the force information, determine the corresponding operation point through the target volume, and determine the rotation angle of the limbs through the angular direction information of the limbs.

[0009] As a further description of the above technical solution, the sensor signal receiving and processing module is used to receive and process various sensor signals, including: receiving gyroscope signals to calculate the center of gravity offset and acceleration information; receiving pressure sensor information, calculating the pressure change, and judging whether the pressure change has a sliding trend; calculating the total contact surface pressure, and estimating the contact surface friction coefficient based on the friction coefficient.

[0010] As a further description of the above technical solution, the coded signal receiving and processing module is used to continuously receive the coded signals of each motor and calculate the current rotation angle, total rotation angle, current position and current speed of the corresponding motor.

[0011] As a further description of the above technical solution, the parallel processor can process multiple tasks simultaneously, and it establishes a data read and write relationship with the memory through the real-time read and write control module; the parallel processor can receive data from the parallel instruction decoding module, the sensor signal receiving and processing module, and the encoding signal receiving and processing module, and calculate the speed and duty cycle of the corresponding motor, and output the motor drive control signal to the corresponding motor drive chip; the parallel processor can call the action information in the memory according to the specific action to realize the standard action, and the action information stored in the memory includes the rotation angle and duty cycle information of each motor.

[0012] As a further description of the above technical solution, if an action is repeatedly executed, the parallel processor can store the information of the action into the memory, name the action, and upload the action to the artificial intelligence model. The next time the artificial intelligence model can directly call the information of the action; during the execution of the action, the parallel processor can receive sensor data in real time and make real-time adjustments.

[0013] As a further description of the above technical solution, the interface module is used to send the motor control information output by the parallel processor to the corresponding motor drive chip, and upload the motor encoding information to the parallel processor.

[0014] As a further description of the above technical solution, the sub-power module interface is a half-duplex or full-duplex interface, which can download generator control information and upload motor encoding information.

[0015] As a further description of the above technical solution, the motor drive chip includes a communication interface, a control information decoding module, a coding information conversion module, a motor position, speed and current loop control module, and a power drive module; the communication interface is used to receive the motor coding information uploaded by the coding information conversion module, and send the motor control information to the control information decoding module; the coding information conversion module is responsible for receiving the motor coding signal fed back by the motor and performing information format conversion; the control information decoding module is used to decode and convert the motor control information sent by the communication interface; the motor position, speed and current loop control module is used to generate three motor pulse control signals PWMU, PWMV and PWMW according to the pulse signal from the control information decoding module and the motor coding information from the coding information conversion module, and output them to the power drive module to obtain three-phase motor drive signals U, V and W.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] Since the power system master control module can process the information of all sub-power modules in parallel, the time for information transmission and processing is reduced, so that the robot can respond to changes in the external environment and internal state more quickly; the parallel control mechanism enables each sub-power module to receive control signals and execute actions at the same time, reducing the delay in action execution and improving the fluency and coordination of actions;

[0018] The motor drive and motor form a sub-power module, which communicates with the power system master control through a standard interface, which is conducive to the long-distance layout of the power module. The parallel processor integrates all motor control operations to improve the coordination of the operation of each motor. At the same time, it receives and processes sensor and motor encoding feedback information in parallel to realize dual-feedback fast-loop control function, so that the center of gravity and movement of the embodied robot can be quickly adjusted and fine-tuned in real time, and problems such as excessive center of gravity offset, stiff and uncoordinated movement are less likely to occur.

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the embodiments of the present invention are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 It is a schematic block diagram of the chip structure of the existing embodied robot power system;

[0022] Figure 2 It is the chip structure of the embodied robot power system proposed in the embodiment of the present invention;

[0023] Figure 3 is a schematic block diagram of the overall control structure of a power system according to an embodiment of the present invention;

[0024] Figure 4 is a schematic block diagram of the internal structure of a motor driver chip according to an embodiment of the present invention;

[0025] Figure 5 It is a control flow diagram of the power system chip structure of an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0027] Please refer to Figure 2An embodiment of the present invention provides a power system chip structure of an embodied robot, including a power system master control module, a memory, and a plurality of sub-power modules connected to the power system master control module through a sub-power module interface; the sub-power module includes a motor and a motor driver chip; the power system master control module can process information of all sub-power modules in parallel, and control each sub-power module in parallel.

[0028] The artificial intelligence model obtains information such as object category, operation position, operation posture, etc. through reasoning based on input information from cameras, instructions, sensors, etc., and then outputs specific action instructions to the power system chip structure based on the received instructions.

[0029] The power system master control module can be implemented by a customized parallel processing chip or a programmable logic gate array FPGA.

[0030] The power system master control module can receive the instructions output by the artificial intelligence model in parallel, receive the sensor and motor encoding information after parallel decoding, obtain the driving information of each motor through parallel calculation, output it to the motor driver chip through the sub-power module interface, and then control the motor rotation. The parallel calculation includes reasonably allocating the working state of each motor according to the artificial intelligence model instructions and the fixed action information in the memory, calculating the speed and angle of each motor, reading the initial control information of each motor from the memory to calculate the distance of the target coordinates and the rotation angle of each motor, receiving the sensor feedback information and calculating the motor rotation parameters, and receiving the motor encoding information through the sub-power module interface to control the speed, position and current loop of the motor drive. The power system master control module communicates with the motor driver chip through the sub-power module interface, and the power system master control module outputs the motor control information to the motor driver chip, and the motor driver chip uploads the motor encoding information to the power system master control module. The motor driver chip receives the speed and duty cycle information from the sub-power module interface and converts it into a pulse signal, and then outputs a 3-phase motor drive signal to drive the motor to rotate. The motor outputs the encoding information to the power system master control module to realize the closed-loop control of the motor.

[0031] In an embodiment of the present invention, a sub-power module is responsible for the rotation control of a joint. The sub-power module communicates with the power system master control using a standard sub-power module interface. The sub-power module interface is a half-duplex or full-duplex interface. The power system master control module outputs the speed and duty cycle information of the control motor to the motor driver chip, and the motor driver chip uploads the motor encoding information to the power system master control module. The power system master control module receives the sensor signal and the uploaded motor encoding information to realize dual-loop motor control and parallel processing structure, which is easy to realize. The present invention proposes a gentle motor control method of first small torque output control and then gradually adjusting to the appropriate torque, so that the embodied robot can quickly and accurately realize various anthropomorphic actions.

[0032] For example, when the embodied robot is walking, the thigh and hand joint motors first work at a fast and small torque, so that the hands and legs can swing easily. When the soles of the feet touch the ground, the force increases, and the coding information fed back by the sub-power module can be quickly used to determine whether the motor is stuck, that is, the power system master control module sends the motor rotation control information, but the actual feedback code information indicates that the motor does not achieve the corresponding rotation according to the control information. The power system master control module can determine the force size through the stuck rotation, re-send the control information through the sub-power module interface, adjust the rotation control information of the motor drive, adjust the frequency and duty cycle of the motor drive pulse, and increase the motor torque, thereby realizing a fast response loop of the motor control. For example, after the sole of the foot touches the ground, assuming that there is a protrusion on the ground that is not recognized by the artificial intelligence model, when the knee joint of the leg is straightened according to the preset control, it has not been fully straightened, but the sole of the foot first encounters the protrusion and is stressed. Since the present invention adopts small torque control, the motor fails to rotate to the set angle after being stressed, that is, a stuck situation occurs. The motor encoding information is uploaded to the power system master control module in real time. After calculation, the power system master control module finds that there is a stuck motor, and then increases the duty cycle in time to increase the torque to stabilize the center of gravity. This is a fast control loop proposed by the present invention, in which a motor initially rotates with a small torque, and then reduces the speed and increases the torque after being stressed to slowly adjust to a suitable torque. The existing robot control technology does not have this fast control loop. Each motor rotates according to a pre-calculated large torque, and the sole of the foot will contact the road protrusion with a large force. The robot will be abnormally stressed and a large center of gravity offset will occur, resulting in unstable control. In the embodiment of the present invention, the power system master control module simultaneously receives sensor information such as pressure sensors, calculates the change of pressure or center of gravity, and determines whether the contact surface with the object slides and whether the center of gravity of the robot is unstable. The power system master control module timely outputs adjustment control information to adjust the robot's movements to avoid sliding and adjust the center of gravity. In this way, the control loop of adjusting the robot's movements through sensors is realized. The above dual-loop control enables the embodied robot to walk quickly and stably.

[0033] For example, when reaching out to pick up an egg, the artificial intelligence model sends an egg-picking instruction and coordinate information to the power system master control module. After receiving the information, the power system calculates whether the robot needs to be moved based on the current position of the robot and the state of the hand. If the hand cannot reach the coordinate point, it controls the walking first, and then controls the action of raising the hand to pick up the egg, or the artificial intelligence model directly sends a walking instruction and an egg-picking instruction. After the power system receives the instruction and decodes it, it first grabs it with the minimum initial force. When it detects that the motor force reaches the set value and the finger detects the pressure, it starts to slowly move the wrist up and determines whether the grasping force is sufficient by detecting the pressure change of the finger pressure sensor array. If the pressure of the finger pressure sensor array changes downward, it means that the egg is sliding, and the motor drive duty cycle is increased, and the finger drive torque is increased until the pressure of the finger pressure sensor is stable. At this time, raise the arm and pick up the egg. In the process of picking up the egg, the changes of the finger pressure sensor are detected in real time to ensure that the egg does not slide. In this way, with the help of a fast-response feedback control loop and a control strategy of gradually increasing the torque, the robot can easily pick up fragile eggs without crushing the eggs due to excessive finger pressure. If the initial operation crushes the egg, the maximum torque of the motor and the pressure parameters of the pressure sensor are stored in the memory to prevent the next operation from exceeding the torque and pressure values.

[0034] like Figure 3 As shown, the power system master control module can be implemented by a customized parallel processing chip or a programmable logic gate array FPGA, including a sensor signal receiving and processing module, a coded signal receiving and processing module, a parallel instruction decoding module, a parallel processor, an interface module, and a real-time read-write control module.

[0035] The parallel instruction decoding module is used to process the instructions output by the artificial intelligence model in real time and in parallel, and calculate the instruction parameters in the processing instructions respectively, including action category, coordinate information, force information, target volume and angular direction information of limbs, select the corresponding motor information according to the action category, determine the rotation angle of the corresponding motor according to the coordinate information, obtain the motor control information through the force information, determine the corresponding operation point through the target volume, and determine the rotation angle of the limbs through the angular direction information of the limbs.

[0036] The sensor signal receiving and processing module is used to receive and process various sensor signals, including: receiving gyroscope signals to calculate the center of gravity offset and acceleration information; receiving pressure sensor information, calculating the pressure change, and judging whether the pressure change has a sliding trend; calculating the total pressure of the contact surface, and estimating the contact surface friction coefficient based on the friction coefficient.

[0037] The encoding signal receiving and processing module is used to continuously receive the encoding signals of each motor and calculate the current rotation angle, total rotation angle (accumulated rotation angle at the beginning of the action), current position and current speed of the corresponding motor.

[0038] The parallel processor can handle multiple tasks at the same time, access the memory data through the real-time read-write control module, read the motor drive information corresponding to the action, and store the repeated actions after the control is completed. The parallel processor receives the decoding results of the parallel instruction, receives the output information of the sensor signal receiving and processing module and the output information of the coded signal receiving and processing module, calculates the motors that need to work and the speed and duty cycle of each motor, and outputs the motor drive control signal to the corresponding motor drive chip. The calculation is the target drive control method, that is, calculating the rotation angle of each motor so that the hands, feet or body parts move to the target coordinate position. The parallel processor can call the action information in the storage according to the specific action to achieve the standard action. The action information in the memory contains the rotation angle and duty cycle information of each motor. If an action is repeatedly executed, the parallel processor can store the action information in the memory, name the action, and upload the action to the artificial intelligence model module. The artificial intelligence model can directly call the action instruction next time. Regardless of the action information obtained by calculation or the action information called from the storage, the sensor data will be received in real time and adjusted in real time during the execution of the action to avoid the occurrence of center of gravity loss, excessive pressure, etc.

[0039] The interface module is used to send the motor control information output by the parallel processor to the corresponding motor drive chip, and upload the motor encoding information to the parallel processor.

[0040] like Figure 4 As shown, the motor drive chip includes a communication interface, a control information decoding module, a coding information conversion module, a motor position, speed and current loop control module, and a power drive module.

[0041] The power drive module, i.e. the motor drive switching device, can use the third-generation semiconductor GaN power device to reduce drive loss, speed up the switch control speed, and achieve fast and accurate control of the motor.

[0042] The communication interface is used to receive the motor encoding information uploaded by the encoding information conversion module, and send the motor control information to the control information decoding module. The encoding information conversion module is responsible for receiving the motor encoding signal fed back by the motor and converting the information format, such as converting the received incremental encoding information into binary encoding information. The control information decoding module is used to decode and convert the motor control information sent by the communication interface, and convert the binary format in the control information into a pulse signal mode. If a total of 4096 levels of resolution are defined, if the control information is 2048, a 50% duty cycle pulse output is output. The motor position, speed and current loop control module is used to generate three motor pulse control signals PWMU, PWMV and PWMW according to the pulse signal from the control information decoding module and the motor encoding information from the encoding information conversion module, and output them to the power drive module to obtain three-phase motor drive signals U, V and W.

[0043] like Figure 5 The figure shows the control flow chart of the power system chip structure of the embodiment of the present invention. First, the artificial intelligence model inputs the instruction: operate the object at the coordinates (x1, y1, z1) with F1 force, the finger is facing the w degree direction, and the volume length, width and height are (a, b, h).

[0044] The parallel instruction decoding module processes the instructions output by the artificial intelligence model in real time and in parallel, and calculates the parameters such as action, coordinates, strength, target volume, angle direction of hands and feet in the processing instructions. The distance is calculated according to the coordinates (x1, y1, z1), and the rotation angle of each motor is calculated to control whether the hand can reach the target coordinate. If it can, the rotation angle of each motor is calculated. If the hand cannot reach the target coordinate point even if the body is turned, it is judged that the distance is not appropriate and the instruction to walk first needs to be output, and the walking direction and distance are calculated at the same time.

[0045] At the same time, the corresponding motor information is retrieved according to the action for calculation. For example, the grasping action needs to control the shoulder, elbow, palm and finger joint motors, and even the body rotation motor. The action sequence is calculated, such as moving the shoulder motor first and then extending the elbow joint, while adjusting the palm joint and finger opening.

[0046] The corresponding motor drive duty cycle is calculated through the force information F1. The power system implements a control strategy of first controlling with a small force and then gradually increasing the force through a feedback loop, thereby reducing the motor drive duty cycle and reducing the initial force from F1 to F0.

[0047] The angle at which the finger needs to be opened and the operating angle of the finger motor are calculated through the object category or the volume information length, width and height (a, b, h) in the instruction.

[0048] According to the finger direction angle information, calculate the angle that the palm needs to rotate, whether it is clockwise or counterclockwise, and calculate the palm joint motor movement angle according to the w degree direction.

[0049] The above parallel decoding results will be input to the parallel processors at the same time.

[0050] The sensor signal receiving and processing module receives gyroscope signals, pressure sensor signals, temperature sensor signals and other signals in parallel, and performs calculations and processing. For example, the gyroscope signal is received to calculate the center of gravity offset and acceleration information. There are multiple pressure sensors distributed in multiple places such as fingers and soles, especially the finger surface is distributed with a pressure sensor array. Receive each pressure sensor, calculate the pressure change, and judge whether the contact surface has a sliding trend by the direction of pressure change. Calculate the total pressure of the contact surface, and estimate the contact surface friction and other information based on the friction coefficient. If the contact surface has a sliding trend, it means that the grip is insufficient. The output force is adjusted, and the motor duty cycle is increased to increase the motor torque, thereby increasing the grip force and increasing the contact pressure between the finger and the egg. Repeat this until the finger contact surface does not show a sliding trend when the hand is slightly raised, then the hand can be raised. If there is a sliding trend during the movement of the hand, the torque is increased to strengthen the grip. Such an adaptive learning strategy can pick up the egg safely and stably. If the grip is too strong during the operation and the egg is crushed, the pressure of the egg is stored in the storage, and the pressure value will be read the next time the egg is grabbed to avoid the finger grip being greater than this value. This forms a closed loop of adaptive learning, which is equivalent to breaking down the action of grabbing an egg into multiple micro-movements. This makes it easy to pick up fragile objects, and the gradual increase in pressure ensures that the object is firmly grasped while avoiding damage to the object due to excessive pressure.

[0051] The encoding signal receiving and processing module continuously receives the encoding signals of each motor, and calculates the current rotation angle, total rotation angle (accumulated rotation angle at the beginning of the action), current position and current speed of the motor.

[0052] The parallel processor can handle multiple tasks at the same time, and access data to the memory in real time through the real-time read-write control module. Receive the decoding results of the parallel instruction, receive the output information of the sensor signal receiving and processing module and the output information of the coded signal receiving and processing, calculate the speed, torque and other parameters of the motors that need to work and the rotation speed of each motor, and output the motor drive control signal to the motor loop control array. The operation is the target drive control mode, that is, calculate the rotation angle of each motor so that the hands, feet or body parts move to the target coordinate position. If there is a sliding trend, the signal is quickly processed in the parallel processor, and the duty cycle information of the joint motor with the sliding trend is increased. The parallel processor can call the action information in the storage according to the specific action to achieve the standard action. The action information in the storage contains information such as the rotation speed, angle and duty cycle of each motor. If an action is repeated, the parallel processor can store the action information in the storage, name the action, and upload the action to the artificial intelligence model module, and the artificial intelligence model can directly call the action instruction next time. Regardless of the action information obtained by calculation or called from storage, sensor data will be received in real time and real-time adjustments will be made during the execution of the action to avoid situations such as loss of center of gravity control and excessive pressure.

[0053] The parallel processor outputs the pulse signal of the driving motor, and realizes the position, speed and current loop control of the motor by receiving the coded information fed back by the motor. It ensures that the motor rotates according to the output control signal and ensures that the motor rotates to the preset position.

[0054] The rotation of the motor is accompanied by the robot's action. When the robot outputs an action, the visual feedback from the camera is input into the artificial intelligence model module. The edge re-outputs the new coordinates (xn, yn, zn) to the power system master control based on the visual feedback. The parallel decoding mode re-decodes the instructions, and the parallel processor recalculates the rotation parameters of each motor. If the artificial intelligence model does not output a new instruction, the power system master control continues to run according to the original instruction until the egg-picking instruction is realized.

[0055] During the continuous control of the motor, the motor encoding signal input is also received to realize the motor operation loop control, including three loop controls of position, speed and current, to ensure that the motor runs accurately according to the control signal and the motor rotation is adjusted in time.

[0056] The embodied robot power system chip structure described in the embodiment of the present invention can enable the power system to process the operating status of each motor in parallel, receive the coded information and real-time sensor signals fed back by each motor in parallel, implement dual-loop control of the motor, and greatly shorten the response time of the motor feedback control loop. This can achieve rapid adjustment and real-time fine-tuning of the operation of each motor, thereby structurally solving the problems of robot center of gravity stability and motion coordination, making the embodied robot's movements more flexible and closer to human movements.

[0057] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A chip structure of an embodied robot power system, characterized in that: It includes a power system master control module, and a plurality of sub-power modules connected to the power system master control module through a sub-power module interface; the sub-power module includes a motor and a motor drive chip; the power system master control module can process information of all sub-power modules in parallel, and control each sub-power module in parallel; The power system master control module includes a sensor signal receiving and processing module, a coded signal receiving and processing module, a parallel instruction decoding module, a parallel processor, an interface module, and a real-time read-write control module; The control method of the power system chip structure includes: The action category, coordinate information, force information, target volume, and hand and foot angle direction information in the instructions output by the artificial intelligence model are decoded in parallel through a parallel instruction decoding module; the distance is calculated according to the coordinate information and the judgment result data of whether to execute the walking action first is obtained; the parallel processor receives the judgment result data of whether to execute the walking action first, the action category, coordinate information, force information, target volume, hand and foot angle direction information, the sensor signal data from the sensor signal receiving and processing module, the motor encoding signal processing output information from the encoding signal receiving and processing module, and the initial information of the motor; The parallel processor calculates the received data to obtain the order of execution of each action, the corresponding action motor array, the speed, position and duty cycle of each motor, and gradually outputs the control signal of each motor; the initial control first reduces the motor control torque; the robot finger surface is distributed with a pressure sensor array, and the direction of pressure change is used to determine whether the contact surface has a tendency to slide. If the contact surface has a tendency to slide, the output force is adjusted to increase the motor torque until the motor torque is stopped when there is just no sliding; if an action is repeatedly executed, the parallel processor stores the action information in the memory, names the action, and directly calls the action instruction next time; the motor control signals output by the parallel processor are directly output as motor drive signals to drive the motor to work after being controlled by the motor operation loop.

2. The chip structure of the embodied robot power system according to claim 1, characterized in that: The coded signal receiving and processing module is used to continuously receive the coded signals of each motor and calculate the current rotation angle, total rotation angle, current position and current rotation speed of the corresponding motor.

3. The chip structure of the embodied robot power system according to claim 1, characterized in that: The parallel processor can handle multiple tasks simultaneously, and it establishes a data read and write relationship with the memory through the real-time read and write control module; the parallel processor can receive data from the parallel instruction decoding module, the sensor signal receiving and processing module, and the encoding signal receiving and processing module, and calculate the speed and duty cycle of the corresponding motor, and output the motor drive control signal to the corresponding motor drive chip; the parallel processor can call the action information in the memory according to the specific action to realize the standard action, and the action information stored in the memory includes the rotation angle and duty cycle information of each motor.

4. The chip structure of the embodied robot power system according to claim 3, characterized in that: If an action is repeatedly executed, the parallel processor can store the information of the action into the memory, name the action, and upload the action to the artificial intelligence model. The artificial intelligence model can directly call the information of the action next time. During the execution of the action, the parallel processor can receive sensor data in real time and make real-time adjustments.

5. The chip structure of the embodied robot power system according to claim 1, characterized in that: The interface module is used to send the motor control information output by the parallel processor to the corresponding motor drive chip, and upload motor encoding information to the parallel processor.

6. The chip structure of the embodied robot power system according to claim 1, characterized in that: The sub-power module interface is a half-duplex or full-duplex interface, which can download generator control information and upload motor encoding information.

7. The chip structure of the embodied robot power system according to claim 6, characterized in that: The motor drive chip includes a communication interface, a control information decoding module, a coding information conversion module, a motor position, speed and current loop control module, and a power drive module; the communication interface is used to receive the motor coding information uploaded by the coding information conversion module, and send the motor control information to the control information decoding module; The coding information conversion module is responsible for receiving the motor coding signal fed back by the motor and performing information format conversion; The control information decoding module is used to decode and convert the motor control information sent by the communication interface; the motor position, speed and current loop control module is used to generate three motor pulse control signals PWMU, PWMV and PWMW according to the pulse signal from the control information decoding module and the motor coding information from the coding information conversion module, and output them to the power drive module to obtain three-phase motor drive signals U, V and W.

Images