Quadruped robot based on bionic joint driving

Abstract

The quadruped robot based on bionic joint driving comprises a leg structure arranged on a quadruped robot body, the leg structure comprises a thigh leg and a shank leg, the thigh leg and the shank leg are rotationally connected, the thigh leg is rotationally connected with a machine body of the quadruped robot body, and the thigh leg is rotationally connected with the machine body of the quadruped robot body. An anti-toppling piece used for preventing toppling during walking is arranged on the shank leg part in a contained mode, and a locking supporting piece is arranged between the thigh leg part and the shank leg part; the anti-toppling part comprises a telescopic air cylinder, a first storage groove is formed in the shank part, the telescopic air cylinder is rotationally connected into the first storage groove through a first motor, and a supporting part is arranged at the tail end of a telescopic rod of the telescopic air cylinder; the device is simple in structure, convenient to operate and capable of supporting the body weight of the quadruped robot and preventing toppling, and the problem of dynamic balance of the robot in a complex environment is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of quadruped robots, and in particular to a quadruped robot based on bionic joint drive. Background Art

[0002] With the development of artificial intelligence technology, the application fields of robots have gradually expanded, and people's functional requirements for robots have also become higher and higher. Mobile robots can be divided into wheeled robots, tracked robots, and legged robots. Legged robots can be further divided into quadruped robots and biped robots. Due to the extremely high technical difficulty and poor adaptability to the ground of biped robots, quadruped robots have become a research hotspot in mobile robots due to their strong adaptability to complex environments.

[0003] In the current market, the front leg structure of a quadruped robot consists of a thigh and a calf. The thigh and the calf rotate forward and backward respectively to complete the movement of the foot end. Most quadruped robots are bionic dogs. In addition to having a thigh and a calf, the foot end of a dog can also rotate forward and backward independently to complete specific functions. For example, when opening a door, it can hold and press the doorknob, press the elevator button, press the switch, etc. Usually, a motor is set at the top of the foot end or the bottom of the calf to drive the foot end to rotate independently to complete the corresponding specific actions.

[0004] However, the movement of a quadruped robot requires high-frequency swinging of its four limbs. The foot end is located at the end of the swing. In order to increase the swing frequency, the rotational inertia needs to be reduced, that is, the weight of the foot end needs to be reduced, and the center of gravity of a single leg is transferred to the upper end of the front leg. However, the foot end also needs to bear the weight of the entire body. Therefore, a large rotational torque is required for the foot end motor, which conflicts with the requirement of reducing the weight of the foot end. Most existing quadruped robots use rigid joint drive and traditional control methods (such as PD control), and there are the following problems: traditional control methods are difficult to compensate for system uncertainties online, resulting in tracking deviations of the motion trajectory and increasing the risk of tipping over. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a quadruped robot based on bionic joint drive that can support the weight of the quadruped robot body and prevent tipping over, and solve the dynamic balance problem of the robot in complex environments.

[0006] The present invention is implemented by the following method: A quadruped robot based on bionic joint drive, including a leg structure provided on the body of the quadruped robot. The leg structure includes a thigh leg and a calf leg. The thigh leg and the calf leg are rotatably connected. The thigh leg is rotatably connected to the body of the quadruped robot. A tipping prevention member for preventing walking and tipping is accommodated on the calf leg. A locking support member is provided between the thigh leg and the calf leg; The tipping prevention member includes a telescopic cylinder. The calf leg is provided with a first receiving groove. The telescopic cylinder is rotatably connected in the first receiving groove through a first motor. A support member is provided at the end of the telescopic rod of the telescopic cylinder.

[0007] Further, both the rotatable connection setting between the thigh leg and the calf leg and the rotatable connection setting between the thigh leg and the body of the quadruped robot are driven by a second motor.

[0008] Further, the support member includes an anti-tipping block. The lower end of the first receiving groove is connected with a second receiving groove. The first receiving groove and the second receiving groove are communicated. A support block corresponding to the second receiving groove is provided at the end of the telescopic rod of the telescopic cylinder. Grooves are provided on both the left and right side surfaces of the support block. The anti-tipping block is rotatably connected in the grooves through a third motor. An anti-slip layer is provided on the lower surface of the anti-tipping block.

[0009] Further, the locking support member includes a first connecting rod and a second connecting rod. A first guide rail groove is provided on the thigh leg. A second guide rail groove is provided at the upper end of the calf leg. One end of the first connecting rod is rotatably connected by embedding in the first guide rail groove. One end of the second connecting rod is rotatably connected by embedding in the second guide rail groove. A limiting groove is provided at the other end of the first connecting rod. A third connecting rod is provided in the limiting groove. A limiting port is provided at the other end of the second connecting rod. A strip-shaped groove is provided on the bottom surface of the limiting port. A fourth motor is provided in the strip-shaped groove. The output end of the fourth motor is connected with a screw rod. Moving blocks are spirally sleeved on both the left and right ends of the screw rod. A rack is provided on the inner side surface of the moving block. A gear meshing with the rack is provided in the limiting port. The other end of the second connecting rod is embedded in the limiting groove. The third connecting rod is connected through the gear. The left half of the screw rod is a left-handed thread, and the right half of the screw rod is a right-handed thread.

[0010] Further, the tipping prevention member includes an RBF neural network compensation module for online approximating system uncertainties and generating a compensation torque.

[0011] Further, the compensation torque dynamically updates the network weights, centers, and basis width parameters through the gradient descent method.

[0012] The beneficial effects of the present invention are as follows: Through the anti-tipping member, the present invention can integrate the tipping index and adaptive control, significantly reducing the tipping risk under complex terrains; through the locking support member, when the quadruped robot is carrying and transporting items, it can prevent the quadruped robot from being overwhelmed due to the overweight of the items, and can realize the limiting and supporting effect on the quadruped robot. It can be actually applied to the quadruped robot, improving the movement stability of the quadruped robot, coping with situations such as the overweight of the entire body, and enhancing the movement stability; significantly improving the anti-interference ability and movement stability of the quadruped robot in complex environments, providing a reliable technical solution for scenarios such as field exploration and disaster rescue; through the bionic joint design, improving the movement compliance, and combining the RBF neural network to online compensate for uncertainties such as ground reaction forces and load changes, realizing high-precision trajectory tracking and dynamic balance control; effectively reducing the tipping risk under complex environments, and being applicable to scenarios such as field exploration and disaster rescue. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic structural diagram of the first state of the present invention.

[0014] Figure 2 It is a schematic structural diagram of the second state of the present invention.

[0015] Figure 3 It is a schematic structural diagram of the locking support member.

[0016] Figure 4 It is a schematic diagram of the usage state of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The present invention will be further described below with reference to the accompanying drawings.

[0018] Please refer to Figures 1 to 4As shown in the figure, the present invention provides an embodiment: a quadruped robot based on bionic joint drive, including a leg structure 2 provided on the quadruped robot body 1. The leg structure 2 includes a thigh leg 21 and a calf leg 22. The thigh leg 21 and the calf leg 22 are rotatably connected. The thigh leg 21 is rotatably connected to the fuselage of the quadruped robot body 1. A tipping prevention member 3 for preventing walking and tipping is accommodated on the calf leg 22. A locking support member 4 is provided between the thigh leg 21 and the calf leg 22. The tipping prevention member 3 includes a telescopic cylinder 31. A first receiving groove 32 is formed in the calf leg 22. The telescopic cylinder 31 is rotatably connected in the first receiving groove 32 through a first motor. A support member 5 is provided at the end of the telescopic rod of the telescopic cylinder 31. By means of the bionic joint design of the thigh leg 21 and the calf leg 22, the movement compliance is improved. Through the action of the tipping prevention member 3, the anti-interference ability and movement stability in a complex environment can be achieved, preventing the quadruped robot from tipping to the side. Through the action of the locking support member 4, when the quadruped robot carries and transports items, the quadruped robot can be supported, avoiding the quadruped robot being overwhelmed due to the overweight of the items.

[0019] Please continue to refer to Figure 1 and Figure 2 As shown in the figure, in an embodiment of the present invention, the rotational connection between the thigh leg 21 and the calf leg 22 and the rotational connection between the thigh leg 21 and the fuselage of the quadruped robot body 1 are both driven by a second motor. By means of the second motor, the rotation of the thigh leg 21 and the calf leg 22 can be realized, thereby realizing the walking drive function of the quadruped robot body 1.

[0020] Please continue to refer to Figure 1 and Figure 2 As shown in the figure, in an embodiment of the present invention, the support member 5 includes an anti-tipping block 51. A second receiving groove 52 is connected to the lower end in the first receiving groove 32. The first receiving groove 32 and the second receiving groove 52 are connected and communicated. A support block 53 corresponding to the second receiving groove 52 is provided at the end of the telescopic rod of the telescopic cylinder 31. Grooves 54 are formed on both the left and right sides of the support block 53. The anti-tipping block 51 is rotatably connected in the grooves 54 through a third motor. An anti-slip layer 55 is provided on the lower surface of the anti-tipping block 51. When in a complex terrain, the telescopic cylinder 31 rotates out of the first receiving groove 32, the telescopic cylinder 31 drives the support block 53 to extend to the side, and then the third motor can drive the anti-tipping block 51 to unfold, so as to realize the support function for the quadruped robot when the quadruped robot tilts, avoiding the quadruped robot from tipping over.

[0021] Please continue to refer to Figures 1 to 3As shown in the figure, in an embodiment of the present invention, the locking support member 4 includes a first connecting rod 41 and a second connecting rod 42. A first guide rail groove 43 is formed on the thigh leg 21, and a second guide rail groove 44 is formed at the upper end of the calf leg 22. One end of the first connecting rod 41 is embedded and rotatably connected in the first guide rail groove 43, and one end of the second connecting rod 42 is embedded and rotatably connected in the second guide rail groove 44. A limiting groove (not shown) is formed at the other end of the first connecting rod 41, and a third connecting rod 45 is arranged in the limiting groove. A limiting port 46 is formed at the other end of the second connecting rod 42, and a strip-shaped groove 47 is formed at the bottom surface of the limiting port 46. A fourth motor (not shown) is arranged in the strip-shaped groove 47, and the output end of the fourth motor is connected with a screw rod 48. Moving blocks 6 are spirally sleeved on the left and right ends of the screw rod 48. A rack 61 is arranged on the inner side surface of the moving block 6, and a gear 62 meshing with the rack 61 is arranged in the limiting port 46. The other end of the second connecting rod 42 is embedded in the limiting groove and arranged, and the third connecting rod 45 passes through the gear 62 and is connected. The left half part of the screw rod 48 is a left-handed thread, and the right half part of the screw rod 48 is a right-handed thread. By the action of the first connecting rod 41 and the second connecting rod 42, support can be provided when the quadruped robot carries and transports items, avoiding the quadruped robot being overwhelmed due to the overweight of the items. Then, the first connecting rod 41 is rotatably connected in the first guide rail groove 43, and the second connecting rod 42 is rotatably connected in the second guide rail groove 44, so that support can be realized when the quadruped robot walks. When the foot of the calf leg 22 of the quadruped robot lands, the fourth motor drives the screw rod 48 to rotate. The rotation of the screw rod 48 can drive the moving blocks 6 to move left and right, realizing the meshing action of the rack 61 and the gear 62, thereby being able to clamp the gear 62, and thus clamp the third connecting rod 45, so that the locking function can be realized, and the fuselage and items of the quadruped robot body 1 can be supported.

[0022] The anti-tipping member includes an RBF neural network compensation module for online approximating system uncertainties and generating a compensation torque. The compensation torque dynamically updates the network weights, centers, and basis width parameters through the gradient descent method.

[0023] It is realized through the following embodiments:

[0024] Upper layer control: Generate gait planning and expected joint trajectories based on the whole-body dynamics model; Lower layer control: Adopt the computed torque method combined with the RBF neural network compensation strategy to dynamically adjust the joint output torque, specifically including: a. Computed torque method: Generate a reference torque based on the PD control law:

[0025] where e = qd - q is the trajectory tracking error, and are the estimated values of the system inertia matrix and the Coriolis force b. RBF neural network compensation: Design an RBF network to approximate the system uncertainties (such as ground reaction force, load change) online, and output the compensation torque ΔT ： Input layer: joint error e, angular velocity error , and ground contact force information; Hidden layer: Gaussian basis function , dynamically adjust the center and the basis width ; Output layer: compensation torque , weights are updated online by the gradient descent method, and the objective function is:

[0026] where ΔU is the deviation between the actual system uncertainty and the estimated value; c. Final control law:

[0027] Anti-tipping strategy Real-time monitor the centroid position and attitude angle of the robot through the inertial measurement unit (IMU), and calculate the tipping risk index ( θ is the pitch / roll angle); When S exceeds the threshold, trigger the gait adjustment mechanism, and redistribute the foot-end contact force in combination with the RBF compensation torque to achieve dynamic balance.

[0028] Install joint encoders at the joints of the quadruped robot: resolution 0.01°, and real-time feedback of joint angle q and angular velocity . Six-axis force sensor: installed at the foot end to measure the ground reaction force (range ±500 N, accuracy ±1%).

[0029] IMU module: integrated with a three-axis accelerometer, gyroscope and magnetometer, sampling frequency 100 Hz, and real-time output of the centroid position (x, y, z) and attitude angle of the robot ( ).

[0030] The main control unit uses an embedded real-time controller (such as NVIDIA Jetson AGX Xavier), runs the ROS 2 system, and the control period is 1 ms; The RBF neural network calculation module is implemented by FPGA to achieve parallel acceleration and ensure the real-time performance of online compensation.

[0031] Implementation of RBF neural network compensation control Network structure and parameter initialization Input layer: 7-dimensional vector , where is the joint trajectory error, is the foot-end contact force; Hidden layer: 20 Gaussian nodes, initial centers evenly distributed in the input space (such as ), basis width ; Output layer: 3-dimensional compensation torque Initial weights randomly distributed in [-0.1, 0.1].

[0032] Online learning algorithm Weight update:

[0033] where the learning rate η = 0.35 and the momentum factor α = 0.05, is the system dynamic error; Basis width and center adjustment:

[0034] Updated every 10 ms to ensure that the network can quickly adapt to environmental changes.

[0035] Control flow Step 1: Generate the desired joint trajectory through the full-body dynamics model (such as the Lagrangian equation)

[0036] Step 2: Collect real-time data

[0037] Step 3: Calculate the reference torque , where

[0038] Step 4: The RBF network outputs the compensation torque , generating the final control instruction

[0039] Step 5: Drive the bionic joint to execute torque output and update the network parameters.

[0040] Define the tipping index , where

[0041] When , trigger the anti-tipping mode.

[0042] Dynamic adjustment mechanism Gait reconstruction: Shorten the stride length to 50% and reduce the center-of-mass height; Moment redistribution: Increase the joint moment of the supporting leg through RBF compensation (e.g., the hip joint moment is increased by 20%) and reduce the output of the swinging leg; Ground adaptation: Adjust the foot landing angle according to F the ground feedback to ensure uniform distribution of the contact force.

[0043] Experimental verification Scenario 1: On a slope terrain (with an inclination angle of 15°), when the RBF compensation is not enabled, the robot overturns within 5 seconds; after enabling, it walks stably for 10 minutes and the trajectory tracking error is reduced to ±0.05 rad; Scenario 2: Random perturbation (lateral impact force of 50 N), the RBF network completes the compensation within 0.2 seconds, and the overturning index S drops from 3.1 to 1.8.

[0044] The above are only the preferred embodiments of the present invention. All equivalent changes and modifications made according to the scope of the patent application of the present invention shall fall within the scope covered by the present invention.

Claims

1. A quadruped robot based on bionic joint drive, characterized in that: It includes a leg structure provided on the quadruped robot body. The leg structure includes a thigh leg and a calf leg. The thigh leg and the calf leg are rotatably connected. The thigh leg is rotatably connected to the fuselage of the quadruped robot body. A tipping prevention member for preventing walking and tipping is accommodated on the calf leg. A locking support member is provided between the thigh leg and the calf leg. The tipping prevention member includes a telescopic cylinder. A first receiving groove is formed in the calf leg. The telescopic cylinder is rotatably connected in the first receiving groove by a first motor. A support member is provided at the end of the telescopic rod of the telescopic cylinder.

2. The quadruped robot based on bionic joint drive according to claim 1, characterized in that: Both the rotatable connection between the thigh leg and the calf leg and the rotatable connection between the thigh leg and the fuselage of the quadruped robot body are driven by a second motor.

3. The quadruped robot based on bionic joint drive according to claim 1, wherein: The support member includes an anti-tipping block. A second receiving groove is connected to the lower end in the first receiving groove. The first receiving groove and the second receiving groove are communicated. A support block corresponding to the second receiving groove is provided at the end of the telescopic rod of the telescopic cylinder. Grooves are formed on the left and right sides of the support block. The anti-tipping block is rotatably connected in the grooves by a third motor. An anti-slip layer is provided on the lower surface of the anti-tipping block.

4. A quadruped robot based on bionic joint drive according to claim 1, characterized in that: The locking support member includes a first connecting rod and a second connecting rod. A first guide rail groove is formed on the thigh leg. A second guide rail groove is formed at the upper end of the calf leg. One end of the first connecting rod is rotatably connected by being embedded in the first guide rail groove. One end of the second connecting rod is rotatably connected by being embedded in the second guide rail groove. A limiting groove is formed at the other end of the first connecting rod. A third connecting rod is provided in the limiting groove. A limiting port is formed at the other end of the second connecting rod. A strip-shaped groove is formed on the bottom surface of the limiting port. A fourth motor is provided in the strip-shaped groove. The output end of the fourth motor is connected to a screw rod. Moving blocks are spirally sleeved on both the left and right ends of the screw rod. A rack is provided on the inner side surface of the moving block. A gear meshing with the rack is provided in the limiting port. The other end of the second connecting rod is embedded in the limiting groove. The third connecting rod is connected through the gear. The left half of the screw rod has a left-handed thread, and the right half of the screw rod has a right-handed thread.

5. The quadruped robot based on bionic joint drive according to claim 1, characterized in that: The tipping prevention member includes an RBF neural network compensation module for online approximating system uncertainties and generating a compensation torque.

6. The quadruped robot based on bionic joint drive according to claim 5, wherein: The compensation torque dynamically updates the network weights, centers, and basis width parameters by the gradient descent method.

Images