A state detection and standing planning method for a humanoid robot after falling down
By installing a piezoelectric thin film sensor and an inertial measurement unit on a humanoid robot, combined with the Mahayana distance calculation method, the problems of state detection and standing planning after falling are solved, and the robot's self-detection and standing ability improvement in complex environments is realized. .
Patent Information
- Application Number
- CN202211034551.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-08-26
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
After falling, humanoid robots cannot self-detect their state and judge the contact between the body parts and the ground, resulting in difficulty in standing. The existing technology mainly focuses on avoiding falls or protective actions, but lacks effective state detection and standing planning methods.
By installing piezoelectric thin film sensors at the key collision points of the robot, combining inertial measurement units and force sensors, the angle and connecting rod position of the computer robot joint motor, the Marstellar distance method is used to detect and plan states to determine whether the robot can stand and control joints. Arrive the minimum state to complete the standing.
The status detection and standing planning of the humanoid robot after falling down is realized, and the robot's adaptability and self-resilience ability in complex environments is improved.
Smart Images

Figure CN115256468B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of humanoid robots, and in particular to a method for detecting a state of a humanoid robot after it falls and planning its standing. Background Art
[0002] The purpose of designing humanoid robots is to help or replace humans to complete certain tasks. However, due to the complexity of the environment, humanoid robots are prone to fall when performing certain tasks. Since humanoid robots cannot sense the position of the body after falling like humans and judge whether the joints are damaged after falling, a method is needed to detect the state of the robot after falling and decide whether to continue to complete the task based on the detected state.
[0003] Existing technologies focus on studying how humanoid robots can avoid falling, or how to complete protective actions when falling is inevitable, but there is little research on robot status detection after falling. Summary of the invention
[0004] In view of this, the present invention provides a state detection and standing planning method for a humanoid robot after falling down.
[0005] The present invention achieves the above technical objectives through the following technical means.
[0006] A state detection and standing planning method for a humanoid robot after falling down:
[0007] When the robot falls, the industrial computer calculates the position and direction of each link based on the robot's direction and the motor angle value corresponding to the joint, which is recorded as S1. At the same time, the motion capture suit obtains the position and direction of each link, which is recorded as S2, and transmits it to the industrial computer.
[0008] If |S1-S2|≤E, the robot body structure is normal, the external environment of the robot is detected, and the robot standing is planned; if |S1-S2|>E, the robot body is abnormal and needs to be repaired immediately; E is the set threshold.
[0009] In the above technical solution, the external environment of the robot is detected, specifically: piezoelectric film sensors are installed at the key collision points of the robot. If the initial value of the piezoelectric film sensor changes, it is determined that the body part corresponding to the key collision point is in contact with the ground.
[0010] In the above technical solution, the key collision points include the forehead, back of the head, front chest, back, front waist, back waist, left and right palms, left and right backs of the hands, left and right knees, front heels and back heels of the robot.
[0011] In the above technical solution, the robot standing is planned, specifically: the distance between the real-time state value of the robot and the reference value corresponding to the state that must be experienced during the standing process is calculated, and the corresponding joints of the robot are controlled to reach the state corresponding to the minimum distance.
[0012] In the above technical solution, the real-time state value of the robot and the reference values corresponding to the states that must be experienced during the standing process include: the motor angle value of the joint, the position and direction of the connecting rod, the data collected by the inertial measurement unit, the data collected by the force sensor and the values of the key collision points.
[0013] In the above technical solution, when the corresponding joint of the robot is controlled to reach the state corresponding to the minimum distance, it is detected whether the joint motor angle value, the position and direction of the connecting rod, the data collected by the inertial measurement unit, the data collected by the force sensor, and the value of the key collision point are the set reference values. If so, continue to the next state transfer, otherwise repeat the state detection, find the optimal state, and control the corresponding joint to reach the optimal state.
[0014] In the above technical solution, the distance is calculated using the Mahalanobis distance calculation method.
[0015] In the above technical solution, the inertial measurement unit is installed on the head of the robot, and the force sensor is installed on the two ankles of the robot.
[0016] The beneficial effects of the present invention are as follows: the present invention first detects the state of the robot after it falls, including posture detection based on the main body and detection based on the external environment, and judges whether the robot has the ability to stand up according to the detected state. If the robot has the ability to stand up, the Mahalanobis distance calculation method is used to calculate the distance between the real-time state value and the reference value corresponding to the state that must be experienced in the standing process, and the corresponding joints of the robot are controlled to reach the state corresponding to the minimum distance value, so as to complete the standing of the robot; the present invention solves the state detection and standing planning problems of the robot in a complex environment, and improves the adaptability of the robot. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 This is a flow chart of the state detection and standing planning method of the humanoid robot after falling down according to the present invention;
[0018] Figure 2 This is a schematic diagram of the structure of the humanoid robot of the present invention;
[0019] Figure 3 This is a schematic diagram of the key collision points of the robot of the present invention;
[0020] Figure 4 This is the standing state transition diagram of the robot described in the present invention. DETAILED DESCRIPTION
[0021] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto.
[0022] When a humanoid robot is performing a task, it is easy to fall down due to the complexity of the environment. After falling down, humans expect the robot to stand up immediately and continue to complete the task. However, since humanoid robots do not have the keen perception function of the human body, after falling down, the robot cannot judge the state of its body or know which part of its body is in contact with the ground, so it is very difficult to stand up.
[0023] In view of the above problems, the present invention provides a method for detecting the state of a humanoid robot after it falls and planning its standing. Figure 1 As shown, the method includes body-based posture detection and external environment-based detection.
[0024] Figure 2 It is a humanoid robot structure designed based on bionic principles. The robot has a total of 20 degrees of freedom (i.e. 20 joints), corresponding to 20 joint encoders and 20 joint motors. An inertial measurement unit (IMU) is installed on the robot head, and a force sensor is installed on each of the robot's ankles. In addition, the robot is also wearing a motion capture suit. The motion capture suit in this embodiment is Noitom's Perception Neuron3. When the motion capture suit is working, it can record the position and direction of the connecting rods between the robot's joints in real time.
[0025] The following specifically introduces a method for detecting the state of a humanoid robot after falling down and planning its standing up according to the present invention.
[0026] Posture detection based on the body: When the robot falls (the detection method is the existing technology), the inertial measurement unit (IMU) obtains the direction of the robot and transmits it to the industrial computer. At the same time, the industrial computer obtains the motor angle values corresponding to the 20 joints of the robot, and then uses the robot's forward kinematics principle to calculate the position and direction of each connecting rod, which is recorded as S1; the process of calculating the position and direction of each connecting rod from the robot's direction and motor angle value is the existing technology. Since the joint encoder is abnormal, the connecting rod breaks or falls off after the robot falls, the position and direction of each connecting rod are obtained according to the motion capture suit installed on the robot, which is recorded as S2 and transmitted to the industrial computer at the same time. After obtaining these two arrays, the position and direction of each connecting rod are compared respectively. When the absolute value of the difference between the two is less than or equal to the set threshold value E, it is considered that the robot body structure is normal and can continue to work; when the absolute value of the difference between the two is greater than the set threshold value E, it is considered that the robot body is abnormal and needs to be repaired immediately.
[0027] Detection based on the external environment: When the posture detection based on the robot is completed and the robot body structure is normal, it is necessary to detect the external environment of the robot to complete the robot's standing task. Piezoelectric film sensors are installed at the robot's key collision points, including the robot's forehead, back of the head, front chest, back, front waist, back waist, left and right palms, left and right backs of the hands, left and right knees, front heels and back heels. Figure 3 The data collected by the piezoelectric film sensor is transmitted to the industrial computer, and the initial value of the piezoelectric film sensor is set to 0. When pressure is applied, the initial value changes to 1, which can be used to determine which part of the body is in contact with the ground. At the same time, the data collected by the force sensor is transmitted to the industrial computer in real time.
[0028] After the body-based posture detection and the external environment-based detection are completed, the motor angle values of the robot's 20 joints, the position and direction of the connecting rod, the data collected by the inertial measurement unit (IMU), the data collected by the force sensor, and the values of the key collision points (i.e., the data collected by the piezoelectric film sensor) can be obtained, and these values can be set as the array Mx.
[0029] Robot standing planning: Regarding the robot standing planning, the robot standing process is divided into several states A, B, C, etc. that must be experienced in the form of state transition. These states include the motor angle values of the robot's 20 joints, the position and direction of the connecting rod, the data collected by the inertial measurement unit (IMU), the data collected by the force sensor, and the values of the key collision points, which are set as arrays M A , M B , M C ...; When the robot wants to stand, according to the state value at this time (that is, the array Mx obtained after the state detection), the reference value M corresponding to the state that must be experienced A , M B , M C ..., use the Mahalanobis distance calculation method to calculate which reference value has the smallest distance from the current state value. Once the minimum value is found, control the corresponding joints of the robot to reach the state corresponding to the minimum value.
[0030] Since the robot may have some interference sliding after reaching the target point (the state corresponding to the minimum value), it needs to be tested again to prevent false arrival. The specific method is: after the robot reaches the target point, it detects whether the angle values of the 20 joint motors, the position and direction of the connecting rod, the data collected by the inertial measurement unit (IMU), the data collected by the force sensor, and the value of the key collision point are the set reference value (M A , M B , M C……), if yes, continue to the next state transfer; if not, repeat the above state detection, and use the Mahalanobis distance calculation method to calculate which reference value has the smallest distance from the state value of the target point, find the optimal state, and control the corresponding joint to reach this state.
[0031] Example: Figure 4 As shown in the lower left corner, when the robot falls, its state is unknown. The robot starts to start body-based posture detection. First, it obtains the direction of the robot based on the inertial measurement unit (IMU). Then, based on the motor angle values of the 20 joints, the robot uses the forward kinematics principle to calculate the position and direction of each connecting rod, and obtains a set of data recorded as S1. Then, the motion capture suit on the robot is started. Since the motion capture suit can transmit the position and direction of each connecting rod to the industrial computer in real time after startup, the transmitted data is recorded as S2. Then, the two sets of data are subtracted and the absolute value is taken. If the absolute value is less than the set threshold E, it means that each connecting rod is normal without breakage or falling off. Then the robot will start detection based on the external environment; otherwise, the robot stops working. If the robot can work normally, it starts to detect the key collision points. Taking the figure in the lower left corner as an example, the piezoelectric film sensor readings corresponding to the back, lower back, right knee, right leg, and heel of the robot's key collision points all change, indicating that these key collision points are in contact with the ground. The value corresponding to this state is set as array Mx = [motor angle values of 20 joints; position and direction of connecting rod; data collected by inertial measurement unit (IMU); data collected by force sensor; value of key collision point]. Since the robot stores several states that the robot must go through during the standing process, such as state A, state B, state C in the figure, etc., the corresponding value is set as M A , M B , M C ...; Then use the Mahalanobis distance to calculate which reference state the state value Mx is closer to at this time. After calculation and comparison, it is found that state A is the closest, so the corresponding joint of the robot is controlled to reach state A. Because the robot's soles have interference sliding in the process of reaching state A, although the joint motor has reached the corresponding angle, it is possible that other parameter values have not been reached, so it needs to be tested again. If the detected state value is correct, it means that the robot has reached the target state and can continue to transfer to the next target state, as shown in state B, followed by state C, and finally standing; if the detected state value is incorrect, for example, it is disturbed during the standing process, and the robot changes from an unknown state to state H. When it reaches state H, it starts to detect whether the angle values of the 20 joint motors, the position and direction of the connecting rod, the data collected by the inertial measurement unit (IMU), the data collected by the force sensor, and the value of the key collision point are the set reference value M. A, it is found that it is not the expected state A, so according to the current state, find the optimal state, the search result is state D, control the robot joints to reach the target state D, then state C, and finally stand.
[0032] The embodiments are preferred implementations of the present invention, but the present invention is not limited to the above-mentioned implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present invention belong to the protection scope of the present invention.
Claims
1. A method for detecting and planning the state of a humanoid robot after it falls down. Features: When the robot falls, the industrial computer calculates the position and direction of each link based on the robot's direction and the motor angle value corresponding to the joint, which is recorded as S1. At the same time, the motion capture suit obtains the position and direction of each link, which is recorded as S2, and transmits it to the industrial computer. If |S1-S2|≤E, the robot body structure is normal, the external environment of the robot is detected, and the robot standing is planned; if |S1-S2|>E, the robot body is abnormal and needs to be repaired immediately; E is the set threshold; Planning the robot to stand is as follows: calculating the distance between the robot's real-time state value and the reference value corresponding to the state that must be experienced during the standing process, and controlling the robot's corresponding joints to reach the state corresponding to the minimum distance; When the corresponding joint of the robot is controlled to reach the state corresponding to the minimum distance, detect whether the joint motor angle value, the position and direction of the connecting rod, the data collected by the inertial measurement unit, the data collected by the force sensor, and the value of the key collision point are the set reference values. If so, continue to the next state transfer, otherwise repeat the state detection, find the optimal state, and control the corresponding joint to reach the optimal state.
2. The method for detecting and planning the state of a humanoid robot after falling down according to claim 1, It is characterized in that The external environment of the robot is detected, specifically: piezoelectric film sensors are installed at the key collision points of the robot. If the initial value of the piezoelectric film sensor changes, it is determined that the body part corresponding to the key collision point has contacted the ground.
3. The method for detecting and planning the state of a humanoid robot after falling down according to claim 2, It is characterized in that The key collision points include the forehead, back of the head, front chest, back, front waist, back waist, left and right palms, left and right backs of the hands, left and right knees, front heels and back heels of the robot.
4. The method for detecting the state of a humanoid robot after falling down and planning its standing up according to claim 1, It is characterized in that The real-time state value of the robot and the reference values corresponding to the states that must be experienced during the standing process include: the motor angle value of the joint, the position and direction of the connecting rod, the data collected by the inertial measurement unit, the data collected by the force sensor and the value of the key collision point.
5. The method for detecting the state of a humanoid robot after falling down and planning its standing up according to claim 1, It is characterized in that The distance is calculated using the Mahalanobis distance calculation method.
6. The method for detecting the state of a humanoid robot after falling down and planning its standing up according to claim 5, It is characterized in that The inertial measurement unit is installed on the head of the robot, and the force sensor is installed on the two ankles of the robot.
