A control method for a quadruped robot crossing concave-convex terrain based on multi-sensor fusion

Abstract

The application provides a kind of control method of four-legged robot crossing concave-convex terrain based on multi-sensor fusion, four-legged robot fuses ranging sensor, encoder, pressure sensor, inertial measurement unit and other sensors, and judges the motion state of itself to complete the expected control action, and cross concave-convex terrain.Firstly, the robot judges the terrain according to the ranging sensor, and controls the motion according to the terrain.When detecting the concave terrain, the motion state of itself is detected by combining the encoder, inertial measurement unit and foot pressure sensor, and the vestibular reflex and the take-off launch are triggered according to the set condition to help the robot leg quickly contact the ground.When detecting the convex terrain, the motion state of itself is also detected by combining the above sensors, and the encounter obstacle reflex and the front and rear leg coordination driving reflex are triggered according to the set condition to adjust the motion posture and help the robot smoothly cross the convex terrain.

Description

Technical Field

[0001] This invention relates to the field of quadruped robot technology, and in particular to a control method for a quadruped robot that traverses uneven terrain based on multi-sensor fusion. Background Technology

[0002] With the development of technology, quadruped robots have found more applications in real life. Compared with wheeled and tracked robots, quadruped robots have better mobility and better traversal capabilities on irregular terrain. Using a traditional central pattern generator (CPG), quadruped robots can have good rhythmic movement. CPG-controlled robots have good stability and traversal capabilities on flat terrain; however, traditional open-loop CPG networks lack self-adjustment for the robot. On irregular terrain, especially uneven terrain, the robot's interaction with the environment is poor, easily leading to dangers such as slippage, jamming, violent vibrations, and even tipping over.

[0003] Therefore, fusing information from multiple sensors, determining the terrain type based on the sensor information, and obtaining the quadruped robot's own motion state can improve the robot's ability to interact with the environment and its dynamic adjustment capabilities.

[0004] The method of this invention incorporates multi-condition judgment and multiple feedback terms into the CPG control strategy to adjust the joint angles and motion posture of the quadruped robot, adapting to uneven terrain. This effectively reduces phenomena such as slippage, jamming, and violent shaking of the robot on uneven terrain, resulting in better control performance and helping the robot adapt to irregular terrain. Summary of the Invention

[0005] To address the problems of poor adaptability to uneven terrain and insufficient interaction with the environment in existing quadruped robots under CPG control methods, this invention provides a quadruped robot control method based on multi-sensor fusion to traverse uneven terrain.

[0006] This invention provides a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion, comprising:

[0007] S1: First, the quadruped robot uses the rhythmic signals output by the Multi-sensor information fusion CPG to plan its gait. Then, through three ranging sensors and an inertial measurement unit, the robot judges and processes the distance values ​​obtained by the ranging sensors from the ground directly below and from the ground diagonally in front of the robot body to identify the terrain type in front of the quadruped robot, including concave terrain, convex terrain, or flat terrain.

[0008] S2: When the terrain is determined to be concave in S1, the quadruped robot simultaneously activates the vestibular reflex and step-off reflex algorithms to ensure that the robot's body does not fluctuate greatly when its feet land, and to adjust its body posture after landing to ensure that the body does not tilt too much, thus preventing excessive shift of the center of gravity and uneven force distribution on the four legs caused by excessive forward and backward tilt angles.

[0009] S3: When the terrain is determined to be raised as described in S1, the quadruped robot assesses the height range of the raised terrain. If the height is less than the robot's leg-lifting height, the robot can directly cross it, requiring only vestibular feedback from the inertial measurement unit for posture adjustment. If the height is greater than the robot's leg-lifting height, the robot needs to activate obstacle detection reflex and hind leg drive reflex to assist in crossing the raised terrain. After crossing the raised terrain with its front legs, the robot adjusts its posture, lowers its center of gravity, and then attempts to cross with its hind legs. The hind leg crossing activates obstacle detection reflex and coordinates with the front leg drive reflex, thus crossing the raised terrain.

[0010] S4: When the terrain is determined to be flat in S1, the quadruped robot walks according to the gait planned by the rhythm signal output by the central pattern generator (CPG).

[0011] According to the present invention, a quadruped robot control method based on multi-sensor fusion for traversing uneven terrain is provided. The quadruped robot uses multi-sensor information fused with actual terrain data to perform feedback adjustment. The feedback quantity is added to the output signal of the central pattern generator (CPG) for actual motion adjustment. Its mathematical model is as follows:

[0012]

[0013] In the formula, X i Y i x represents the actual hip and knee joint angle input. hi y ki The offset represents the hip and knee joint angle signals output by the CPG oscillator. hi offset ki The vestibular reflex feedback values ​​for the hip and knee joints. hi feedback ki Feedback from the hip and knee joints, used for the robot's reaction to missteps or obstacles, is used to control its interaction with the outside world based on terrain changes, thereby achieving dynamic feedback to adapt to the terrain.

[0014] According to the present invention, a control method for a quadruped robot that traverses uneven terrain based on multi-sensor fusion is provided. The three ranging sensors are respectively installed on the head of the robot body, the bottom of the front body, and the bottom of the rear body. The bottom ranging sensor is used to measure the height of the robot body from the ground, and the head ranging sensor is used to measure the distance between the robot and a set angle diagonally in front of the robot.

[0015] According to the present invention, a control method for a quadruped robot that traverses uneven terrain based on multi-sensor fusion is provided, wherein the ranging sensors include, but are not limited to, ultrasonic sensors, infrared sensors, laser sensors, etc.

[0016] According to the present invention, a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion is provided, wherein the distance processing and calculation method is as follows:

[0017]

[0018] Where DS2 is the distance value obtained by the bottom ranging sensor of the quadruped robot, l is the length of a single leg of the quadruped robot, θ represents the initial angle of the quadruped robot joint, threshold2 is the height threshold set according to the actual situation, and max_height is the maximum height difference set according to the robot's own size. The depressions and protrusions defined here are within the robot's crossing capability.

[0019] In this calculation method, the distance value obtained by the sensor is compared with the distance from the bottom of the robot's body to the soles of its feet. If the distance value obtained by the sensor is greater than the distance from the bottom of the robot's body to the soles of its feet, and the difference is greater than a set threshold, it is determined to be a concave terrain. Similarly, if the distance value obtained by the sensor is less than the distance from the bottom of the robot's body to the soles of its feet, and the difference is less than the set threshold, it is determined to be a convex terrain. When the difference between the distance value obtained by the sensor and the distance from the bottom of the robot's body to the soles of its feet is within the set threshold range, it is determined to be a flat terrain. The size of the set threshold is based on the range of fluctuation of the robot in flat terrain to prevent the robot from misjudging the terrain.

[0020] According to the present invention, a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion is provided, further determining the distance processing method.

[0021]

[0022] If the robot determines that DS² - 2lcosθ ≥ max_height, then the terrain is too deep a depression and cannot be crossed, requiring a change of route. If the robot determines that DS² - 2lcosθ ≤ -max_height, then the terrain is too high a protrusion and cannot be crossed, requiring a change of route.

[0023] According to the present invention, a quadruped robot control method based on multi-sensor fusion for traversing uneven terrain is provided. When the terrain is determined to be uneven, the quadruped robot simultaneously activates the vestibular reflex and the slip reflex, thereby enabling simultaneous posture adjustment and rapid foot landing, which helps to reduce the severe vibration and body imbalance caused by slipping.

[0024] This invention provides a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion. The vestibular reflex activated by the quadruped robot in uneven terrain mimics a biological mechanism for maintaining balance during movement. This vestibular reflex is applied to the dynamic posture adjustment of the quadruped robot in uneven terrain. The control method is as follows:

[0025] Vestibular reflex calculation method

[0026]

[0027]

[0028] offset h The offset is the offset of the robot's hip joint balance position. k α represents the offset of the knee joint balance position of the quadruped robot. α represents the terrain slope. Since the body posture angle is approximately linearly related to the slope, the posture angle balance coefficient h1 can be obtained. Then, by mapping the posture angle to the joint angle change, the joint angle balance coefficient h2 can be obtained.

[0029] This invention provides a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion, which includes a step-off reflex activated when the quadruped robot is in a concave terrain. When the DS2 (No. 2) ranging sensor detects that the terrain in front of the robot is concave, the method combines foot pressure sensors and an encoder to determine whether the robot's feet touch the ground when entering a support phase. If the feet do not touch the ground, the step-off reflex is activated to prevent severe vibrations and imbalances caused by stepping off the ground.

[0030] According to the present invention, a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion is provided. The method combining foot pressure sensors and encoders involves acquiring feedback values ​​from the foot pressure sensors simultaneously when the encoder detects that the quadruped robot's hip and knee joints have reached the angle for entering the support phase. If the foot pressure sensor feedback indicates no ground contact, it is marked as a misstep; if the feedback indicates ground contact, it is marked as a normal footing.

[0031] Its judgment method can be expressed as:

[0032]

[0033] Where i represents the i-th leg, x hi y represents the hip angle of the i-th leg. ki θ represents the knee angle of the i-th leg. sth The joint angle θ represents the phase transition of the hip joint. stk The joint angle Δθ represents the phase transition of the knee joint. h This indicates setting a threshold for the range of motion of the hip joint, Δθ k This indicates setting the threshold for the range of motion of the knee joint, touchsenor i This represents the feedback value of the pressure sensor at the foot end of the i-th leg. A value of 0 indicates that the leg is not touching the ground, and a value of 1 indicates that the leg is touching the ground.

[0034] According to the present invention, a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion is provided. The control method for the "stepping reflex" involves providing feedback values ​​to the knee and hip joints of the robot's leg when the swinging leg enters the support phase but has not yet touched the ground, causing it to quickly extend downwards until it touches the ground. Simultaneously, at the moment of stepping, the robot's center of gravity is lowered in real time to prevent tipping due to large height differences. The mathematical model for stepping reflex used in this invention is as follows:

[0035]

[0036] In the formula, feedback hi feedback ki The feedback values ​​from the hip and knee joints in the robot's misstep reaction are denoted by n, where n is the amplitude adjustment coefficient, and x is the value of the change. hi Let x be the hip angle value of the i-th leg. s =θ sth -Δθ h x s This represents the lower limit of the range of motion of the hip joint.

[0037] According to the present invention, a quadruped robot control method based on multi-sensor fusion for crossing uneven terrain is provided. When it is determined that the height of the uneven terrain is less than the current leg lifting height, the robot can directly cross the uneven terrain, only needing to adjust the vestibular reflex posture during the crossing process.

[0038] According to the present invention, a quadruped robot control method based on multi-sensor fusion for traversing uneven terrain is provided. When it is determined that the height of the uneven terrain is greater than the current leg lifting height, the quadruped robot will simultaneously activate obstacle detection reflex and hind leg drive.

[0039] According to the present invention, a quadruped robot control method based on multi-sensor fusion for traversing uneven terrain is provided. When it is determined that the height of the uneven terrain is greater than the current leg lifting height, the quadruped robot initiates an obstacle encounter reflex action as follows:

[0040] Step 1: First, swing the hip joint of the front leg backward, reducing the angle to A1, while simultaneously swinging the knee joint upward, increasing the angle to B1, to complete the contraction and raising of the leg movement.

[0041] Step 2: Swing your front leg hip forward, increasing the angle to A2, while keeping the knee joint unchanged, to complete the obstacle course.

[0042] Step 3: Keeping the hip joint angle of the front leg unchanged, swing the knee joint down, reducing the angle to B3, and complete the landing action;

[0043] Step 4: Continue to swing the hip and knee joints to the angle planned by CPG.

[0044] The values ​​of A1, A2, B1, and B3 are determined by the robot's own dimensions and the height of the raised terrain.

[0045] According to the present invention, a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion is provided. The obstacle encounter and reaction implementation method planned by the quadruped robot, combined with the angle planned by the CPG, can be calculated as follows:

[0046]

[0047] In the formula, A h A is the hip joint obstacle reflex accommodation coefficient. k x is the knee joint obstacle encounter reflex accommodation coefficient. max x represents the maximum range of motion of the hip joint. hi Output hip joint rotation angle to CPG, y ki Output the knee joint angle to the CPG. A h A k A is an adjustable parameter used to control the magnitude of change in feedback adjustment. h =0.3, A k =0.3.

[0048] According to the present invention, a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion is provided. The condition for initiating obstacle detection reflex is that the ranging sensor determines that there is a raised terrain ahead. The front leg initiates obstacle detection reflex when the foot pressure sensor determines that it has entered the swing phase.

[0049] According to the present invention, a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion is provided. When the terrain is determined to be uneven with a height greater than the current leg lifting height, the quadruped robot initiates a hind leg drive reflex method. This method involves the hind legs retracting simultaneously with the front legs' obstacle detection reflex, maintaining the robot's body with the front higher than the rear. As the robot's front legs complete the landing action, the hind legs quickly return to their original posture, generating a forward thrust that helps the robot stably traverse uneven terrain.

[0050] According to the present invention, a quadruped robot control method based on multi-sensor fusion for traversing uneven terrain is provided. When it is determined that the height of the uneven terrain is greater than the current leg lifting height and the front leg traverses the uneven terrain, the posture is adjusted, and the rear leg initiates obstacle detection reflex and the front leg coordinates the drive.

[0051] According to the present invention, a control method for a quadruped robot that traverses uneven terrain based on multi-sensor fusion is provided, wherein the obstacle encounter reflex control method of the hind legs is the same as that of the front legs.

[0052] According to the present invention, a control method for a quadruped robot based on multi-sensor fusion to traverse uneven terrain is provided. The method involves coordinated front leg drive, where the front legs retract to lower the robot's forebody posture while maintaining the original forward movement control mode. This allows the hindbody to be propelled forward by the forebody, reducing the load on the hind legs. This helps the hind legs traverse uneven terrain.

[0053] According to the present invention, a control method for a quadruped robot based on multi-sensor fusion to traverse uneven terrain is provided, wherein the hind legs perform obstacle detection reflex when the foot pressure sensor determines that the robot is entering the swing phase.

[0054] According to the present invention, a quadruped robot control method based on multi-sensor fusion for traversing uneven terrain is provided. When the terrain is determined to be flat, the quadruped robot moves forward according to the original CPG motion plan.

[0055] The beneficial effects of this invention are as follows:

[0056] 1) The ranging sensor used in this invention can enhance the robot's perception of the terrain by judging the current terrain, thereby making preparations in advance to cross uneven terrain, and has a better perception capability.

[0057] 2) This invention integrates a ranging sensor, a foot pressure sensor, an inertial measurement unit, and an encoder to provide the quadruped robot with external terrain information and its own motion information for traversing uneven terrain. The self-motion information includes the robot's current posture angle, foot contact with the ground, joint angles, etc., which provides powerful assistance for the robot to perform various reflex movements and posture adjustments.

[0058] 3) The present invention simultaneously activates vestibular feedback and step reflex in concave terrain, allowing the quadruped robot to adjust its posture while its legs are rapidly extending downwards. This improves the stability, passability, and autonomous adjustment capabilities of the quadruped robot, making it more robust in changing environments.

[0059] 4) This invention simultaneously activates obstacle detection reflexes and hind leg coordinated drive when encountering raised terrain, helping the robot quickly traverse such terrain and reducing time and energy consumption. The hind leg drive method allows the robot to cross raised terrain even with lower torque, making it more adaptable to its environment.

[0060] 5) The present invention simultaneously activates the vestibular reflex and coordinates the foreleg drive when encountering raised terrain, which helps the robot's hind legs to quickly cross raised terrain. By lowering the center of gravity of the forebody and moving the center of gravity forward, it helps the hind legs to better cross raised terrain.

[0061] 6) In this invention, the robot combines multiple reflection methods through information from various sensors, which makes it more stable and robust when crossing uneven terrain. Attached Figure Description

[0062] Figure 1 This is a flowchart of the steps of the control method for a quadruped robot that traverses uneven terrain based on multi-sensor fusion, as described in this invention.

[0063] Figure 2 This is a Simulink control model diagram in an embodiment of the present invention;

[0064] Figure 3 This is a schematic diagram of the quadruped robot structure and sensor installation positions on the Webots platform in this embodiment of the invention. Figure 1 ;

[0065] Figure 4 This is a schematic diagram of the quadruped robot structure and sensor installation positions on the Webots platform in this embodiment of the invention. Figure 2 ;

[0066] Figure 5 This is the timing diagram of the robot's walking phase in this invention;

[0067] Figure 6 This is a schematic diagram of a concave terrain model (left) and a convex terrain model (right) on the Webots platform in an embodiment of the present invention;

[0068] Figure 7 This is a schematic diagram of the terrain detection using the ranging sensor of the quadruped robot according to the present invention;

[0069] Figure 8 This is a diagram showing the relationship between sensor signals and reflected motion under concave terrain, as described in this invention.

[0070] Figure 9 This is a schematic diagram of the quadruped robot's reaction to stepping into concave terrain and the change in leg movements of the forelegs.

[0071] Figure 10 This is a posture diagram of a robot on the Webots platform in this embodiment of the invention, which activates vestibular reflex and slip reflex when encountering concave terrain.

[0072] Figure 11This is a waveform diagram of the joint angles of the robot on the Webots platform in this embodiment of the invention when it encounters a concave terrain and initiates the vestibular reflex and the slip reflex.

[0073] Figure 12 This is a diagram showing the relationship between sensor signals and reflected motion of the quadruped robot under raised terrain.

[0074] Figure 13 This is a schematic diagram of the quadruped robot of the present invention reacting to obstacles on raised terrain and changing leg movements.

[0075] Figure 14 This is a diagram showing the front leg activation obstacle encounter reflex and hind leg driving posture of the robot on the Webots platform when encountering raised terrain in this embodiment of the invention.

[0076] Figure 15 This is a waveform diagram of the joint angles of the front legs initiating obstacle encounter reflex and the rear legs driving when the robot on the Webots platform encounters a raised terrain in an embodiment of the present invention.

[0077] Figure 16 This is a diagram showing the hind leg obstacle avoidance reflex and front leg coordinated driving posture of the robot on the Webots platform when encountering raised terrain in this embodiment of the invention.

[0078] Figure 17 This is a waveform diagram of the joint angles of the robot on the Webots platform encountering a raised terrain in this embodiment of the invention, showing the obstacle encounter reflex initiated by the hind legs and the coordinated drive reflex of the front legs.

[0079] Figure 18 This is a Simulink control model diagram of the feedback adjustment part of the quadruped robot in an embodiment of the present invention. Detailed Implementation

[0080] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0081] like Figure 1 This is a flowchart illustrating a control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion, provided by the present invention, including the following steps:

[0082] S1: First, the quadruped robot uses the rhythmic signals output by the Multi-sensor information fusion CPG to plan its gait. Then, through three range sensors and an inertial measurement unit, the robot's body distance from the ground directly below and from the ground diagonally in front is judged and processed based on the robot's own size and the distance values ​​obtained by the distance sensors to the ground. This allows the robot to distinguish the terrain type in front of it, including concave terrain, convex terrain, or flat terrain.

[0083] S2: When the terrain is determined to be concave in S1, the quadruped robot simultaneously activates the vestibular reflex and step-off reflex algorithms to ensure that the robot's body does not fluctuate greatly when its feet land, and to adjust its body posture after landing to ensure that the body does not tilt too much, thus preventing excessive shift of the center of gravity and uneven force distribution on the four legs caused by excessive forward and backward tilt angles.

[0084] S3: When the terrain is determined to be raised in S1, the quadruped robot assesses the height range of the raised terrain. If the height of the raised terrain is less than the height at which the quadruped robot lifts its legs, it can directly cross it, only needing to adjust its posture using vestibular reflexes via the inertial measurement unit. If the height of the raised terrain is greater than the height at which the quadruped robot lifts its legs, the quadruped robot needs to activate obstacle detection reflexes and hind leg drive reflexes to assist in crossing the raised terrain. After the front legs cross the raised terrain, the robot adjusts its posture to lower its center of gravity before attempting to cross with its hind legs. The hind leg crossing activates obstacle detection reflexes and coordinates with the front legs' drive reflexes to cross the raised terrain.

[0085] S4: When the terrain is determined to be flat in S1, the quadruped robot walks according to the gait planned by the rhythm signal output by the central pattern generator (CPG).

[0086] In one embodiment of the present invention, a co-simulation platform combining Webots and Simulink is used. A quadruped robot structural model, a terrain model, and various sensors are built in Webots to provide a virtual environment for algorithm verification. Webots and Simulink can establish co-simulation via API. Webots provides the physics engine, executes instructions, and generates motion data. Simulink receives the data generated in Webots, performs data calculations for the motion control algorithm, and provides the calculation results to the Webots platform. Figure 2 This is a Simulink control model diagram established in the example of the present invention. The inputs are the relevant gait and control parameters, and the output signals are the angles of each joint.

[0087] First, a quadruped robot structural model was built on the Webots platform, and various sensors were installed on the robot's body, such as... Figure 3 and 4 As shown in the illustration, this invention employs a quadruped robot construction method with thirteen degrees of freedom. Each leg has three degrees of freedom: lateral hip joint, hip joint, and knee joint. In addition, there is one degree of freedom at the junction of the forequarters and posterior trunk with the spine. Adding this degree of freedom allows the quadruped robot to have more posture variations and better adaptability to complex terrain. The actual structure of the quadruped robot may differ, and the sensor mounting method may vary, but it must meet the requirements for the control algorithm to operate.

[0088] Furthermore, a CPG mathematical model is built in Simulink to output rhythmic joint angle values. This section introduces the CPG mathematical model as a basis for use and does not constitute a limitation on the invention. The CPG mathematical model used in this invention is as follows:

[0089]

[0090] r i 2 =x hi 2 +y ki 2 (1.2)

[0091]

[0092]

[0093] In the formula, x hi It is the signal quantity of the oscillator output hip joint angle, y ki It is the knee joint angle signal output by the oscillator; (1.1) the second term on the right is the coupling term between oscillators; θ i j This refers to the relative phase between oscillators i and j; R(θ) i j ) is a rotation matrix used to represent the phase coupling relationship between oscillators; ω sw ω st These are the swing phase frequency and the support phase frequency of the quadruped robot, respectively; parameter a determines ω in ω sw and ω st The rate of change between them is a positive constant; β is the loading factor.

[0094] Furthermore, this invention requires the fusion of multi-sensor information based on the actual terrain to provide feedback adjustment for the robot. The feedback quantity is adjusted by adding it to the output signal, and its mathematical model is as follows:

[0095]

[0096] In the formula, X i Y i The actual hip and knee joint angle input, offset hi offset ki The vestibular reflex feedback values ​​for the hip and knee joints. hi feedback ki Feedback from the hip and knee joints, used for the robot's reaction to missteps or obstacles, is used to control its interaction with the outside world based on terrain changes, thereby achieving dynamic feedback to adapt to the terrain.

[0097] When traversing uneven terrain, the quadruped robot primarily employs a four-beat Walk gait, which offers better stability and maneuverability on uneven surfaces. The phase flow diagram is shown below. Figure 5 As shown in the diagram. LF represents the left front leg, LH represents the left hind leg, RF represents the right front leg, and RH represents the right hind leg. On flat terrain, the quadruped robot primarily uses the gait sequence LF->RH->LH->RF. On uneven terrain, it primarily uses the gait sequence LF->RF->LH->RH. The gait sequence used by the quadruped robot on uneven terrain takes into account the impact of the terrain on stability. Ensuring that the left and right front legs traverse the uneven terrain first, followed by the hind legs crossing the terrain, makes the robot more stable and easier to control.

[0098] In one embodiment, a simulation model is established in the Webots simulation model as follows: Figure 6 The depressions and ridges shown have a height difference of 5 cm.

[0099] like Figure 3 and 4 As shown, three distance sensors are installed at the robot's head, the bottom of the front body, and the bottom of the rear body, respectively. The head distance sensor measures the distance between the robot and a set angle in front of it, mainly for early terrain perception. In this example, the set angle is 45 degrees, and it measures the terrain elevation 18.8 cm in front of the robot. The distance sensors at the bottom of the front and rear bodies measure the height of the robot's front and rear bodies from the ground.

[0100] Furthermore, the ranging sensor can be an ultrasonic sensor, an infrared sensor, a laser sensor, etc. In this embodiment of the invention, an infrared ranging sensor is used. The DS1 infrared ranging sensor is mounted on the head of the quadruped robot, the DS2 infrared ranging sensor is mounted on the bottom of the front body, perpendicular to and facing the ground, and the DS3 infrared ranging sensor is mounted on the bottom of the rear body, perpendicular to and facing the ground, used to measure the distance from the bottom of the robot's body to the ground. Figure 7The diagram shown is a schematic diagram of terrain detection using a ranging sensor on a quadruped robot in an embodiment of the present invention.

[0101] After the robot completes the distance measurement, it proceeds to the judgment stage, processing the obtained distance data to determine the terrain type and detailed parameters ahead. The distance processing and calculation method is as follows:

[0102]

[0103] Where DS2 is the distance value obtained by the bottom ranging sensor of the quadruped robot, l is the length of a single leg of the quadruped robot, θ represents the initial angle of the quadruped robot joint, threshold2 is the height threshold set according to the actual situation, and max_height is the maximum height difference set according to the robot's own size. The depressions and protrusions defined here are within the robot's crossing capability.

[0104] In this calculation method, the distance value obtained by the sensor is compared with the distance from the bottom of the robot's body to the soles of its feet. If the distance value obtained by the sensor is greater than the distance from the bottom of the robot's body to the soles of its feet, and the difference is greater than a set threshold, it is determined to be a concave terrain. Similarly, if the distance value obtained by the sensor is less than the distance from the bottom of the robot's body to the soles of its feet, and the difference is less than the set threshold, it is determined to be a convex terrain. When the difference between the distance value obtained by the sensor and the distance from the bottom of the robot's body to the soles of its feet is within the set threshold range, it is determined to be a flat terrain. The size of the set threshold is based on the range of fluctuation of the robot in flat terrain to prevent the robot from misjudging the terrain.

[0105] Furthermore,

[0106]

[0107] If the robot determines that DS² - 2lcosθ ≥ max_height, then the terrain is too deep a depression and cannot be crossed, requiring a change of route. If the robot determines that DS² - 2lcosθ ≤ -max_height, then the terrain is too high a protrusion and cannot be crossed, requiring a change of route.

[0108] When the terrain is determined to be concave, the quadruped robot simultaneously activates its vestibular reflex and slip reflex, enabling it to adjust its posture and land quickly at the same time, thus helping to reduce the severe vibration and body imbalance caused by slipping.

[0109] The vestibular reflex activated by a quadruped robot in concave terrain mimics a biological mechanism for maintaining balance during movement and is commonly used for posture adjustment on slopes. After adjustment, the quadruped robot's body posture angle and slope will have an approximately linear relationship, i.e., Δα = h1α, where the posture angle balance coefficient h1 = 0.24, Δα is the body posture angle (here referring to the robot's pitch angle, which can be directly read through the inertial measurement unit), and α is the terrain slope. The relationship between the change in the robot's joint balance position Δθ and the body posture angle Δα is then established.

[0110]

[0111] Where l is the length of the robot's leg, L is the length of the robot's body, and θ0 is the original equilibrium position of the joint. In one embodiment of the invention, the lengths of the robot's thigh and lower leg are approximately l = 0.13m, and the body length L = 0.4m. Calculations show that Δθ and Δα have an approximately linear relationship.

[0112] Δθ=h2Δα(1.9)

[0113] The joint angle balance coefficient h2 was found to be 4.30.

[0114] The vestibular reflex described here is applied to the dynamic posture adjustment of a quadruped robot in uneven terrain. The control method is as follows:

[0115] Vestibular reflex calculation method

[0116]

[0117]

[0118] offset h The offset is the offset of the robot's hip joint balance position. k α represents the offset of the quadruped robot's knee joint balance position, and α represents the terrain slope.

[0119] The quadruped robot activates a slip-reflex mechanism in concave terrain. Based on the DS2 (No. 2) ranging sensor detecting a concave terrain in front of the robot, and combined with foot pressure sensors and encoders, it determines whether the robot's feet touch the ground when entering the support phase. If not, the slip-reflex mechanism is activated to prevent severe vibrations and imbalance caused by slipping.

[0120] The method combining foot pressure sensors and encoders is as follows: when the encoder detects that the quadruped robot's hip and knee joints have reached the angle to enter the support phase, it simultaneously acquires the feedback value from the foot pressure sensors. If the foot pressure sensor feedback indicates no ground contact, it is marked as a misstep; if the feedback indicates ground contact, it is marked as a normal foot landing.

[0121] Its judgment method can be expressed as:

[0122]

[0123] Where i represents the i-th leg, x hi y represents the hip angle of the i-th leg. ki θ represents the knee angle of the i-th leg. sth The joint angle θ represents the phase transition of the hip joint. stk The joint angle Δθ represents the phase transition of the knee joint. h This indicates setting a threshold for the range of motion of the hip joint, Δθ k This indicates setting the threshold for the range of motion of the knee joint, touchsenor i This represents the feedback value of the pressure sensor at the foot end of the i-th leg. A value of 0 indicates that the leg is not touching the ground, and a value of 1 indicates that the leg is touching the ground.

[0124] The control method for the misstep reflex involves providing feedback values ​​to the knee and hip joints of the robot's leg at the moment the swinging leg is about to enter the support phase but has not yet touched the ground. This prompts the leg to quickly extend downwards until it touches the ground. Simultaneously, the robot's center of gravity is lowered in real time at the moment of misstep to prevent tipping due to a large drop. Specifically, the rapid downward extension of the leg can be a varying joint feedback value during the process before the foot touches the ground, or it can be an instantaneous feedback value. The mathematical model for the misstep reflex used in this invention is as follows:

[0125]

[0126] In the formula, feedback hi feedback ki The feedback values ​​from the hip and knee joints in the robot's misstep reaction are denoted by n, where n is the amplitude adjustment coefficient, and x is the value of the change. hi Let x be the hip angle value of the i-th leg. s =θ sth -Δθ h x s This represents the lower limit of the range of motion of the hip joint.

[0127] After calculating the feedback values ​​for the hip and knee joints, these values ​​are added to the joint angle values ​​of the hip and knee joints output by the CPG model to adjust the robot's motion posture. The feedback method of this invention involves adjusting the joint angle change waveform output by the CPG according to the feedback conditions.

[0128] When the quadruped robot determines that the depression is less than a certain set threshold, it only needs to adjust its posture by performing vestibular reflex.

[0129] When the controller determines that the elevation difference of the depression exceeds a certain set threshold, the quadruped robot simultaneously activates its vestibular reflex and slip reflex. For example... Figure 8 As shown in the figure, the relationship between sensor signals and reflected motion under the concave terrain is illustrated.

[0130] like Figure 9 The diagram shows the leg movements of a quadruped robot activating its vestibular reflex and slip reflex in concave terrain. The dashed line connecting A-A1-A2-A3 represents the movement trajectory at the hip joint of the thigh, and the dashed line connecting B-B1-B2-B3 represents the movement trajectory at the foot.

[0131] like Figure 10 The figure shows the motion posture changes of a quadruped robot when it traverses a concave terrain in the Webots simulation environment, triggered by the vestibular reflex and the slip reflex.

[0132] like Figure 11 The image shows the waveform of joint angle changes when a quadruped robot simultaneously activates its vestibular reflex and air-stepping reflex. The graph shows that the balance positions of the hip and knee joints continuously adjust with changes in the pitch angle, maintaining a certain degree of stability. The rapidly changing portion represents the air-stepping reflex, which allows the legs to quickly extend downwards, preventing violent vibrations of the robot body.

[0133] When the terrain is determined to be a raised area with a height less than the current leg lift height, the raised area can be crossed directly, and only the vestibular reflex posture needs to be adjusted during the crossing process.

[0134] like Figure 12 The diagram shows the relationship between sensor signals and reflected motion under raised terrain. The conditions for initiating obstacle detection reflection are that the range sensor determines that there is raised terrain ahead, and the forelegs initiate obstacle detection reflection when the foot pressure sensor determines that they have entered the swing phase.

[0135] When the terrain is determined to be a raised area with a height greater than the current leg-raising height, the quadruped robot initiates the obstacle detection reflex action as follows:

[0136] Step 1: First, swing the hip joint of the front leg backward, reducing the angle to A1, while simultaneously swinging the knee joint upward, increasing the angle to B1, to complete the contraction and raising of the leg movement.

[0137] Step 2: Swing your front leg hip forward, increasing the angle to A2, while keeping the knee joint unchanged, to complete the obstacle course.

[0138] Step 3: Keeping the hip joint angle of the front leg unchanged, swing the knee joint down, reducing the angle to B3, and complete the landing action;

[0139] Step 4: Continue to swing the hip and knee joints to the angle planned by CPG.

[0140] The values ​​A1, A2, B1, and B3 are determined by the robot's own dimensions and the height of the raised terrain. For example... Figure 13 The diagram shown illustrates the changes in leg movement in response to obstacles encountered on raised terrain, as described above. The dotted lines represent the trajectory of the foot.

[0141] In one embodiment of the present invention, the obstacle encounter and reflex implementation method planned by the quadruped robot, combined with the angle planned by the CPG, can be calculated as follows:

[0142]

[0143] In the formula, A h A is the hip joint obstacle reflex accommodation coefficient. k x is the knee joint obstacle encounter reflex accommodation coefficient. max x represents the maximum range of motion of the hip joint. i Output hip joint rotation angle to CPG, y i Output the knee joint angle to the CPG. A h A k As an adjustable parameter, it is used to control the range of change in feedback adjustment. In one embodiment of the present invention, A h =0.3, A k =0.3.

[0144] When the terrain is determined to be higher than the current leg-raising height, the quadruped robot initiates its hind leg drive mechanism by retracting its hind legs while the front legs react to the obstacle, keeping the robot's body positioned with the front higher than the back. As the robot's front legs complete their landing motion, the hind legs quickly return to their original posture, generating a forward thrust that helps the robot stably traverse the raised terrain.

[0145] When the robot determines that the height of a raised area is greater than its current leg-raising height, it will simultaneously activate obstacle detection and hind leg drive reflexes. For example... Figure 14 The image shows the posture diagrams of the quadruped robot in Webots simulation, including the front leg initiation obstacle encounter reflection and the hind leg drive reflection.

[0146] like Figure 15 The figure shows the waveform of joint angle changes when a quadruped robot simultaneously activates obstacle detection reflex and hind leg drive reflex.

[0147] As shown in the diagram, when the quadruped robot's front legs traverse raised terrain, the hip and knee joint angles change according to the previously described preset movements. The hip joint angle decreases, the thigh swings backward, and the knee joint angle increases, causing the lower leg to swing upward, thus increasing the leg lift height. After lifting the leg, the hip joint angle increases again, and the knee joint angle decreases, completing the landing motion. Simultaneously, during the front leg lift, the hind leg's knee joint angle increases and the hip joint angle decreases, helping the body lean backward and improving the robot's posture. After the front leg has essentially completed the landing motion, the hind leg quickly decreases the knee joint angle and increases the hip joint angle back to its original state, generating a momentary forward thrust that helps the robot's front legs stably and quickly traverse the raised terrain.

[0148] When the terrain is determined to be a raised area with a height greater than the current leg lift height and the front leg crosses the raised area, the posture is adjusted, and the hind leg initiates the obstacle encounter reflex and the front leg coordinate drive reflex.

[0149] The obstacle avoidance reflex control method of the hind legs is the same as that of the front legs.

[0150] The foreleg coordination drive method involves retracting the forelegs to lower the robot's forebody posture while maintaining the original forward movement control method. This allows the hindbody to be propelled forward by the forebody, reducing the load on the hind legs. This helps the hind legs traverse raised terrain.

[0151] When the pressure sensor at the foot of the hind leg detects that it has entered the swing phase, it performs an obstacle detection reflex.

[0152] like Figure 16 The image shown is a motion posture diagram of a quadruped robot in Webots simulation, depicting the coordinated driving of its front legs and the obstacle-encroaching reflex of its hind legs.

[0153] like Figure 17 The figure shows the waveforms of joint angle changes during the coordinated actuation of the front legs and the obstacle detection reflex of the hind legs in a quadruped robot. As can be seen from the figure, the hip joint angle of the front legs decreases while the knee joint angle increases, thereby lowering the posture of the forelegs and reducing the weight-bearing capacity of the hind legs. The hind legs, on the other hand, increase the lifting height by increasing the knee joint angle and decreasing the hip joint angle, and then land by increasing the hip joint angle and decreasing the knee joint angle, thus completing the obstacle detection reflex.

[0154] like Figure 18 The diagram shown is a Simulink control model of the feedback adjustment part of the quadruped robot in an example of the present invention. The main input signals are the distance value measured by the ranging sensor, the judgment value of the foot pressure sensor, the pitch angle value of the inertial measurement unit, and the joint angle value output by the encoder. The main output signals are the offset of various reflection adjustments.

[0155] The beneficial effects of this invention are as follows: The ranging sensor used in this invention can enhance the robot's perception of the terrain by judging the current terrain, thereby making preparations in advance to cross uneven terrain, and has a better perception capability.

[0156] This invention integrates a ranging sensor, a foot pressure sensor, an inertial measurement unit, and an encoder to provide quadruped robots with external terrain information and their own motion information for traversing uneven terrain. The self-motion information includes the robot's current posture angle, foot contact with the ground, joint angles, etc., which provides powerful assistance for the robot to perform various reflex movements and posture adjustments.

[0157] This invention activates vestibular feedback and step reflex simultaneously in concave terrain, allowing the quadruped robot to adjust its posture while its legs rapidly extend downwards. This improves the quadruped robot's stability, maneuverability, and autonomous adjustment capabilities, making it more robust in changing environments.

[0158] This invention simultaneously activates obstacle detection reflexes and hind leg coordinated drive when encountering raised terrain, helping the robot quickly traverse such terrain and reducing time and energy consumption. The hind leg drive method allows the robot to cross raised terrain even with lower torque, making it more adaptable to its environment.

[0159] This invention activates the vestibular reflex and coordinates the foreleg drive simultaneously when encountering raised terrain, helping the robot's hind legs to quickly cross raised terrain. By lowering the center of gravity of the forebody and moving the center of gravity forward, it helps the hind legs to better cross raised terrain.

[0160] In this invention, the robot combines multiple reflection methods through information from various sensors, giving it better stability and robustness when traversing uneven terrain.

[0161] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion, characterized in that, Includes the following steps: S1: First, the quadruped robot uses the rhythmic signals output by the multi-sensor fusion central mode generator to plan its gait. Then, through the range sensor and inertial measurement unit, based on the quadruped robot's own size and the distance values ​​of the robot's body from the ground directly below and from the ground diagonally in front obtained by the range sensor, the distance values ​​are judged and processed to distinguish the type of terrain in front of the quadruped robot, including concave terrain, convex terrain, or flat terrain. S2: When the terrain is determined to be concave in S1, the quadruped robot simultaneously activates the vestibular reflex and step-off reflex algorithms to ensure that the quadruped robot does not cause a large fluctuation in the body when its feet land, and to complete the body posture adjustment after landing to ensure that the body does not tilt too much, preventing excessive shift of the center of gravity and uneven force on the four legs caused by excessive forward and backward tilt angles. S3: When the terrain is determined to be raised in S1, the quadruped robot judges the height range of the raised terrain. If the height of the raised terrain is less than the current leg height of the quadruped robot, the quadruped robot directly crosses it, and only needs to adjust its posture by using vestibular reflex based on the inertial measurement unit. If the height of the raised terrain is greater than the current leg height of the quadruped robot, the quadruped robot needs to activate obstacle detection reflex and hind leg drive reflex to help complete the front leg crossing. After the front leg crosses the raised terrain, the robot adjusts its posture to lower the center of gravity of the forequarters, and then the hind leg crosses. The hind leg crossing activates obstacle detection reflex and front leg coordinated drive reflex to cross the raised terrain. S4: When the terrain is determined to be flat in S1, the quadruped robot walks according to the gait planned by the rhythm signal output by the central pattern generator CPG. The vestibular reflex is applied to the dynamic posture adjustment of quadruped robots in uneven terrain, and its control method is as follows: Vestibular reflex calculation method: offset h The offset is the offset of the robot's hip joint balance position. k α is the offset of the knee joint balance position of the quadruped robot, and α is the terrain slope. Since the body posture angle is approximately linearly related to the slope, the posture angle balance coefficient h1 is obtained. The posture angle is mapped to the joint angle change to obtain the joint angle balance coefficient h2. The control method for the misstep reflex is as follows: At the moment the robot's swinging leg enters the support phase but has not yet touched the ground, a feedback value is given to the knee and hip joints of the robot's leg, causing it to quickly extend downwards until it touches the ground. Simultaneously, at the instant of misstepping, the robot's center of gravity is lowered in real time to prevent the robot from tipping over due to large height differences in concave terrain. The mathematical model for the misstep reflex is as follows: In the formula, feedback hi feedback ki The feedback values ​​from the hip and knee joints in the robot's misstep reaction are denoted by n, where n is the amplitude adjustment coefficient, and x is the value of the change. hi Let x be the hip angle value of the i-th leg. s =θ sth -Δθ h x s θ represents the lower limit of the range of motion of the hip joint. sth The joint angle Δθ represents the phase switching of the hip joint. h This indicates setting a threshold for the range of motion of the hip joint.

2. The control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion according to claim 1, characterized in that, The robot's motion is adjusted by fusing multi-sensor information based on the actual terrain. The feedback is added to the output signal of the central pattern generator (CPG) for actual motion adjustment. The mathematical model is as follows: In the formula, X i Y i x represents the actual hip and knee joint angle input. hi y ki The offset represents the hip and knee joint angle signals output by the CPG oscillator. hi offset ki The feedback values ​​for the vestibular reflex at the hip and knee joints. hi feedback ki Feedback from the hip and knee joints, used for the robot's reaction to missteps or obstacles, is used to control its interaction with the outside world based on terrain changes, thereby achieving dynamic feedback to adapt to the terrain.

3. The control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion according to claim 1, characterized in that, The method for determining the terrain type in front of the quadruped robot by judging the distance value is as follows: Where DS2 is the distance value collected by the ranging sensor at the bottom of the quadruped robot's forebody, l is the length of a single leg of the quadruped robot, θ represents the initial angle of the quadruped robot's joint, threshold2 is the height threshold set according to the actual situation, and max_height is the maximum height difference set according to the robot's own size. The depressions and protrusions defined here are within the robot's crossing capability. In this calculation method, the distance value collected by the ranging sensor is compared with the distance from the bottom of the robot's body to the bottom of its feet. If the distance value obtained by the sensor is greater than the distance from the bottom of the robot's body to the bottom of its feet, and the difference is greater than a set threshold, it is judged as a concave terrain. Similarly, if the distance value obtained by the sensor is less than the distance from the bottom of the robot's body to the bottom of its feet, and the difference is less than the set threshold, it is judged as a convex terrain. When the difference between the distance value obtained by the sensor and the distance from the bottom of the robot's body to the bottom of its feet is within the set threshold range, it is judged as a flat terrain. The size of the set threshold is based on the range of fluctuation of the robot under flat terrain to prevent the robot from misjudging the terrain. Further judgment methods are as follows: When the robot determines that DS2-2lcosθ≥max_height, it considers the terrain to be too deep and cannot be crossed, and the route needs to be changed. When the robot determines that DS2-2lcosθ≤-max_height, it considers the terrain to be too high and cannot be crossed, and the route needs to be changed.

4. The control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion according to claim 1, characterized in that, The quadruped robot activates a step-off reflex when it is in a concave terrain. When the DS2 ranging sensor detects that the terrain in front of the robot is concave, it uses foot pressure sensors and encoders to determine whether the feet touch the ground when the robot enters the support phase. If the feet do not touch the ground, the step-off reflex is activated to prevent the robot from experiencing severe vibrations and imbalances caused by stepping off the ground.

5. The control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion according to claim 4, characterized in that, The judgment method combining foot pressure sensor and encoder is as follows: when the encoder detects that the hip and knee joints of the quadruped robot have reached the angle of entering the support phase, the feedback value of the foot pressure sensor is obtained at the same time. If the foot pressure sensor feedback is that it is not touching the ground, it is marked as stepping into the air; if the feedback is that it has touched the ground, it is marked as normal foot landing. The judgment method is expressed as follows: Where i represents the i-th leg, x hi y represents the hip angle of the i-th leg. ki θ represents the knee angle of the i-th leg. stk The joint angle Δθ represents the phase transition of the knee joint. k This indicates setting the threshold for the range of motion of the knee joint, touchsenor i This represents the feedback value of the pressure sensor at the foot end of the i-th leg. A value of 0 indicates that the leg is not touching the ground, and a value of 1 indicates that the leg is touching the ground.

6. The control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion according to claim 1, characterized in that, When the terrain is determined to be a raised area with a height greater than the current leg-raising height, the quadruped robot initiates the obstacle detection reflex action as follows: Step 1: First, swing the hip joint of the front leg backward, reducing the angle to A1, while simultaneously swinging the knee joint upward, increasing the angle to B1, to complete the contraction and raising of the leg movement. Step 2: Swing your front leg hip forward, increasing the angle to A2, while keeping the knee joint unchanged, to complete the obstacle course. Step 3: Keeping the hip joint angle of the front leg unchanged, swing the knee joint down, reducing the angle to B3, and complete the landing action; Step 4: Continue swinging the hip and knee joints to the angle planned by CPG; The values ​​of A1, A2, B1, and B3 are determined by the robot's own dimensions and the height of the raised terrain.

7. The control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion according to claim 1, characterized in that, The obstacle-detection and reflex implementation method planned for the quadruped robot, combined with the angle planned by CPG, is calculated as follows: In the formula, A h A is the hip joint obstacle reflex accommodation coefficient. k x is the knee joint obstacle encounter reflex accommodation coefficient. max x represents the maximum range of motion of the hip joint. hi Output hip joint angle to CPG, y ki Output the knee joint angle for CPG, A h A k It is an adjustable parameter used to control the magnitude of changes in feedback regulation.

8. The control method for a quadruped robot traversing uneven terrain based on multi-sensor fusion according to claim 1, characterized in that, The conditions for triggering obstacle detection reflection are: the ranging sensor determines that there is a raised terrain in front; the front leg triggers obstacle detection reflection when the foot pressure sensor determines that it has entered the swing phase; the hind leg triggers obstacle detection reflection when the foot pressure sensor determines that it has entered the swing phase. The obstacle detection reflection control method of the hind leg is the same as that of the front leg. When the terrain is determined to be a raised area with a height greater than the current leg lifting height, the quadruped robot initiates a rear leg drive reflex method. At the same time as the front leg performs an obstacle detection reflex, the rear leg retracts, keeping the robot's body in a front-high, rear-low position. When the robot's front leg completes the landing action, the rear leg quickly returns to its original posture, generating a forward thrust that helps the robot stably cross the raised terrain. The foreleg coordinated drive reflex method is as follows: the forelegs retract, which lowers the robot's forebody posture, while maintaining the original forward movement control method, so that the hindbody can be driven forward by the forebody, reducing the load on the hind legs and helping the hind legs cross raised terrain.

Images