Adaptive humanoid robot gait optimization system based on deep learning

Abstract

The invention relates to the technical field of gait optimization systems, in particular to a self-adaptive humanoid robot gait optimization system based on deep learning, which comprises an analysis module, a control module and a communication module, the analysis module is used for processing related information of the step length, the gradient, the friction force and the load and transmitting the information to the control module; the control module obtains a climbing step length reference index according to related information of the step length, the gradient, the friction force and the load and transmits the climbing step length reference index to the communication module; the communication module transmits the climbing step length reference index to a robot foot control module; and the robot foot control module adjusts the climbing parameters according to the climbing step length reference index. The robot can automatically adjust gaits according to external environment changes such as gradient, friction force and load, the climbing requirement is met, and the self-adaptive capacity is high.

Description

Technical Field

[0001] The present invention relates to the technical field of gait optimization systems, and particularly to an adaptive humanoid robot gait optimization system based on deep learning. Background Art

[0002] The gait optimization system mainly dynamically optimizes the gait of the robot by real-time detecting the gait parameters of the robot (such as step length, step frequency, support angle, joint movement) and combining with sensor data (such as slope, friction, center of gravity position, etc.), so that it can adapt to complex environments, reduce energy consumption, and improve stability.

[0003] The application document with the publication number of CN103895020B discloses a method and system for controlling the gait of a robot, including: determining whether the robot is walking and the direction in which the robot is walking; measuring the amount of time that the sole of one foot of the robot is in contact with the ground; calculating a hypothetical reaction force applied to the sole of the foot using trigonometric functions, where the trigonometric functions use the measured amount of time that the sole of the foot is in contact with the ground as a period; and applying the calculated hypothetical reaction force to the Jacobian transpose matrix and converting the hypothetical reaction force into a driving torque for the lower limb joints of the robot.

[0004] The existing gait control depends on a static mathematical model and lacks adaptability. Summary of the Invention

[0005] The purpose of the present invention is to propose an adaptive humanoid robot gait optimization system based on deep learning for the above-mentioned deficiencies.

[0006] The present invention adopts the following technical solutions:

[0007] An adaptive humanoid robot gait optimization system based on deep learning, the system includes an analysis module, a control module and a communication module; the analysis module is used to process information related to step length, slope, friction and load, and transmit it to the control module; the control module obtains a climbing step length reference index according to the information related to step length, slope, friction and load, and transmits it to the communication module; the communication module transmits the climbing step length reference index to the robot foot control module; the robot foot control module adjusts the climbing parameters according to the climbing step length reference index.

[0008] Optionally, the analysis module includes a slope analysis sub-module, a load analysis sub-module and a data storage sub-module;

[0009] The slope analysis sub-module is used to analyze and obtain the total slope length, total slope height, slope angle detected each time, and slope surface friction, and transmit them to the control module; the load analysis sub-module is used to analyze and obtain the total load of the robot, and transmit it to the control module; the data storage sub-module is used to store the leg length of the robot, and transmit it to the control module;

[0010] The control module obtains the load adjustment coefficient according to the total load of the robot, obtains the total number of slope detections according to the total slope length and total slope height, obtains the average slope angle according to the slope angle detected each time and the total number of slope detections, obtains the slope adjustment coefficient according to the average slope angle, obtains the step length ratio index according to the leg length of the robot, obtains the flat ground step length reference index according to the step length ratio index and the leg length of the robot, obtains the climbing step length calculation index according to the flat ground step length reference index, slope adjustment coefficient, average slope angle, slope surface friction, load adjustment coefficient and total load of the robot, and obtains the climbing step length reference index according to the flat ground step length reference index and the climbing step length calculation index.

[0011] Optionally, the slope analysis sub-module includes a visual analysis unit, a slope analysis unit and a friction analysis unit; the visual analysis unit is used to analyze and obtain the total slope length and total slope height, and transmit them to the control module; the slope analysis unit is used to analyze and obtain the slope angle detected each time, and transmit it to the control module; the friction analysis unit is used to analyze and obtain the slope surface friction, and transmit it to the control module.

[0012] Optionally, the visual analysis unit includes a camera, an image pre-processor, a slope recognizer and a slope calculator; the camera is used to capture an initial image; the image pre-processor uses the method of Gaussian filtering to reduce the noise of the initial image; the slope recognizer identifies the boundary of the slope by the method of edge detection; the slope calculator obtains the total slope length and total slope height according to the boundary of the slope, and transmits them to the control module.

[0013] Optionally, the friction analysis unit includes a force sensor and a friction calculator; the force sensor is used to detect and obtain the normal force and tangential friction force; the friction calculator obtains the slope angle detected each time according to the normal force and tangential friction force, and transmits it to the control module.

[0014] Optionally, the control module calculates the climbing step length reference index, which satisfies the following formula: where LT is the climbing step length reference index,

[0015] L pd is the flat ground step length reference index, L opt is the climbing step length calculation index.

[0016] The beneficial effects achieved by the present invention are:

[0017] 1. The robot can automatically adjust its gait according to external environmental changes such as slope, friction, and load, adapt to the needs of climbing slopes, has strong adaptability, and enhances stability.

[0018] 2. The optimized gait data can be remotely transmitted to achieve remote control and collaborative optimization of multiple robots.

[0019] 3. Optimizing the gait reduces energy waste and makes the robot more efficient during long-term operation.

[0020] To enable a further understanding of the features and technical content of the present invention, please refer to the following detailed description of the present invention and the accompanying drawings. However, the provided drawings are only for reference and illustration and are not used to limit the present invention. Description of the Drawings

[0021] Figure 1 It is a schematic diagram of the overall structure of the present invention;

[0022] Figure 2 It is a schematic diagram of the structure of the analysis module in the present invention;

[0023] Figure 3 It is a schematic diagram of the structure of the slope analysis sub-module in the present invention;

[0024] Figure 4 It is a schematic diagram of the structure of the visual analysis unit in the present invention;

[0025] Figure 5 It is a schematic diagram of the structure of the friction analysis unit in the present invention;

[0026] Figure 6 It is an effect diagram of the present invention;

[0027] Figure 7 It is a schematic diagram of the overall structure of the second embodiment of the present invention;

[0028] Figure 8 It is a schematic diagram of the structure of the actual step length detection sub-module in the second embodiment of the present invention;

[0029] Figure 9 It is a schematic diagram of the structure of the coordinate detection sub-module in the second embodiment of the present invention. Detailed Embodiments

[0030] The following are specific embodiments to illustrate the implementation manners of the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of the present invention. Additionally, the drawings of the present invention are only for simple schematic illustration and are not drawn according to actual dimensions. The following embodiments will further detail the related technical content of the present invention, but the disclosed content is not used to limit the protection scope of the present invention.

[0031] Embodiment 1: This embodiment provides an adaptive gait optimization system for humanoid robots based on deep learning, as combined Figures 1 to 6 as shown.

[0032] An adaptive gait optimization system for humanoid robots based on deep learning, the system includes an analysis module, a control module, and a communication module; the analysis module is used to process information related to step length, slope, friction, and load, and transmit it to the control module; the control module obtains a climbing step length reference index based on the information related to step length, slope, friction, and load, and transmits it to the communication module; the communication module transmits the climbing step length reference index to the robot foot control module; the robot foot control module adjusts the climbing parameters according to the climbing step length reference index.

[0033] Specifically, adjusting the climbing parameters is equivalent to adjusting the step length of the robot, that is, the length of each step of the robot.

[0034] Optionally, the analysis module includes a slope analysis sub-module, a load analysis sub-module, and a data storage sub-module;

[0035] The slope analysis sub-module is used to analyze and obtain the total length of the slope, the total height of the slope, the slope angle detected each time, and the slope surface friction, and transmit it to the control module; the load analysis sub-module is used to analyze and obtain the total load of the robot, and transmit it to the control module; the data storage sub-module is used to store the leg length of the robot, and transmit it to the control module;

[0036] The control module obtains a load adjustment coefficient according to the total load of the robot, obtains the total number of slope detections according to the total length and total height of the slope, obtains the average slope angle according to the slope angle detected each time and the total number of slope detections, obtains a slope adjustment coefficient according to the average slope angle, obtains a step length ratio index according to the leg length of the robot, obtains a flat ground step length reference index according to the step length ratio index and the leg length of the robot, obtains a climbing step length calculation index according to the flat ground step length reference index, slope adjustment coefficient, average slope angle, slope surface friction, load adjustment coefficient, and total load of the robot, and obtains a climbing step length reference index according to the flat ground step length reference index and the climbing step length calculation index.

[0037] Optionally, the slope analysis sub-module includes a visual analysis unit, a slope analysis unit, and a friction analysis unit; the visual analysis unit is configured to analyze and obtain the total slope length and the total slope height, and transmit them to the control module; the slope analysis unit is configured to analyze and obtain the slope angle detected each time, and transmit it to the control module; the friction analysis unit is configured to analyze and obtain the slope surface friction force, and transmit it to the control module.

[0038] Optionally, the visual analysis unit includes a camera, an image pre-processor, a slope identifier, and a slope calculator; the camera is configured to capture an initial image; the image pre-processor uses Gaussian filtering to reduce the noise of the initial image; the slope identifier identifies the boundary of the slope by edge detection; the slope calculator obtains the total slope length and the total slope height based on the boundary of the slope, and transmits them to the control module.

[0039] Optionally, the friction analysis unit includes a force sensor and a friction calculator; the force sensor is configured to detect and obtain the normal force and the tangential friction force; the friction calculator obtains the slope angle detected each time based on the normal force and the tangential friction force, and transmits it to the control module.

[0040] Optionally, the control module calculates a climbing step length reference index, which satisfies the following formula:

[0041] where LT is the climbing step length reference index,

[0042] L pd is the flat ground step length reference index, and L opt is the climbing step length calculation index.

[0043] Optionally, when the control module calculates the climbing step length reference index, it satisfies the following formula:

[0044] L opt = L pd × (1 - k jd × θ + 0.03 × F + k fz × M); L pd = k bc × L leg ;

[0045]

[0046] k fz = 0.01 × e 0.02×M .

[0047] where k jd is the slope adjustment coefficient, θ is the average slope angle, F is the slope surface friction force, and k fzis the load adjustment coefficient, M is the total load of the robot; k bc is the step length proportionality exponent, L leg is the leg length of the robot; N is the total number of slope detections, j n is the slope angle detected at the nth time; cd is the total length of the slope, h is the total height of the slope.

[0048] When the control module calculates the reference index of the climbing step length, refer to the following program code:

[0049]

[0050]

[0051]

[0052] Specifically, when calculating the reference index of the climbing step length, it is for the climbing situation with a single slope, and the case where the slope fluctuates in a wave shape is not considered.

[0053] Transmit the calculated reference index of the climbing step length to the foot control module of the robot. The robot can adjust the climbing step length in a timely manner according to the actual situation, improving the climbing stability and efficiency. The optimized step length helps to maintain the center of gravity stability and reduce the occurrence of tilting or slipping; in addition, optimizing the step length can ensure that the gait of the robot conforms as much as possible to the natural movement mode of the human body or animals, thereby improving the comfort of its movement and extending the service life of the robot.

[0054] The units of the reference index of the climbing step length, the reference index of the flat ground step length, and the calculation index of the climbing step length are all meters.

[0055] The unit of the average slope angle is degrees.

[0056] The unit of the slope surface friction force is Newton.

[0057] The unit of the total load of the robot is kilograms. The total load of the robot refers to the weight of the robot itself plus the total amount of the load carried by the robot.

[0058] The unit of the leg length of the robot is meters;

[0059] The units of the total length of the slope and the total height of the slope are both meters.

[0060] The above units are just an example. Those skilled in the art can set different units according to actual needs when implementing this solution.

[0061] This embodiment solves the problem of poor adaptability of the traditional gait optimization system. The robot can automatically adjust its gait according to external environmental changes such as slope, friction, and load, adapt to the climbing requirements, has strong adaptability, and enhances stability.

[0062] Embodiment 2: This embodiment includes all the content of Embodiment 1 and provides an adaptive gait optimization system for humanoid robots based on deep learning. Combined with Figures 7 to 9 as shown.

[0063] An adaptive gait optimization system for humanoid robots based on deep learning, the system further includes a stability evaluation module; the stability evaluation module is used to evaluate the information of whether the robot has good or poor climbing stability and transmit it to the communication module; the communication module transmits the information of whether the robot has good or poor climbing stability to the robot foot control module; the robot foot control module pauses the climbing operation of the robot according to the information of poor climbing stability of the robot.

[0064] Optionally, the stability evaluation module includes an actual step length detection sub-module, an information storage sub-module, a coordinate detection sub-module, a data calculation sub-module, and a stability judgment sub-module; the actual step length detection sub-module is used to detect and obtain the actual left foot step length of the robot during climbing and the actual right foot step length of the robot during climbing, and transmit them to the data calculation sub-module; the information storage sub-module is used to store the total number of robot components and the weight of each robot component, and transmit them to the data calculation sub-module; the coordinate detection sub-module is used to detect and obtain the coordinates of each component of the robot when standing on flat ground and the coordinates of each component of the robot during climbing, and transmit them to the data calculation sub-module; the data calculation sub-module obtains the center of gravity coordinate factor of the robot during climbing according to the total number of robot components, the weight of each robot component, and the coordinates of each component of the robot during climbing, obtains the center of gravity coordinate factor of the robot when standing on flat ground according to the total number of robot components, the weight of each robot component, and the coordinates of each component of the robot when standing on flat ground, and obtains the stability index of the robot during climbing according to the actual left foot step length of the robot during climbing, the actual right foot step length of the robot during climbing, the center of gravity coordinate factor of the robot when standing on flat ground, and the center of gravity coordinate factor of the robot during climbing, and transmits the stability index of the robot during climbing to the stability judgment sub-module; the stability judgment sub-module obtains the information of whether the robot has good or poor climbing stability according to the stability index of the robot during climbing and transmits it to the communication module.

[0065] Specifically, when the stability judgment sub-module makes a judgment, it refers to the following principle: when the stability index of the robot during climbing is greater than or equal to the selected threshold of the stability index of the robot during climbing, it means that the climbing stability of the robot is poor; when the stability index of the robot during climbing is less than the selected threshold of the stability index of the robot during climbing, it means that the climbing stability of the robot is good; the selected threshold of the stability index of the robot during climbing is set by those skilled in the art.

[0066] Optionally, the actual step detection sub-module includes a trajectory tracking unit and a step calculation unit; the trajectory tracking unit is used to detect the movement trajectories of the left foot and the right foot and obtain movement information; the step calculation unit obtains the actual left foot step length of the robot when climbing a slope and the actual right foot step length of the robot when climbing a slope according to the movement information, and transmits them to the data calculation sub-module.

[0067] Optionally, the coordinate detection sub-module includes a coordinate setting unit, an initial position detection unit, an attitude tracking unit, a climbing position detection unit, and a data transmission unit; the coordinate setting unit is used to set a reference coordinate; the initial position detection unit is used to detect and obtain the coordinates of each component when the robot stands on flat ground; the attitude tracking unit is used to track the attitude change of the robot when climbing a slope; the climbing position detection unit obtains the coordinates of each component of the robot when climbing a slope according to the attitude change of the robot when climbing a slope; the data transmission unit transmits the coordinates of each component when the robot stands on flat ground and the coordinates of each component of the robot when climbing a slope to the data calculation sub-module.

[0068] Optionally, when the data calculation sub-module calculates the stability index of the robot climbing a slope, the following formula is satisfied:

[0069] STAB = |L left -L right | + ||A - B||;

[0070] Where, STAB is the stability index of the robot climbing a slope, L left is the actual left foot step length of the robot when climbing a slope, L right is the actual right foot step length of the robot when climbing a slope, A is the center of gravity coordinate factor of the robot when standing on flat ground, B is the center of gravity coordinate factor of the robot when climbing a slope; I is the total number of robot components, m i is the weight of the i-th component of the robot, rb i is the coordinate of the i-th component of the robot when standing on flat ground, ra i is the coordinate of the i-th component of the robot when climbing a slope.

[0071] When the data calculation sub-module calculates the stability index of the robot climbing a slope, refer to the following program code:

[0072]

[0073]

[0074]

[0075] Specifically, although the corresponding climbing step - length reference index has been set in the first embodiment, during the actual walking of the robot, the step - lengths of the left and right feet may not be exactly the same. Especially during the movement process, there may be slight differences. This difference will affect the gait stability of the robot. In some cases, it may even cause the robot to lose balance or walk unsteadily.

[0076] The smoothness of the robot's walking can be judged by the absolute value of the difference between the actual step - length of the left foot when the robot climbs the slope and the actual step - length of the right foot when the robot climbs the slope. Generally, the difference in the length of the left and right steps of the robot is small. The units of both the actual step - length of the left foot when the robot climbs the slope and the actual step - length of the right foot when the robot climbs the slope are meters. The methods of calculating the actual step - length of the left foot when the robot climbs the slope and the actual step - length of the right foot when the robot climbs the slope are the same. Therefore, the "actual step - length of the left foot when the robot climbs the slope" is used for analysis and explanation. The "step - length" refers to the distance from the moment when one sole touches the ground in each step to the moment when the next sole touches the ground. The "actual step - length of the left foot when the robot climbs the slope" refers to the distance from the moment when the left foot touches the slope surface (the starting position of the left foot) to the moment when the left foot touches the slope surface again.

[0077] When calculating the center - of - gravity coordinate factor of the robot when climbing the slope, the data at a certain moment when the robot climbs the slope is selected. Before calculating the center - of - gravity coordinate factor of the robot when standing on flat ground and the center - of - gravity coordinate factor of the robot when climbing the slope, a reference coordinate needs to be set, and corresponding reference points (which can be the feet of the robot) are set on the reference coordinate.

[0078] The total number of robot components includes all components of the robot and the loaded items. The following is an example (the example does not uniquely limit the robot components). For example, structural components (body, joints, limbs, power components, etc.), fixing components (screws, nuts, etc.), and loaded items (backpacks, tools, etc.).

[0079] The unit of the weight of each robot component is kilograms.

[0080] Before calculating the coordinates of each component of the robot when standing on flat ground and the coordinates of each component of the robot when climbing the slope, corresponding marking points can be marked on the corresponding components, and then the corresponding coordinates can be obtained through visual analysis. And these coordinates are three - dimensional coordinates.

[0081] The above units are just examples. Those skilled in the art can set different units according to actual needs when implementing this solution.

[0082] This embodiment solves the problem that the traditional gait optimization system has a relatively single function. A stability evaluation module is newly added, which can evaluate the stability of the robot during the climbing process in real - time and judge whether its gait is suitable for the current slope, friction, and load conditions.

[0083] The above-disclosed content is only the preferred and feasible embodiment of the present invention, and does not limit the protection scope of the present invention. Therefore, all equivalent technical changes made by using the content of the specification and drawings of the present invention are included in the protection scope of the present invention. In addition, with the development of technology, the elements therein can be updated.

Claims

1. An adaptive gait optimization system for humanoid robots based on deep learning, characterized in that, The system includes an analysis module, a control module, and a communication module; The analysis module is used to process information related to step length, slope, friction, and load, and transmit it to the control module; The control module obtains a climbing step length reference index based on the information related to step length, slope, friction, and load, and transmits it to the communication module; The communication module transmits the climbing step length reference index to the robot foot control module; The robot foot control module adjusts the climbing parameters according to the climbing step length reference index.

2. The adaptive humanoid robot gait optimization system based on deep learning according to claim 1, characterized in that The analysis module includes a slope analysis sub-module, a load analysis sub-module, and a data storage sub-module; The slope analysis sub-module is used to analyze and obtain the total slope length, total slope height, slope angle detected each time, and slope surface friction, and transmit it to the control module; The load analysis sub-module is used to analyze and obtain the total load of the robot, and transmit it to the control module; The data storage sub-module is used to store the leg length of the robot, and transmit it to the control module; The control module obtains a load adjustment coefficient according to the total load of the robot, obtains the total number of slope detections according to the total slope length and total slope height, obtains the average slope angle according to the slope angle detected each time and the total number of slope detections, obtains a slope adjustment coefficient according to the average slope angle, obtains a step length ratio index according to the leg length of the robot, obtains a flat ground step length reference index according to the step length ratio index and the leg length of the robot, obtains a climbing step length calculation index according to the flat ground step length reference index, slope adjustment coefficient, average slope angle, slope surface friction, load adjustment coefficient, and total load of the robot, and obtains a climbing step length reference index according to the flat ground step length reference index and the climbing step length calculation index.

3. The adaptive gait optimization system for a humanoid robot based on deep learning according to claim 2, wherein The slope analysis sub-module includes a visual analysis unit, a slope analysis unit, and a friction analysis unit; The visual analysis unit is used to analyze and obtain the total slope length and total slope height, and transmit it to the control module; The slope analysis unit is used to analyze and obtain the slope angle detected each time, and transmit it to the control module; The friction analysis unit is used to analyze and obtain the slope surface friction, and transmit it to the control module.

4. The adaptive humanoid robot gait optimization system based on deep learning according to claim 3, characterized in that, The visual analysis unit includes a camera, an image pre-processor, a slope recognizer, and a slope calculator; The camera is used to capture an initial image; The image pre-processor uses the Gaussian filtering method to reduce the noise of the initial image; The slope recognizer identifies the boundary of the slope by the edge detection method; The slope calculator obtains the total slope length and total slope height according to the boundary of the slope, and transmits it to the control module.

5. The adaptive humanoid robot gait optimization system based on deep learning according to claim 3, wherein, The friction analysis unit includes a force sensor and a friction calculator; The force sensor is used to detect and obtain the normal force and tangential friction force; The friction calculator obtains the slope angle detected each time according to the normal force and tangential friction force, and transmits it to the control module.

6. The adaptive gait optimization system for a humanoid robot based on deep learning according to claim 2, wherein, The control module calculates the climbing step length reference index, satisfying the following formula: Among them, LT is the reference index of the ramp step size, and L pd is the reference index of the flat step size, and L opt is the calculation index of the ramp step size.

Images