Speed map-based guidance control method and brain-like force sensing glove

Abstract

The application discloses a kind of guiding control method and brain-like force sense glove based on speed map, the method includes constructing mixed vector field and virtual potential field;Combined with mixed vector field and virtual potential field, construct global convergent vector field as speed map;Obtain guide coordinates, and input speed map to obtain guide information;Based on guide information and the current speed of robot, obtain error information;Substitute it into controller to realize robot control.The application designs a global convergent vector as guide map in advance, guides system state to tend to be stable, so that robot is attracted to desired trajectory when there is disturbance, compensates the uncertainty and interference of control system with minimum calculation amount, significantly reduces the transient overshoot in flexible interaction process;Design three-dimensional force accurate estimation based on brain-like force sense glove, combined with guide control method and admittance relationship, realize flexible interaction of any type and any number of robots.

Description

Technical Field

[0001] This invention relates primarily to the field of robotics, and more specifically, to a guidance and control method based on a velocity map and a brain-like sensory glove. Background Technology

[0002] In the fields of robotics and human-computer interaction, achieving efficient, stable, and flexible navigation and control of robots in complex environments has always been a research hotspot and challenge. Traditional robot control methods often rely on precise models and strict control conditions. However, in practical applications, robots often face challenges such as environmental uncertainties, external interference, and complex dynamic interactions. These factors often lead to decreased control performance or even system instability. Especially in situations requiring high-precision flexible interaction, such as medical surgical robots, precision manufacturing, and human-computer interaction devices, extremely high demands are placed on the robot's control accuracy, response speed, and stability. However, most existing robot control technologies are based on classical control theory, which is insufficient to effectively cope with complex and ever-changing interaction scenarios. Particularly in flexible interaction processes requiring rapid response and high-precision control, traditional control methods often suffer from unstable control effects and large transient overshoot.

[0003] Furthermore, with the continuous development of robotics technology, robots are no longer merely tools for performing single tasks, but are increasingly involved in collaborative work with humans or other robots. In human-computer interaction, compliant interaction is difficult due to limitations in robot model, size, and number, and force sensors are expensive, making it difficult to equip every robot with them. Therefore, developing an interactive device capable of achieving high-precision, flexible interaction between humans and multiple robots has become an urgent problem to be solved in the field of human-computer interaction. Summary of the Invention

[0004] The purpose of this invention is to provide a guidance control method based on a velocity map and a neuromorphic force-sensing glove. This method utilizes a globally convergent vector field as a guidance map to guide the robot's tracking trajectory at any position, compensating for all disturbances in the robot system with minimal control force. Furthermore, in compliant interaction, the guidance control method effectively reduces overshoot during force contact while compensating for dynamic disturbances. Further, to address the problem of human-multi-machine interaction, a neuromorphic force-sensing glove is proposed, enabling coordinated work between a human and any number, model, and size of robotic arms.

[0005] The technical solution of the present invention is as follows:

[0006] On one hand, this invention discloses a guidance and control method based on a speed map, the method comprising the following steps:

[0007] S1. Construct a reasonable hybrid vector field, in which all points converge to the desired point;

[0008] S2. Construct a suitable virtual potential field and combine it with a hybrid vector field to achieve smooth convergence;

[0009] S3. Combine the hybrid vector field and the virtual potential field to construct a global convergent vector field, so that all points in the map can smoothly converge to the target point, and use it as a velocity map.

[0010] S4. Based on the predefined trajectory, obtain the guidance coordinates through coordinate transformation equations, and input the velocity map to obtain guidance information in real time;

[0011] S5. Based on the guidance information and the robot's current speed, the error is obtained;

[0012] S6. By substituting the error information into the controller, robot control is achieved.

[0013] Preferably, the globally convergent vector field in step S3 is obtained from the hybrid vector field constructed in step S1 and the virtual potential field constructed in step S2. The hybrid vector field is constructed by superimposing multiple vector fields, all of which are obtained based on predefined paths. The target path of the vector field is defined by the intersection of several hypersurfaces, and each hypersurface is given by the zero level set of a smooth function. The globally convergent vector field is expressed as:

[0014] In formula (1) This is a globally convergent vector field, i.e., a velocity map, where n is the dimension of the vector field, s is the scaling factor, n0 is the number of planes with zero values ​​in the vector field, S represents all axes in n-dimensional space, and I represents all axes excluding x. w The set of axes of the axis variable, X G For vector field space, For each vector field in the aforementioned hybrid vector field; the X G Represented as:

[0015] X G =X=[x1, x2, ..., x n ] T (2)

[0016] In equation (2), X is the robot coordinate space; S and I are represented as:

[0017] S = x k , k=1,…,n (3)

[0018] I = S\x w =i j , j = 1, ..., n-1 (4)

[0019] The error of the vector field is expressed as:

[0020]

[0021] In formula (5) yes The vector of partial derivatives of is expressed as:

[0022]

[0023] In formula (6) It is a Kronecker function, expressed as:

[0024]

[0025] The hybrid vector field consists of n vector fields, and the desired paths of these vector fields can be represented as follows:

[0026]

[0027] In formula (8) The paths intersecting the n-1 planes; the mixed vector field can be represented as:

[0028]

[0029] In equation (9), sgn(·) is the sign function, and r0 and k j All are normal numbers, ⊥ φ (X G ) is the outer product of all vectors, ⊥ φ (X G = [v1, ..., v] n ] T , It is a vector ⊥ φ (X G The sum of all elements in ), d w , which is the virtual potential field constructed in step S2;

[0030] Preferably, the virtual potential field in step S2 is represented as:

[0031]

[0032] In equation (10), e represents the natural logarithm.

[0033] Preferably, the two-dimensional representation of the mixed vector field in step S1 is as follows:

[0034]

[0035] In equation (11), A is a [0 1; -1 0] matrix. It is a vector The sum of all elements in the expression.

[0036] Preferably, the coordinate transformation equation in step S4 is expressed as:

[0037] X g (t)=X(t)-X d (t) (12)

[0038] In the formula X g X(t) represents the position of the end effector after coordinate transformation, and X(t) represents the position of the robot in space. d (t) represents the desired position; according to formula (1), the guidance information is expressed as

[0039] Preferably, the error in step S5 is expressed as:

[0040]

[0041] In equation (13), k g X is a positive constant. g Let X be the position of the robot in the global convergence vector field space coordinate system. g =XX d X d This represents the desired trajectory.

[0042] Preferably, the controller in step S6 is represented as:

[0043]

[0044] In equation (14), k c It is a positive number.

[0045] Preferably, the robot control in step S6 involves substituting the controller into the robot dynamics model for compliant trajectory tracking. The robot dynamics model is represented as follows:

[0046]

[0047] In equation (15), M(q)∈R n×n Represents the inertia matrix. Let G(q) ∈ R represent the centripetal force and Coriolis force vectors. n Represents the gravity matrix; vector q, and Represent the position, velocity, and acceleration of the joint, respectively; vector τ∈R n τ e ∈Rn and τ d ∈R n These represent the input torque, the external torque in the workspace, and the unknown bounded disturbance, respectively; M(q), G(q) consists of a nominal part and a bounded uncertain part, as follows:

[0048]

[0049] The robot dynamics model can be represented as:

[0050]

[0051] In equation (17), D(t) is

[0052] The τ is represented as: τ=J(q) T u (18)

[0053] In equation (18), J(q) is the Jacobian matrix;

[0054] The τ e It can be represented as: τ e =J(q) T F e (19)

[0055] In equation (19), F e External force;

[0056] The positive kinematic mapping between the end effector and the joint angle is X = L(q). Therefore, the robotic arm dynamics model can be expressed as:

[0057]

[0058] The robot control achieves compliant human-machine interaction through admittance relations, which are expressed as follows:

[0059]

[0060] In equation (21) M d It is the inertia of expectation, B d It is damping, K d It is the stiffness matrix, X r , and These are the reference trajectory position, velocity, and acceleration; the reference trajectory is a shaped trajectory used for smooth interaction; the coordinate transformation equation is expressed as:

[0061] X g (t)=x(t)-X r (t) (22)

[0062] The X g (t) is limited in range by a saturation function, and is expressed as follows:

[0063]

[0064] In equation (23), K T It is a threshold. Representing vectors Each element in the set is restricted to (-K) T K T );

[0065] The error E of the robotic arm is expressed as

[0066]

[0067] Then, substitute equations (22) and (24) into equation (14) to obtain controller u, and then substitute u into the robotic arm dynamics model equation (20) to realize real-time control of the robotic arm.

[0068] On the other hand, the present invention also discloses a brain-like force-sensing glove, which adopts the speed map-based guidance and control method described above, and controls the robot in step S6 to perform compliant trajectory tracking by substituting the controller into the robot dynamics model; including a force estimation algorithm based on a brain-like neural network, an upper sleeve surface (1), an upper film (2), a piezoresistive material (3), a lower film (4) and a lower sleeve surface (5), a controller, and a main control console; the upper sleeve surface (1) is a glove-shaped structure fixedly connected to the side of the upper film (2), and the upper film (2) is provided with wires for detection. The resistance value of the piezoresistive material (3) is measured. The resistance value of the piezoresistive material (3) can be adjusted according to the pressure change. The piezoresistive material (3) is fixed to the upper film (2) and the lower film (4). The lower film (4) is provided with a wire for detecting the resistance value of the piezoresistive material (3). The lower film (4) is provided with a protrusion at each position of the piezoresistive material to increase the detection range. The lower film (4) is fixedly connected to the lower sleeve surface (5). The lower sleeve surface (5) is fixedly connected to the side of the upper sleeve surface (1). The controller can collect the resistance change of the piezoresistive material (3) to obtain the pressure value.

[0069] Preferably, the controller includes a neuromorphic chip, a microcontroller, and a wireless transmission module. The neuromorphic chip is used to deploy the force estimation algorithm based on the neuromorphic neural network. The force estimation algorithm based on the neuromorphic neural network outputs a multi-channel pressure sensing sequence by pre-collecting multiple piezoresistive materials (3) of the neuromorphic force sensing glove under different directions and forces. During the acquisition process, a three-dimensional pressure sensor is used as the correct label to construct a neuromorphic force sensing glove dataset. Then, the acquired force sensing sequence is converted into a pulse sequence and input into a spiking neural network to train a neuromorphic three-dimensional force sensing model. The three-dimensional force is obtained by inputting the force sensing sequence of the neuromorphic force sensing glove into the neuromorphic three-dimensional force sensing model in real time. The three-dimensional force signal is then sent to the main control console through the wireless transmission module.

[0070] Preferably, the main controller collects the three-dimensional force perception of the brain-like force-sensing glove in real time during the human-computer interaction process, converts the force into the workspace of each robot for trajectory updates, and controls the robot through a speed map-based guidance control method to achieve human-machine collaborative tasks.

[0071] This invention provides a velocity map-based guidance control method and a brain-like force-sensing glove, which have the following beneficial effects: By pre-designing a global convergence vector as a guidance map, the system state tends to stabilize, thereby attracting the robot to the desired trajectory under disturbances. This compensates for the uncertainty and interference of the control system with minimal computation and significantly reduces transient overshoot during flexible interaction. Furthermore, a brain-like neural network-based force-sensing glove is designed for accurate estimation of three-dimensional forces. Combined with the guidance control method and admittance relation, this enables flexible interaction of any type and number of robots. Attached Figure Description

[0072] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0073] Figure 1 This is a flowchart of a speed map-based guidance and control method according to the present invention;

[0074] Figure 2 This is a schematic diagram of the structure of the brain-like sensory glove described in this invention;

[0075] Figure 3 This is a schematic diagram of the trajectory tracking error of D0 under interference-free conditions; (a) GC tracking error; (b) ANIC tracking error; (c) SOBC tracking error; (d) external force;

[0076] Figure 4Here are schematic diagrams of the trajectory tracking errors of time-varying disturbance D1; (a) GC tracking error; (b) ANIC tracking error; (c) SOBC tracking error; (d) GC actual trajectory;

[0077] Figure 5 This is a schematic diagram of the trajectory tracking error of the hybrid disturbance D2; (a) GC tracking error; (b) ANIC tracking error; (c) SOBC tracking error; (d) GC actual trajectory;

[0078] Figure 6 This is a schematic diagram of the trajectory tracking error under strong perturbation D3; (a) GC tracking error; (b) ANIC tracking error; (c) SOBC tracking error; (d) GC vector field error;

[0079] Figure 7 These are the tracking errors of the PD controller under different disturbances; (a) D0; (b) D1; (c) D2; (d) D3.

[0080] Explanation of reference numerals in the attached figures:

[0081] Upper sleeve surface (1), upper thin film (2), piezoresistive material (3), lower thin film (4), and lower sleeve surface (5). Root mean square error (RMS), root mean square error of the x-axis (zx), root mean square error of the y-axis (zy), and root mean square error of the z-axis (zz), reference trajectory (Xr), end effector position of the robotic arm (X), controller of the present invention (GC), adaptive neural network controller (ANIC), state observer-based controller (SOBC), PD controller (PDC), no disturbance (D0), time-varying disturbance (D1), mixed disturbance (D2), strong disturbance (D3). Detailed Implementation

[0082] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0083] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0084] like Figure 1 This embodiment provides a guidance and control method based on a speed map, the method including the following steps:

[0085] S1. Construct a reasonable hybrid vector field, in which all points converge to the desired point;

[0086] S2. Construct a suitable virtual potential field and combine it with a hybrid vector field to achieve smooth convergence;

[0087] S3. Combine the hybrid vector field and the virtual potential field to construct a global convergent vector field, so that all points in the map can smoothly converge to the target point, and use it as a velocity map.

[0088] S4. Based on the predefined trajectory, obtain the guidance coordinates through coordinate transformation equations, and input the velocity map to obtain guidance information in real time;

[0089] S5. Based on the guidance information and the robot's current speed, the error is obtained;

[0090] S6. By substituting the error information into the controller, robot control is achieved.

[0091] Furthermore, the globally convergent vector field in step S3 is obtained from the hybrid vector field constructed in step S1 and the virtual potential field constructed in step S2. The hybrid vector field is constructed by superimposing multiple vector fields, all of which are obtained based on predefined paths. The target path of the vector field is defined by the intersection of several hypersurfaces, and each hypersurface is given by the zero level set of the smooth function. The globally convergent vector field is expressed as:

[0092]

[0093] In formula (1) This is a globally convergent vector field, i.e., a velocity map, where n is the dimension of the vector field, s is the scaling factor, n0 is the number of planes with zero values ​​in the vector field, S represents all axes in n-dimensional space, and I represents all axes excluding x. w The set of axes of the axis variable, X G For vector field space, For each vector field in the aforementioned hybrid vector field; the X G Represented as:

[0094] X G =X=[x1, x2, ..., x n ] T (2)

[0095] In equation (2), X is the robot coordinate space; S and I are represented as:

[0096] S = x k , k=1,...,n (3)

[0097] I = S\x w =i j , j = 1, ..., n-1 (4)

[0098] The error of the vector field is expressed as:

[0099]

[0100] In formula (5) yes The vector of partial derivatives of is expressed as:

[0101]

[0102] In formula (6) It is a Kronecker function, expressed as:

[0103]

[0104] The hybrid vector field consists of n vector fields, and the desired paths of these vector fields can be represented as follows:

[0105]

[0106] In formula (8) The paths intersecting the n-1 planes; the mixed vector field can be represented as:

[0107]

[0108] In equation (9), sgn(·) is the sign function, and r0 and k j All are normal numbers, ⊥ φ (X G ) is the outer product of all vectors, ⊥ φ (X G = [v1, ..., v] n ] T , It is a vector ⊥ φ (X G The sum of all elements in ), d w The virtual potential field constructed in step S2.

[0109] Furthermore, the virtual potential field in step S2 is represented as:

[0110]

[0111] In equation (10), e represents the natural logarithm.

[0112] Furthermore, the two-dimensional representation of the mixed vector field in step S1 is as follows:

[0113]

[0114] In equation (11), A is a [0 1; -1 0] matrix. It is a vector The sum of all elements in the expression.

[0115] Furthermore, the coordinate transformation equation in step S4 is expressed as:

[0116] X g (t)=X(t)-X d (t) (12)

[0117] In the formula X g X(t) represents the position of the end effector after coordinate transformation, and X(t) represents the position of the robot in space. d (t) represents the desired position; according to formula (1), the guidance information is expressed as

[0118] Furthermore, the error in step S5 is expressed as:

[0119]

[0120] In equation (13), k g X is a positive constant. g Let X be the position of the robot in the global convergence vector field space coordinate system. g =XX d X d This represents the desired trajectory.

[0121] Furthermore, the controller mentioned in step S6 is represented as follows:

[0122]

[0123] In equation (14), k c It is a positive number.

[0124] Furthermore, the robot control in step S6 involves substituting the controller into the robot dynamics model for compliant trajectory tracking. The robot dynamics model is represented as follows:

[0125]

[0126] In equation (15), M(q)∈R n×n Represents the inertia matrix. Let G(q) ∈ R represent the centripetal force and Coriolis force vectors. n Represents the gravity matrix; vector q, and Represent the position, velocity, and acceleration of the joint, respectively; vector τ∈R n τ e ∈R n and τ d∈R n These represent the input torque, the external torque in the workspace, and the unknown bounded disturbance, respectively; M(q), G(q) consists of a nominal part and a bounded uncertain part, as follows:

[0127]

[0128] The robot dynamics model can be represented as:

[0129]

[0130] In equation (17), D(t) is

[0131] The τ is represented as: τ=J(q) T u (18)

[0132] In equation (18), J(q) is the Jacobian matrix;

[0133] The τ e It can be represented as: τ e =J(q) T F e (19)

[0134] In equation (19), F e External force;

[0135] The positive kinematic mapping between the end effector and the joint angle is X = L(q). Therefore, the robotic arm dynamics model can be expressed as:

[0136]

[0137] The robot control achieves compliant human-machine interaction through admittance relations, which are expressed as follows:

[0138]

[0139] In equation (21) M d It is the inertia of expectation, B d It is damping, K d It is the stiffness matrix, X r , and These are the reference trajectory position, velocity, and acceleration; the reference trajectory is a shaped trajectory used for smooth interaction; the coordinate transformation equation is expressed as:

[0140] X g (t)=X(t)-X r (t) (22)

[0141] The Xg (t) is limited in range by a saturation function, and is expressed as follows:

[0142]

[0143] In equation (23), K T It is a threshold. Representing vectors Each element in the set is restricted to (-K) T K T );

[0144] The error E of the robotic arm is expressed as

[0145]

[0146] Then, substitute equations (22) and (24) into equation (14) to obtain controller u, and then substitute u into the robotic arm dynamics model equation (20) to realize real-time control of the robotic arm.

[0147] like Figure 3 As shown, a brain-like force-sensing glove includes a force estimation algorithm based on a brain-like neural network, an upper sleeve (1), an upper film (2), a piezoresistive material (3), a lower film (4), and a lower sleeve (5), a controller, and a main control console. The upper sleeve (1) is a glove-shaped structure fixedly connected to the side of the upper film (2). The upper film (2) is provided with wires for detecting the resistance value of the piezoresistive material (3). The piezoresistive material (3) can change its resistance value according to pressure. The piezoresistive material (3) is fixed to the upper film (2) and the lower film (4). The lower film (4) is provided with wires for detecting the resistance value of the piezoresistive material (3). The lower film (4) has protrusions at each piezoresistive material position to increase the detection range. The lower film (4) is fixedly connected to the lower sleeve (5), and the lower sleeve (5) is fixedly connected to the side of the upper sleeve (1). The controller can collect the resistance change of the piezoresistive material (3) to obtain the pressure value. The robot control in step S6 is to substitute the controller into the robot dynamics model for compliant trajectory tracking.

[0148] A dynamic model of a robotic arm is selected for simulation comparison. The dynamic model is represented as follows:

[0149]

[0150] in,

[0151] A=sin(q_1)sin(q_2)+cos(q_1)cos(q_2)$, $B

[0152] =cos(q_1)sin(q_2)-sin(q_1)cos(q2)

[0153] g = 9.8 kg / N

[0154] Each link is 1m long, and the first and second links have masses of 5kg and 10kg respectively. Comparative experiments were conducted using PDC, SOBC, and ANIC. The proposed GC parameters are k... c =4000, k g =400, s=1 / 3, q1=q2=1 / 2rad, r o =1,k j =1,K T =0.5 and j=1,2.

[0155] Furthermore, the controller includes a neuromorphic chip, a microcontroller, and a wireless transmission module. The neuromorphic chip is used to deploy the force estimation algorithm based on a neuromorphic neural network. The force estimation algorithm based on a neuromorphic neural network outputs a multi-channel pressure sensation sequence by pre-collecting multiple piezoresistive materials (3) of the neuromorphic force-sensing glove under different directions and forces. During the acquisition process, a three-dimensional pressure sensor serves as the correct label to construct a neuromorphic force-sensing glove dataset. Then, the acquired force sensation sequence is converted into a pulse sequence and input into a spiking neural network to train a neuromorphic three-dimensional force perception model. The three-dimensional force is obtained by inputting the force sensation sequence of the neuromorphic force-sensing glove into the neuromorphic three-dimensional force perception model in real time. The three-dimensional force signal is then sent to the main control console via the wireless transmission module. During human-computer interaction, the main controller collects the three-dimensional force sensation of the neuromorphic force-sensing glove in real time, converts the force into the workspace of each robot for trajectory updates, and controls the robot through a speed map-based guidance control method to achieve human-machine collaborative tasks.

[0156] This example verifies the performance of GC through four sets of experiments. The expected trajectory of the final actuator is... The parameter of the admittance relationship is M. d = diag([1, 1]), B d =diag([40,40]) and K d =diag([300,800]), during the time interval 5-7s, the external force is F. e =[20sin(t),20sin(t)] T The perturbation is D0 = [0, 0] T , D1=[(t)+400(t), 600(t)+200(t)] T , and

[0157] Figure 3The performance of different controllers in a noise-free environment was demonstrated. GC, PDC, ANIC, and SOBC showed strong trajectory tracking performance, with PDC exhibiting the best tracking capability.

[0158] Figure 4 and 5 The chart represents the tracking performance of various controllers under time-varying and mixed disturbances. It is evident that the tracking performance of ANIC and SOBC significantly deteriorates due to disturbances, with ANIC exhibiting better disturbance immunity than SOBC. Both SOBC and ANIC still exhibit considerable overshoot even without external forces. When contact forces are present, mixed disturbances exacerbate the overshoot of ANIC. Notably, GC's performance is unaffected by disturbances, maintaining low computational complexity and high tracking accuracy, and effectively suppressing overshoot caused by contact forces.

[0159] However, as Figure 7 As shown, the PDC exhibits significant trajectory tracking errors under various interference conditions, indicating insufficient anti-interference capability. Figure 6 As shown, when the external force F e During interaction and disengagement, both ANIC and SOBC exhibit considerable overshoot, while GC effectively suppresses overshoot, resulting in smoother trajectory tracking. Tables 1 and 2 further highlight GC's ability to achieve high-precision control with minimal control input in disturbed environments. Table 1 shows the root mean square error for different controllers. Table 2 shows the runtime for 2000 iterations for different controllers.

[0160] Table 1

[0161]

[0162] Table 2

[0163]

[0164] Although strong disturbances are rare, they can occur in actual control processes due to equipment failure or other factors. Figure 7 This demonstrates the performance of different controllers under such severe disturbances. Under strong disturbances, neither SOBC nor ANIC can effectively track the trajectory. Therefore, GC maintains high control accuracy even under strong disturbances.

[0165] In all experiments, the speed map-based guidance control method proposed in this invention has the shortest response time, while the neural network-based controller has the longest response time. Furthermore, the guidance control scheme effectively compensates for all uncertainties and disturbances with minimal computational load.

[0166] In the following description, terms such as “inner,” “outer,” “upper,” “lower,” “left,” and “right” are used only to indicate orientation or positional relationship for the convenience of describing the embodiments and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

Claims

1. A speed map-based guidance control method characterized by comprising: The method comprises the following steps: S1, constructing a mixed vector field, all points in the vector field can converge to the desired point; S2, constructing a virtual potential field, and combining the mixed vector field to achieve smooth convergence; S3, combining the mixed vector field and the virtual potential field to construct a global convergence vector field, so that all points in the map can smoothly converge to the target point, and the global convergence vector field is taken as a velocity map; S4, according to the predefined trajectory, the guidance coordinates are obtained through the coordinate transformation equation, and the guidance coordinates are input into the velocity map to obtain the guidance information in real time; S5, based on the guidance information and the current speed of the robot, error information is obtained; The error of step S5 is represented as In equation (13), k g X is a positive constant. g Let X be the position of the robot in the coordinate system of the convergent vector field in the whole space. g =XX d X d X represents the desired trajectory; X is the robot's position in space. S6, the error information is brought into the controller to realize robot control; The controller of step S6 is represented as: k in formula (14) is a normal number; c is a normal number; is a guide information.

2. The speed map-based guidance control method according to claim 1, characterized by, The global convergence vector field of step S3 is obtained from the mixed vector field of step S1 and the virtual potential field of step S2, the mixed vector field is constructed by superimposing a plurality of vector fields, the plurality of vector fields are obtained based on the predefined path, the target path of the plurality of vector fields is defined by the intersection of several hypersurfaces, and each hypersurface is given by the zero level set of a smooth function; the global convergence vector field is represented as: In formula (1) is a full convergent vector field, i.e. a velocity map, n is the dimension of the vector field, s is a scale factor, n0 is the number of planes in the vector field that are zero, S represents all axes in n-dimensional space, I represents the set of axes that do not include x w is the set of axes of the axis variable, X G is the vector field space, is each vector field in the mixed vector field; the X G is represented as: X G = X = [x1, x2,..., x n ] T (2) In formula (2), X is the robot coordinate space; S and I are represented as: S = x k , k = 1,..., n (3) I = S\x w = ij, j = 1,..., n - 1 (4) Wherein, the error of the vector field is represented as: The partial derivative vector of the function (5) is denoted as: is the partial derivative vector of the function (5) is denoted as: In formula (6) is a Kronecker delta function, expressed as: The mixed vector field is composed of n vector fields, and the expected path of the vector fields can be represented as In formula (8) is a path intersecting n-1 planes; the mixed vector field can be represented as: In formula (9), sgn(·) is a sign function, r0and k j are normal numbers, and φ (X G ) is an outer product of all vectors, and φ (X G ) = [v1,..., v n ] T , is a sum of all elements in the vector φ (X G ), d w is a virtual potential field constructed in step S2; the virtual potential field is represented as: In formula (10), e represents the natural logarithm.

3. The speed map based guidance control method according to claim 2, characterized by, The representation of the mixed vector field of step S1 in the two-dimensional space is: A is a [0 1; -1 0] matrix in equation (11), is the sum of all elements in the vector is the sum of all elements in the vector 4. The speed map based guidance control method according to claim 2, characterized by, The coordinate transformation equation of step S4 is represented as: X g (t) = X(t) - X d (t) (12) where Xg(t) is the position of the end effector after coordinate transformation, X(t) is the position of the robot in space, X d (t) is the desired position; according to equation (1), the guidance information is represented as 5. The speed map based guidance control method according to claim 4, characterized by, The robot control of step S6 is to bring the controller into the robot dynamics model for compliant trajectory tracking, the robot dynamics model is a robot dynamics model, and is represented as: M(q) e R in formula (15) n×n represents an inertia matrix, represents a centripetal force and Coriolis force vector, G(q) e R n represents a gravity matrix; Vector q, and denote the position, velocity and acceleration of the joint, respectively; vector τ∈R n , τ e ∈R n and τ d ∈R n denote the input torque, external torque in the workspace and unknown bounded disturbance, respectively; M(q) = A(q) + B(q)u G(q) = A(q) + B(q)u G(q) = A(q) + B(q)u The robot dynamics model can be represented as: D(t) in formula (17) is The τ can be represented as: τ = J(q) T u (18) In formula (18), J(q) is a Jacobian matrix; The tau e may be expressed as: τ e = J(q) T F e (18) F in formula (18) e is an external force; the robot control achieves compliant human-robot interaction through a mobility relationship, which is expressed as: In equation (19) M d It is the inertia of expectation, B d It is damping, K d It is the stiffness matrix, X r , and These are the reference trajectory position, velocity, and acceleration; the reference trajectory is a shaped trajectory used for smooth interaction; the coordinate transformation equation is expressed as: X g (t) = X(t) - X r (t) (19) The X g (t) Range defined by a saturation function, expressed as wherein K T is a threshold value, denotes each element of the vector which are restricted to (-K T , K T ).

6. A brain-like dexterous glove, characterized by, A velocity map-based guidance control method according to claim 1 is adopted, and the controller is brought into the robot dynamics model for compliant trajectory tracking; The method comprises a brain-like neural network-based force estimation algorithm, an upper sleeve (1), an upper film (2), a piezoresistive material (3), a lower film (4) and a lower sleeve (5), a controller and a total control console; the upper sleeve (1) is fixedly connected with the upper film (2) on the side surface in a glove-shaped structure, the upper film (2) is provided with a wire for detecting the resistance value of the piezoresistive material (3), the piezoresistive material (3) can change the resistance value according to the pressure, the piezoresistive material (3) is fixed with the upper film (2) and the lower film (4), the lower film (4) is provided with a wire for detecting the resistance value of the piezoresistive material (3), the lower film (4) is provided with a protrusion at each piezoresistive material position for increasing the detection range; the lower film (4) is fixedly connected with the lower sleeve (5), and the lower sleeve (5) and the upper sleeve (1) are fixedly connected on the side surface; the controller can collect the resistance change of the piezoresistive material (3) to obtain the pressure value.

7. The brain-like intelligence glove according to claim 6, wherein The controller comprises a brain-like chip, a micro single-chip microcomputer and a wireless transmission module, the brain-like chip is used to deploy the brain-like neural network-based force estimation algorithm; the brain-like neural network-based force estimation algorithm outputs a multi-channel pressure sensation sequence of the brain-like force glove through a plurality of piezoresistive materials (3) under different directions and forces, a three-dimensional pressure sensor is used as a correct label during the acquisition process, and a brain-like force glove dataset is constructed; then, the acquired force sensation sequence is converted into a pulse sequence to input a pulse neural network, a brain-like three-dimensional force sensation perception model is trained, and three-dimensional force is obtained by inputting the force sensation sequence of the brain-like force glove into the brain-like three-dimensional force sensation perception model in real time; And the three-dimensional force signal is sent to the total control console through the wireless transmission module.

8. A cerebriform dexterity glove according to claim 7, wherein During human-computer interaction, the total control console acquires the three-dimensional force sensation of the brain-like force glove in real time, converts the force to the working space of each robot for trajectory updating, controls the robot through a speed map-based guidance control method, and realizes human-robot collaborative tasks.

Images