A dynamic multi-objective collaborative optimization method for climbing operation robots

Abstract

The present invention discloses a dynamic multi-objective collaborative optimization method for a climbing working robot, comprising the following steps: determining the optimization variables of the kinematic characteristics and dynamic characteristics of the climbing working robot according to the working process and working indicators of the climbing working robot; obtaining the kinematic equation and dynamic equation of the climbing working robot according to the working process of the climbing working robot; establishing the optimization objective function of maximizing the working efficiency and the optimization objective function of maximizing the stability according to the working requirements of the climbing working robot; determining the constraint conditions of the kinematic characteristics and dynamic characteristics optimization variables by using a mechanism analysis method; obtaining a regression analysis data model of key indicators of the climbing working robot by using a deep data-driven algorithm, and describing the constraint relationship between the optimization variables and uncertainty factors, and obtaining the uncertainty constraint conditions of the optimization variables; determining the global collaborative optimization objective of the system, and establishing the global collaborative optimization model of the system; solving the global collaborative optimization model of the system, and obtaining the optimal solution set of the optimization variables.

Description

Technical Field

[0001] The invention relates to the fields of forestry equipment and artificial intelligence technology, and in particular to a dynamic multi-objective collaborative optimization method for a climbing and pruning robot. Background Art

[0002] my country's forestry covers a large area, with a large amount of work, and the forestry industry is developing rapidly. During the growth of forests, trees need to be pruned reasonably to improve the quality and yield of wood and enhance the economic benefits of forests. At present, forest pruning operations are mainly manual, with high labor intensity, low efficiency and high risk. With the continuous improvement of robot technology, forest pruning operations have put forward new requirements for the mechanization and automation of forestry equipment. Intelligent climbing robots will become an important means of large-scale standing tree pruning operations.

[0003] In recent years, many research institutions have conducted research on tree pruning machines. Beijing Forestry University has designed a spiral-rising tree pruning machine, which uses an elastic clamping mechanism to hold the tree. The transmission mechanism is purely mechanical and the control method is manual remote control. The degree of automation of the whole machine is low; Shandong Forestry University has developed a spiral climbing pruning robot suitable for fast-growing forests, and used PID control to solve the tilt problem of the pruning robot during the climbing process; Guangxi University has designed a climbing tree pruning machine that climbs in a spiral along the tree trunk. For the spiral climbing pruning machine with a nonlinear system with an input dead zone, the RBFNNs neural network is used to approximate the unknown nonlinear function, and the virtual control signal and adaptive control law of the control system are constructed to realize the adaptive control of the robot's tree holding, climbing and tool cutting speed; a control method with an adaptive threshold strategy is proposed based on the event trigger mechanism, which greatly improves the efficiency and energy utilization of the control system, thereby further improving the working efficiency and automation level of tree climbing and pruning equipment.

[0004] At present, the pruning models developed in China are basically in the manual remote control mechanized operation mode, and the degree of automation and intelligence is not high. The pruning robot is a complex system with nonlinearity, strong multivariable coupling and uncertainty. It not only requires good adaptability in the climbing process, but also requires coordinated control and efficient operation of the climbing and pruning processes. However, due to the complexity of the system working process and the unclear mechanism of uncertainty, it is impossible to take into account all the uncertainties of various environments and working conditions when designing the collaborative optimization control algorithm. In the process of the robot's specific execution of tasks, the changes and uncertainties of the model, the coupling between mechanisms, the various frictions of the rotating system, the changes in the external environment and working conditions, and other factors make the global collaborative optimization modeling of the robot very difficult, which is the current bottleneck problem of the climbing autonomous operation robot technology.

[0005] The present invention proposes a dynamic multi-objective collaborative optimization method for a climbing working robot based on uncertainty analysis. With the goal of stable operation and efficient operation of the robot's operating process, the method analyzes the effective coordination mechanism of the mutual coupling between the system characteristics of the climbing and pruning robot with uncertain influencing factors, the operating state and the environmental conditions based on a deep data-driven algorithm, establishes a dynamic multi-objective collaborative optimization model and solution method for the robot under complex environmental conditions, realizes the collaborative optimization control of various mechanisms such as climbing, holding, and pruning during the operating process, improves the overall performance of the climbing and pruning robot, and improves the working efficiency, automation, and intelligence level of tree climbing and pruning equipment. Summary of the invention

[0006] To achieve the above object, the technical solution of the present invention is:

[0007] A dynamic multi-objective collaborative optimization method for a climbing operation robot comprises the following steps:

[0008] Analyze the working principle of the climbing robot. Determine the global optimization goal of the climbing robot, decompose the complex optimization problem of the robot into multiple sub-optimization problems such as climbing, pruning, and holding, explore the uncertain constraint rules between the sub-optimization problems, build a dynamic multi-objective collaborative optimization model for the climbing robot, determine its solution method, and obtain the global optimal operating parameters of the robot during the operation process.

[0009] According to the operation process and operation index of the climbing operation robot, the kinematic and dynamic characteristics optimization variables are determined. The kinematic characteristics optimization variables include the climbing speed V 攀 , climbing acceleration α 攀 , climbing displacement x 攀 、Drive wheel angular velocity ω 轮 and the driving wheel angular acceleration α 轮 The dynamic characteristic optimization variables include the climbing arm clamping force F 夹 , Climbing arm friction F 摩 The output torque T of the driving wheel motor 电机 ; Define optimization variable X p =[X1,X2,X3…X n ] T , where X p is the optimization variable, which is a matrix composed of acceleration, speed, force, power, acceleration influence coefficient, and speed influence coefficient. n ∈[1,n];

[0010] According to the operation process of the climbing robot, its kinematic equation and dynamic equation are obtained;

[0011] According to the operation requirements of the climbing operation robot, the optimization objective function of maximizing the operation efficiency and the optimization objective function of maximizing the stability are established;

[0012] The mechanism analysis method is used to determine the constraints of the kinematic and dynamic characteristics optimization variables, including the tree diameter R, climbing speed V 攀 and clamping force F;

[0013] According to the virtual prototype experimental data, physical prototype test data and actual operation data, with the climbing robot's operating indicators as output and optimization variables and other factors as input, a deep data-driven algorithm is used to obtain the climbing robot's key indicator regression analysis data model, and the constraint relationship between the optimization variables and uncertainty factors is described to obtain the optimization variable uncertainty constraint conditions;

[0014] Taking the high operating efficiency, high stability and strong adaptability of the climbing robot as the global collaborative optimization goal of the system, a global collaborative optimization model of the system was established using the normalization method, multi-objective optimization method and entropy weight objective assignment method.

[0015] The multi-modal adaptive differential evolution algorithm is used to solve the global collaborative optimization model of the system and obtain the optimal solution set of optimization variables.

[0016] The climbing robot is used to complete the tasks of automatic climbing and fixed-point pruning. The kinematic equation is used to analyze the motion of climbing and pruning, which is expressed as follows:

[0017]

[0018] In the formula, H p is the height that the pruning machine climbs, h1 represents the height of the working point, and h0 represents the height of the starting point; L d Indicates the distance that the cutter needs to be adjusted as the tree diameter changes, d0 is the trunk diameter at the starting point, and d is the trunk diameter at the working point; is the rate of reduction of tree diameter per unit height, and d1 is the diameter of a tree at a certain height.

[0019] The dynamic equation is used to define the requirements of the clamping force, the driving force and the driving torque provided by the motor during the climbing process of the climbing robot. The dynamic equation is expressed as follows:

[0020]

[0021] Where F is the clamping force, F X1 is the normal force of the first clamping wheel, F X2 is the normal force of the second clamping wheel; F M The driving force provided by the motor, F g is the normal resistance, G is the weight of the climbing robot, and f j Acceleration resistance; T Mis the driving torque, T g The torque required to overcome rolling resistance, T G The torque required to overcome gravity, T j The torque required to overcome the acceleration resistance.

[0022] The operation efficiency maximization optimization objective function is expressed as follows:

[0023]

[0024] In the formula, f1(x) represents the work efficiency function, x represents the optimization variable, and p i (x) represents the task completion index function, q i (x) represents the single-task power consumption index function, r i (x) represents the collaborative index function, P represents the total task quantization coefficient, Q represents the total task power loss experimental value, R represents the empirical collaborative quantization value, λ1, λ2 and λ3 all represent weight coefficients, N1 represents the number of effective jobs, N2 represents the number of power-consuming tasks, and N3 represents the number of collaborative jobs;

[0025] The stability maximization optimization objective function is expressed as follows:

[0026]

[0027] Where f2(x) represents the stability function, x represents the optimization variable, and T i (x) represents the component motor output torque ripple function, E i (x) represents the operating efficiency fluctuation function, F i (x) represents the climbing driving force function, T s represents the empirical expected value, E s Represents the operating efficiency index, F s represents the climbing control driving force, F R represents the empirical regression value of the climbing driving force, λ4, λ5 and λ5 all represent weight coefficients, N4 represents the number of motors, N5 represents the number of work tasks, and N6 represents the number of driving force sources.

[0028] The constraints of the kinematic and dynamic characteristics optimization variables specifically include:

[0029] The diameter R of the tree must satisfy:

[0030] R min ≤R≤R max ,

[0031] In the formula, R max and R min Respectively represent the maximum radius and minimum radius of the trunk;

[0032] The climbing speed V of the climbing robot must meet the following requirements:

[0033] V min ≤V≤V max ,

[0034] Where V min and V max Respectively represent the minimum speed and maximum speed of the climbing operation robot;

[0035] The clamping force F of the climbing robot push rod must meet the following requirements:

[0036] F min ≤F≤F max ,

[0037] In the formula, F min and F max They represent the minimum clamping force and maximum clamping force of the push rod respectively.

[0038] Before describing the constraint relationship between optimization variables and uncertainty factors, analyze the uncertainty factors in the climbing operation process:

[0039] Based on the robot's working principle, simulation and experimental data, analyze the uncertainty factors and their sources related to the robot's motion trajectory, motion posture, workspace, and operation process;

[0040] Deep data-driven algorithms are used to explore the influencing mechanism of uncertainty factors in the robot's operation process under extreme working conditions, establish the mutual coupling and mutual influence mechanism of the robot's kinematic characteristic parameters, dynamic characteristic parameters and control system parameters, and analyze the impact of environmental conditions changes, parameter uncertainty in the model and time-varying characteristics on the system's operating performance.

[0041] According to the influencing characteristics of uncertainty factors in the robot's operation process under extreme working conditions, high-dimensional space convolution, high-dimensional space mapping, dynamic constraint penetration and feature extraction are combined to construct the specific constraints of the optimization problem. The feature modeling, data evolution rules, operation state mapping and recursive calculation methods are deeply integrated to obtain the numerical range of the optimization variables. The deep kernel extreme learning machine is used as the carrier to construct the key indicator regression analysis data model of the dynamic multi-objective collaborative optimization model of the climbing pruning robot.

[0042] The optimization variable uncertainty constraint is expressed as follows:

[0043]

[0044] Where, X p The optimization variable is a matrix composed of acceleration, velocity, force, power, acceleration influence coefficient, velocity influence coefficient, etc. prepresents the synergy coefficient vector, which is determined by the optimization problem, u toatal Represents the cumulative value vector of the synergy coefficient, represented by u p and the accumulation rule; F() is the synergy coefficient vector feature mapping function, D(X p ) represents the constraint function related to the motion speed, S(X p ) represents the state value calculation function of the optimization variable, W(X p ) is the empirical constraint formula for the optimization variable, H(X p ) is the evaluation function of the empirical constraint function and the synergy coefficient feature, R(X p ) represents the calculation function of the correlation coefficient, α, β, γ are the correlation calculation coefficients, a, b, c are the empirical coefficients, and v, w are the coefficients related to the evaluation system.

[0045] The system global collaborative optimization model is expressed as follows:

[0046]

[0047] Where, X p is the optimization variable, Ω(X p ) represents the optimization function, u p represents the synergy coefficient vector, u toatal represents the cumulative value vector of the synergy coefficient, F() is the synergy coefficient vector feature mapping function, D(X p ) represents the constraint function related to the motion speed, Θ(X p ) represents the function related to the uncertainty characteristics, S(X p ) represents the state value calculation function of the optimization variable, W(X p ) is the empirical constraint formula for the optimization variable, H(X p ) is the evaluation function of the empirical constraint function and the synergy coefficient feature, R(X p ) represents the calculation function of the correlation coefficient, α, β, γ are the correlation calculation coefficients, a, b, c are the empirical coefficients, and v, w are the coefficients related to the evaluation system.

[0048] In the process of solving the multi-objective optimization algorithm, in view of the complexity of the solution process of the multi-objective optimization algorithm, based on the optimization solution that penalizes the violation of the constraint conditions, a cost function for solving the uncertain multi-objective optimization problem is constructed to evaluate the rationality of the solution of the solution process. The cost function is expressed as follows:

[0049]

[0050] In the formula, f b Represents the basic fitness calculation formula, which is constructed by optimizing the objective function. g(X) represents the fitness function, ε is a small constant to avoid the denominator of the calculation process being 0, Z i Indicates the penalty factor for violating the maximum acceleration limit. When the acceleration is greater than the physical limit value, a corresponding penalty coefficient will be assigned. V i Indicates the penalty factor for violating the maximum speed limit. When the running speed is greater than the maximum speed limit, a corresponding penalty coefficient will be assigned. i It indicates the penalty factor for the acceleration change rate violating the acceleration change restriction condition. When the acceleration amplitude change rate is greater than the specified threshold, a corresponding penalty coefficient will be assigned. i It represents the penalty factor for violating the maximum power limit. When the average power is greater than the physical limit value, a corresponding penalty coefficient will be assigned.

[0051] The process of solving the system global collaborative optimization model using the multi-modal adaptive differential evolution algorithm is as follows:

[0052] The differential evolution algorithm scaling factor is reconstructed with the best fitness in the neighborhood, and combined with the crossover operator to mutate the individuals of the differential evolution algorithm.

[0053] An intelligent inference algorithm is used to calculate the degree of coordination of each individual collaborative evaluation index. The optimal solution is searched in a separate direction in each search process according to the size of the degree of coordination. After a certain number of iterations, the optimal solutions of the multi-objective optimization problem are sorted, and finally the solution corresponding to the maximum degree of coordination is selected as the satisfactory solution to this optimization problem.

[0054] The process of obtaining the best solution set of optimization variables includes:

[0055] The optimization problem is to solve the global collaborative optimization model of the system. The satisfactory solution of the optimization problem is:

[0056]

[0057] ψ(X p )=aΦ(u pa )+bF(s,pa)X p +cS(X p ),

[0058] In the formula, ψ(X p ) represents the optimization condition of the optimization problem, u pa The collaborative evaluation index value vector representing the solution set of the optimization problem is determined by the specific optimization problem. pa ) represents the characteristic mapping function from the collaborative evaluation vector to the solution space. The calculation result is a vector or matrix. s represents the quantitative coefficient of the evaluation process. F(s,u pa ) represents the collaborative evaluation index value vector u pa The quantization function, S(X p) represents the regularization function of the solution set of the optimization problem;

[0059] Finally, we get the solution set of the optimization problem, namely:

[0060]

[0061] The beneficial effects of this hair are:

[0062] At present, the research on climbing and pruning robots mainly focuses on the innovation of climbing structure design, kinematic analysis of climbing process, dynamic modeling and path planning, obstacle avoidance control, etc. The theoretical methods for global collaborative optimization and adaptive control of the operation process of climbing and pruning robots to improve the overall coordination performance of the whole machine are still lacking, and the contradiction between high efficiency, stability and adaptability of robot climbing and pruning operations is prominent. At the same time, although there have been many studies on collaborative optimization and adaptive control of industrial robots and some mobile operation robots at home and abroad, due to the different working characteristics, working mechanisms and working environment conditions of each robot, the uncertain influence mechanism of the mutual coupling of internal and external factors of the robot is different, and the variety of problem changes, it is difficult to have a universal dynamic multi-objective optimization and control method that is suitable for solving the multi-objective global optimization and adaptive control problems of various robots.

[0063] The dynamic multi-objective global collaborative optimization method based on uncertainty analysis of the present invention comprehensively considers the mutual coupling mechanism of the mechanical system, kinematic and dynamic system, control system and environmental conditions of the standing tree climbing and pruning robot, as well as the influence of uncertain factors. It can effectively eliminate the internal and external uncertainties of the pruning robot, greatly improve the global coordination performance and control performance of the system, and is of great significance to significantly improving the overall performance of the climbing and pruning robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a flow chart of a dynamic multi-objective collaborative optimization method for a climbing operation robot;

[0065] Figure 2 This is the main structural diagram of the climbing operation robot;

[0066] In the figure: 1-crawler mechanism, 2-climbing motor, 3-ring track motor, 4-ring track mechanism, 5-pitch adjustment mechanism, 6-pitch adjustment motor, 7-pruning mechanism, 8-pruning motor, 9-frame, 10-electric push rod, 11-clamping mechanism;

[0067] Figure 3 It is a flow chart of the model building and solution method of the present invention. DETAILED DESCRIPTION

[0068] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0069] The present invention discloses a dynamic multi-objective collaborative optimization method for a climbing operation robot, the process is as follows: Figure 1 As shown, the following steps are included:

[0070] Analyze the working principle of the climbing robot. Determine the global optimization goal of the climbing robot, decompose the complex optimization problem of the robot into multiple sub-optimization problems such as climbing, pruning, and holding, explore the uncertain constraint rules between the sub-optimization problems, build a dynamic multi-objective collaborative optimization model for the climbing robot, determine its solution method, and obtain the global optimal operating parameters of the robot during the operation process.

[0071] According to the operation process and operation index of the climbing operation robot, the kinematic and dynamic characteristics optimization variables are determined. The kinematic characteristics optimization variables include the climbing speed V 攀 , climbing acceleration α 攀 , climbing displacement x 攀 、Drive wheel angular velocity ω 轮 and the driving wheel angular acceleration α 轮 The dynamic characteristic optimization variables include the climbing arm clamping force F 夹 , Climbing arm friction F 摩 The output torque T of the driving wheel motor 电机 ; Define optimization variable X p =[X1,X2,X3…X n ] T , where X p is the optimization variable, which is a matrix composed of acceleration, speed, force, power, acceleration influence coefficient, and speed influence coefficient. n ∈[1,n];

[0072] According to the operation process of the climbing robot, its kinematic equation and dynamic equation are obtained;

[0073] According to the operation requirements of the climbing operation robot, the optimization objective function of maximizing the operation efficiency and the optimization objective function of maximizing the stability are established;

[0074] The mechanism analysis method is used to determine the constraints of the kinematic and dynamic characteristics optimization variables, including the tree diameter R, climbing speed V 攀 and clamping force F;

[0075] According to the virtual prototype experimental data, physical prototype test data and actual operation data, with the climbing robot's operating indicators as output and optimization variables and other factors as input, a deep data-driven algorithm is used to obtain the climbing robot's key indicator regression analysis data model, and the constraint relationship between the optimization variables and uncertainty factors is described to obtain the optimization variable uncertainty constraint conditions;

[0076] Taking the high operating efficiency, high stability and strong adaptability of the climbing robot as the global collaborative optimization goal of the system, a global collaborative optimization model of the system was established using the normalization method, multi-objective optimization method and entropy weight objective assignment method.

[0077] The multi-modal adaptive differential evolution algorithm is used to solve the global collaborative optimization model of the system and obtain the optimal solution set of optimization variables.

[0078] The specific implementation process is as follows:

[0079] Step 1: Analysis of the working principle of climbing pruning.

[0080] The main devices of the climbing pruning robot include power source, climbing crawler, driving wheel, clamping wheel, clamping mechanism, tool cutting mechanism and control system. The main structure of the machine is as follows Figure 2 As shown in the figure, the global optimization goal of the pruning robot is determined, and the complex optimization problem of the robot is decomposed into multiple sub-optimization problems such as climbing, pruning, and holding. The uncertain constraint rules between the sub-optimization problems are explored, and a dynamic multi-objective collaborative optimization model of the pruning robot and its solution method are constructed to obtain the global optimal operating parameters of the robot during the operation process and improve the comprehensive performance of the robot operation.

[0081] The working process of the climbing pruning robot is as follows:

[0082] The robot's motion joints close and hold the tree trunk tightly, and the clamping wheel clamps the pruning machine to the tree trunk. After the clamping force detected by the pressure sensor reaches the expected value, the pruning machine is in a force balance state;

[0083] The controller issues an expected control command to the drive motor, and the drive motor drives the drive wheel to start climbing up along the tree trunk. The drive motor controls the speed of the drive wheel to control the rising speed of the pruning machine, and at the same time controls the cutter to saw the side branches at a certain speed;

[0084] After the robot reaches the desired height, the clamping device holds the tree trunk tightly, drives the motor to reverse, and smoothly descends to the bottom of the tree at a preset speed. At this point, the pruning operation of a single tree is completed.

[0085] According to the operation process and operation index of the climbing operation robot, the kinematic and dynamic characteristics optimization variables are determined. The kinematic characteristics optimization variables include the climbing speed V 攀 , climbing acceleration α 攀 , climbing displacement x 攀 、Drive wheel angular velocity ω 轮 and the driving wheel angular acceleration α 轮 The dynamic characteristic optimization variables include the climbing arm clamping force F 夹 , Climbing arm friction F 摩The output torque T of the driving wheel motor 电机 ; Define optimization variable X p =[X1,X2,X3…X n ] T , where X p is the optimization variable, which is a matrix composed of acceleration, speed, force, power, acceleration influence coefficient, and speed influence coefficient. n ∈[1,n].

[0086] According to the operation process of the climbing robot, its kinematic equation and dynamic equation are obtained:

[0087] The designed climbing and pruning robot mainly completes the tasks of automatic climbing and fixed-point pruning. The motion of climbing and pruning is analyzed, and the kinematic equation is expressed as follows:

[0088]

[0089] In the formula, H p is the height that the pruning machine climbs, h1 represents the height of the working point, and h0 represents the height of the starting point; L d Indicates the distance that the cutter needs to be adjusted as the tree diameter changes, d0 is the trunk diameter at the starting point, and d is the trunk diameter at the working point; is the rate of reduction of tree diameter per unit height, and d1 is the diameter of a tree at a certain height.

[0090] In order for the climbing robot to climb and work on trees smoothly, stably and safely, the clamping force of the climbing robot during climbing and the driving force and driving torque provided by the motor must meet the requirements. The dynamic equation is expressed as follows:

[0091]

[0092] Where F is the clamping force, F X1 is the normal force of the first clamping wheel, F X2 is the normal force of the second clamping wheel; F M The driving force provided by the motor, F g is the normal resistance, G is the weight of the climbing robot, and f j Acceleration resistance; T M is the driving torque, T g The torque required to overcome rolling resistance, T G The torque required to overcome gravity, T j The torque required to overcome the acceleration resistance.

[0093] According to the operation requirements of the climbing robot, the optimization objective function of maximizing operation efficiency and the optimization objective function of maximizing stability are established:

[0094] The optimization objective function for maximizing operating efficiency is expressed as follows:

[0095]

[0096] In the formula, f1(x) represents the work efficiency function, x represents the optimization variable, and p i (x) represents the task completion index function, q i (x) represents the single-task power consumption index function, r i (x) represents the collaborative index function, P represents the total task quantization coefficient, Q represents the total task power loss experimental value, R represents the empirical collaborative quantization value, λ1, λ2 and λ3 all represent weight coefficients, N1 represents the number of effective jobs, N2 represents the number of power-consuming tasks, and N3 represents the number of collaborative jobs;

[0097] The stability maximization optimization objective function is expressed as follows:

[0098]

[0099] Where f2(x) represents the stability function, x represents the optimization variable, and T i (x) represents the component motor output torque ripple function, E i (x) represents the operating efficiency fluctuation function, F i (x) represents the climbing driving force function, T s represents the empirical expected value, E s Represents the operating efficiency index, F s represents the climbing control driving force, F R represents the empirical regression value of the climbing driving force, λ4, λ5 and λ5 all represent weight coefficients, N4 represents the number of motors, N5 represents the number of work tasks, and N6 represents the number of driving force sources.

[0100] The mechanism analysis method is used to determine the constraints of the kinematic and dynamic characteristics optimization variables, including the tree diameter R, climbing speed V 攀 And clamping force F:

[0101] (1) Tree diameter R: In order to ensure that the climbing robot can cling to the tree trunk well during the climbing process, the trunk diameter must meet the following requirements:

[0102] R min ≤R≤R max

[0103] In the formula, R max and R min Respectively represent the maximum radius and minimum radius of the trunk;

[0104] (2) Climbing speed V: In order to ensure the efficiency of pruning and the stability of the machine, the climbing speed of the climbing robot must meet the following requirements:

[0105] V min ≤V≤V max

[0106] Where V min and V max Respectively represent the minimum speed and maximum speed of the climbing operation robot;

[0107] (3) Clamping force F: In order to ensure that the climbing robot can climb up and down smoothly, the clamping force of the push rod needs to meet the following requirements:

[0108] F min ≤F≤F max

[0109] In the formula, F min and F max They represent the minimum clamping force and maximum clamping force of the push rod respectively.

[0110] Step 2: Method for establishing dynamic multi-objective collaborative optimization model of climbing pruning robot.

[0111] Dynamic multi-objective global collaborative optimization method based on uncertainty analysis, model building and solution methods such as Figure 3 As shown in the figure, the collaborative optimization method comprehensively considers the mutual coupling relationship between the robot's mechanical system, kinematic and dynamic systems, control systems, and environmental conditions, as well as the influence of uncertainty factors. It can effectively eliminate internal and external uncertainties in the robot and greatly improve its global coordination and control performance.

[0112] (1) Analysis of uncertainty factors in climbing operation process

[0113] Based on the working principle of the robot, simulation and experimental data, the uncertainty influencing factors and their sources related to the robot's motion trajectory, motion posture, workspace, climbing and pruning process are analyzed; a deep data-driven algorithm is used to explore the influencing mechanism of uncertainty factors in the robot's operation process under extreme working conditions such as tree trunk bending, tree pits, tree tumors, and branch stubble. The robot's kinematic characteristic parameters such as speed, acceleration, displacement, and driving wheel angular velocity, angular acceleration, and dynamic characteristic parameters such as clamping force, friction, torque, and control system parameters are coupled and influenced by each other, and the influence of environmental conditions changes, parameter uncertainty in the model, and time-varying characteristics on the system's operating performance are analyzed.

[0114] (2) Constructing a regression analysis data model for key indicators of the optimization model based on a deep data-driven algorithm

[0115] According to the influencing characteristics of uncertainty factors in the robot's operation process under extreme working conditions such as trunk bending, tree pits, tree tumors, and branch stumps, high-dimensional space convolution, high-dimensional space mapping, dynamic constraint penetration and feature extraction are combined to construct the specific constraints of the optimization problem. The feature modeling, data evolution laws, operation state mapping and recursive calculation methods are deeply integrated to obtain the numerical range of optimization variables. The deep kernel extreme learning machine is used as the carrier to construct a key indicator regression analysis data model for the dynamic multi-objective collaborative optimization model of the climbing pruning robot.

[0116] The output function expression of the extreme learning machine is:

[0117]

[0118] h i (x) = G(a i ,b i ,x)a i ∈R d ,b i ∈R

[0119] Where, β=[β1,β2,...,β L ] T represents the output layer weight vector, where β i ∈R m ; h(x)=[h1(x),h2(x),...,h L (x)] represents the nonlinear feature map of the extreme learning machine, and a is a i The matrix composed of b is b i A vector composed of j = 1, 2, 3…N, a T a=I,b T b=1.

[0120] According to the kernel method, the kernel function is used to map the features to obtain a deep kernel extreme learning machine, and the output function is as follows:

[0121]

[0122] In the formula, f N (x) represents the output function of the deep kernel extreme learning machine; Ω ELM represents the kernel mapping matrix; I represents the unit diagonal matrix; C represents the penalty coefficient; K(x,x i ),i=1,2,…,n represents kernel function; Y represents sample target.

[0123] The influence of system deterministic parameters and their constraints are analyzed by mechanism analysis. Based on experimental test and virtual simulation data and real-time environmental condition detection data, feature modeling, data evolution law, operation state mapping and recursive calculation method are deeply integrated to explore the uncertain constraint law between each sub-optimization problem. The specific constraint conditions of the optimization problem are constructed by combining high-dimensional space convolution, high-dimensional space mapping, dynamic constraint penetration and feature extraction.

[0124]

[0125] (3) Establishment of a dynamic multi-objective collaborative optimization model for a climbing and pruning robot

[0126] Aiming at the complex optimization problem of pruning robots with strong multi-factor coupling, nonlinearity and uncertainty, the collaborative optimization problem of the whole machine is decomposed into multiple sub-optimization problems such as climbing, pruning and holding. Taking the high efficiency, good stability and strong adaptability of the robot as the global collaborative optimization goal of the system, the mechanism analysis is used to clarify the influence law and constraint restrictions of the system's deterministic parameters; based on experimental test and virtual simulation data and real-time environmental condition detection data, feature modeling, data evolution law, operation state mapping and recursive calculation methods are deeply integrated to explore the uncertain constraint laws between each sub-optimization problem, and the specific constraint conditions of the optimization problem are constructed by combining high-dimensional space convolution, high-dimensional space mapping, dynamic constraint penetration and feature extraction. The dynamic multi-objective collaborative optimization model of the climbing pruning robot is constructed, which is expressed as follows:

[0127]

[0128] Where, X p The optimization variable is a matrix composed of acceleration, velocity, force, power, acceleration influence coefficient, velocity influence coefficient, etc. p ) represents the optimization function; u p represents the synergy coefficient vector, which is determined by the optimization problem; u toatal Represents the cumulative value vector of the synergy coefficient, represented by u p and the accumulation rule; F() is the synergy coefficient vector feature mapping function, D(X p ) represents the constraint function related to the motion speed, Θ(X p ) represents the function related to the uncertainty characteristics, S(X p ) represents the state value calculation function of the optimization variable, W(X p ) is the empirical constraint formula for the optimization variable, H(X p ) is the evaluation function of the empirical constraint function and the synergy coefficient feature, R(X p) represents the calculation function of the correlation coefficient; α, β, γ are the correlation calculation coefficients, a, b, c are the empirical coefficients, and v, w are the coefficients related to the evaluation system.

[0129] Step 3: Establish a robot dynamic multi-objective collaborative optimization solution strategy with complex constraints based on the multimodal adaptive differential evolution algorithm.

[0130] (1) Differential evolution algorithm: The mutant individuals and target individuals are mixed to generate test individuals. The vector individuals in the population are taken as target individuals one by one. New individuals are reproduced through mutation, crossover and selection. The competition mechanism of retaining good individuals and eliminating bad individuals guides the search process to approach the global optimal solution of the optimization problem. The basic operation of solving optimization problems based on differential evolution algorithm is to start with a random population, generate a new generation of individuals of differential evolution algorithm according to mutation mechanism, crossover mechanism and selection mechanism, and repeat the cycle until the solution of the optimization problem is obtained.

[0131] Initialization process: Each individual position information in the population represents a candidate solution to the optimization problem. The position information of the population members needs to be initialized. The initial individual position distribution is expressed as follows:

[0132] x i,j =x j,min +rand i,j [0,1]·(x j,max -x j,min )

[0133] x i =[x i1 ,x i2 ,x i3 ,…,x iD ] T

[0134] In the formula, x i =[x i1 ,x i2 ,x i3 ,…,x iD ] T It indicates that the individual composition of the differential evolution algorithm is a vector, which is a vector containing multiple dimensions. The number of dimensions is the number of optimization variables. D is the dimension of the optimization problem, x j,min and x j,max is the minimum boundary value and maximum boundary value of the solution space in the jth dimension;

[0135] Mutation process: In the differential algorithm, the individuals of the current generation population are taken as the parent individuals. The parent individuals communicate with each other in a certain mutation mode to generate mutant individuals. This process realizes the optimal evolution of the population. The mutation process of the multimodal differential evolution algorithm is expressed as follows:

[0136]

[0137] In the formula, rand(0,1) represents a random number ranging from 0 to 1, CM represents the probability of individual random retention in the mutation process, SC represents the scaling factor, θ represents the adaptive mutation factor of the multimodal adaptive algorithm, and F best (k) is the optimal fitness of the current era, F best (k-1) is the optimal fitness of the previous generation;

[0138] Crossover process: To further enhance the diversity of the population, after performing mutation operations on the individuals in the population, the position information of the individuals in the population is crossed one by one. The crossover process of the multimodal adaptive differential evolution algorithm is expressed as follows:

[0139]

[0140] In the formula, CR represents the probability that the current individual’s dimension data is replaced by the corresponding dimension data of the global optimal individual, x ij represents the current individual, x bestj represents the global optimal individual.

[0141] (2) Cost function construction: In view of the complexity of the solution process of the multi-objective optimization algorithm, a cost function for solving uncertain multi-objective optimization problems is constructed based on the optimization solution that penalizes violations of constraints, and the rationality of the solution process is evaluated to solve comprehensive optimization problems that only have optimization forms but no specific mathematical analysis. The cost function expression is expressed as follows:

[0142]

[0143] In the formula, f b Represents the basic fitness calculation formula, which is constructed by optimizing the objective function. g (X) represents the fitness function, ε is a small constant to avoid the denominator of the calculation process being 0, Z i Indicates the penalty factor for violating the maximum acceleration limit. When the acceleration is greater than the physical limit value, a corresponding penalty coefficient will be assigned. V i Indicates the penalty factor for violating the maximum speed limit. When the running speed is greater than the maximum speed limit, a corresponding penalty coefficient will be assigned. i It indicates the penalty factor for the acceleration change rate violating the acceleration change restriction condition. When the acceleration amplitude change rate is greater than the specified threshold, a corresponding penalty coefficient will be assigned. i It indicates the penalty factor for violating the maximum power limit. When the average power is greater than the physical limit, a corresponding penalty coefficient will be assigned.

[0144] (3) The scaling factor of the differential evolution algorithm is reconstructed with the optimal fitness of the neighborhood, and it is combined with the crossover operator to mutate the individuals of the differential evolution algorithm. In the mutation process, in addition to the mutation adaptive capability, the local and global search directions are taken into account, which effectively improves the solution accuracy, solution diversity and solution convergence speed of the differential evolution algorithm.

[0145] (4) Since the various objective functions are interconnected through decision variables and the topological structure is very complex, the optimization result of the multi-objective optimization problem is a non-dominated solution set. In order to obtain a satisfactory solution to the optimization problem, the intelligent inference algorithm is used to calculate the synergy degree of each individual's collaborative evaluation index. According to the size of the synergy degree, the optimal solution is searched in a separate direction in each search process. After a certain number of iterations, the optimal solutions of the multi-objective optimization problem are sorted, and finally the solution corresponding to the maximum synergy degree is selected as the satisfactory solution to this optimization problem.

[0146] Step 4: Intelligent perception recursive decision maker.

[0147] Since the various objective functions are interconnected through decision variables and the topological structure is very complex, the optimization result of the multi-objective optimization problem is a non-dominated solution set. In order to obtain a satisfactory solution to the optimization problem, the intelligent inference algorithm is used to calculate the synergy degree of each individual's collaborative evaluation index. According to the size of the synergy degree, the optimization solution is searched in a separate direction in each search process. After a certain number of iterations, the optimization solutions of the multi-objective optimization problem are sorted, and finally the solution corresponding to the maximum synergy degree is selected as the satisfactory solution to this optimization problem.

[0148]

[0149] ψ(X p )=aΦ(u pa )+bF(s,pa)X p +cS(X p )

[0150] In the formula, ψ(X p ) represents the optimization condition of the optimization problem, u pa The collaborative evaluation index value vector representing the solution set of the optimization problem is determined by the specific optimization problem. pa ) represents the characteristic mapping function from the collaborative evaluation vector to the solution space. The calculation result is a vector or matrix. s represents the quantitative coefficient of the evaluation process. F(s,u pa ) represents the collaborative evaluation index value vector u pa The quantization function, S(X p ) represents the regularization function of the solution set of the optimization problem.

[0151] Finally, we get the solution set of the optimization problem, namely:

[0152]

Claims

1. A dynamic multi-objective collaborative optimization method for a climbing robot, characterized in that: The following steps are involved: According to the operation process and operation index of the climbing operation robot, the kinematic and dynamic characteristics optimization variables are determined. The kinematic characteristics optimization variables include the climbing speed V 攀 , climbing acceleration α 攀 , climbing displacement x 攀 、Drive wheel angular velocity ω 轮 and the driving wheel angular acceleration α 轮 The dynamic characteristic optimization variables include the climbing arm clamping force F 夹 , Climbing arm friction F 摩 The output torque T of the driving wheel motor 电机 ; Define optimization variable X p =[X1,X2,X3…X n ] T , where X p is the optimization variable, which is a matrix composed of acceleration, speed, force, power, acceleration influence coefficient, and speed influence coefficient. n ∈[1,n]; According to the operation process of the climbing robot, its kinematic equation and dynamic equation are obtained; According to the operation requirements of the climbing operation robot, the optimization objective function of maximizing the operation efficiency and the optimization objective function of maximizing the stability are established; The mechanism analysis method is used to determine the constraints of the kinematic and dynamic characteristics optimization variables, including the tree diameter R, climbing speed V 攀 and clamping force F; According to the virtual prototype experimental data, physical prototype test data and actual operation data, with the climbing robot's operating indicators as output and optimization variables and other factors as input, a deep data-driven algorithm is used to obtain the climbing robot's key indicator regression analysis data model, and the constraint relationship between the optimization variables and uncertainty factors is described to obtain the optimization variable uncertainty constraint conditions; Taking the high operating efficiency, high stability and strong adaptability of the climbing robot as the global collaborative optimization goal of the system, a global collaborative optimization model of the system was established using the normalization method, multi-objective optimization method and entropy weight objective assignment method. The multi-modal adaptive differential evolution algorithm is used to solve the global collaborative optimization model of the system and obtain the optimal solution set of optimization variables.

2. The dynamic multi-objective collaborative optimization method of a climbing operation robot according to claim 1, characterized in that: The climbing robot is used to complete the tasks of automatic climbing and fixed-point pruning. The kinematic equation is used to analyze the motion of climbing and pruning, which is expressed as follows: In the formula, H p is the height that the pruning machine climbs, h1 represents the height of the working point, and h0 represents the height of the starting point; L d Indicates the distance that the cutter needs to be adjusted as the tree diameter changes, d0 is the trunk diameter at the starting point, and d is the trunk diameter at the working point; is the rate of reduction of tree diameter per unit height, and d1 is the diameter of a tree at a certain height.

3. The dynamic multi-objective collaborative optimization method of a climbing operation robot according to claim 1, characterized in that: The dynamic equation is used to define the requirements of the clamping force, the driving force and the driving torque provided by the motor during the climbing process of the climbing robot. The dynamic equation is expressed as follows: Where F is the clamping force, F X1 is the normal force of the first clamping wheel, F X2 is the normal force of the second clamping wheel; F M The driving force provided by the motor, F g is the normal resistance, G is the weight of the climbing robot, and f j Acceleration resistance; T M is the driving torque, T g The torque required to overcome rolling resistance, T G The torque required to overcome gravity, T j The torque required to overcome the acceleration resistance.

4. The dynamic multi-objective collaborative optimization method of a climbing operation robot according to claim 1, characterized in that: The operation efficiency maximization optimization objective function is expressed as follows: In the formula, f1(x) represents the work efficiency function, x represents the optimization variable, and p i (x) represents the task completion index function, q i (x) represents the single-task power consumption index function, r i (x) represents the collaborative index function, P represents the total task quantization coefficient, Q represents the total task power loss experimental value, R represents the empirical collaborative quantization value, λ1, λ2 and λ3 all represent weight coefficients, N1 represents the number of effective jobs, N2 represents the number of power-consuming tasks, and N3 represents the number of collaborative jobs; The stability maximization optimization objective function is expressed as follows: Where f2(x) represents the stability function, x represents the optimization variable, and T i (x) represents the component motor output torque ripple function, E i (x) represents the operating efficiency fluctuation function, F i (x) represents the climbing driving force function, T s represents the empirical expected value, E s Represents the operating efficiency index, F s represents the climbing control driving force, F R represents the empirical regression value of the climbing driving force, λ4, λ5 and λ5 all represent weight coefficients, N4 represents the number of motors, N5 represents the number of work tasks, and N6 represents the number of driving force sources.

5. The dynamic multi-objective collaborative optimization method of a climbing operation robot according to claim 1, characterized in that: The constraints of the kinematic and dynamic characteristics optimization variables specifically include: The diameter R of the tree must satisfy: R min ≤R≤R max , In the formula, R max and R min Respectively represent the maximum radius and minimum radius of the trunk; The climbing speed V of the climbing robot must meet the following requirements: In min ≤V≤V max , Where V min and V max Respectively represent the minimum speed and maximum speed of the climbing robot; The clamping force F of the climbing robot push rod must meet the following requirements: F min ≤F≤F max , In the formula, F min and F max They represent the minimum clamping force and maximum clamping force of the push rod respectively.

6. The dynamic multi-objective collaborative optimization method of a climbing operation robot according to claim 1, characterized in that: The optimization variable uncertainty constraint is expressed as follows: In the formula, X p is the optimization variable, which is a matrix composed of acceleration, speed, force, power, acceleration influence coefficient, and speed influence coefficient; u p represents the synergy coefficient vector, which is determined by the optimization problem, u toatal Represents the cumulative value vector of the synergy coefficient, represented by u p Determined by the accumulation rule; F() is the eigenvalue mapping function of the synergy coefficient vector, D(X p ) represents the constraint function related to the motion speed, S(X p ) represents the state value calculation function of the optimization variable, W(X p ) is the empirical constraint formula for the optimization variable, H(X p ) is the evaluation function of the empirical constraint function and the synergy coefficient feature, R(X p ) represents the calculation function of the correlation coefficient, α, β, γ are the correlation calculation coefficients, a, b, c are the empirical coefficients, and v, w are the coefficients related to the evaluation system.

7. The dynamic multi-objective collaborative optimization method of a climbing operation robot according to claim 6, characterized in that: The system global collaborative optimization model is expressed as follows: In the formula, X p is the optimization variable, Ω(X p ) represents the optimization function, u p represents the synergy coefficient vector, u toatal represents the cumulative value vector of the synergy coefficient, F() is the synergy coefficient vector feature mapping function, D(X p ) represents the constraint function related to the motion speed, Θ(X p ) represents the function related to the uncertainty characteristics, S(X p ) represents the state value calculation function of the optimization variable, W(X p ) is the empirical constraint formula for the optimization variable, H(X p ) is the evaluation function of the empirical constraint function and the synergy coefficient feature, R(X p ) represents the calculation function of the correlation coefficient, α, β, γ are the correlation calculation coefficients, a, b, c are the empirical coefficients, and v, w are the coefficients related to the evaluation system.

8. The dynamic multi-objective collaborative optimization method of a climbing operation robot according to claim 7, characterized in that: In the process of solving the multi-objective optimization algorithm, based on the optimization solution that penalizes the violation of the constraint conditions, a cost function for solving the uncertain multi-objective optimization problem is constructed to evaluate the rationality of the solution in the solution process. The cost function is expressed as follows: In the formula, f b Represents the basic fitness calculation formula, which is constructed by optimizing the objective function. g (X) represents the fitness function, ε is a small constant to avoid the denominator of the calculation process being 0, Z i Indicates the penalty factor for violating the maximum acceleration limit. When the acceleration is greater than the physical limit value, a corresponding penalty coefficient will be assigned. V i Indicates the penalty factor for violating the maximum speed limit. When the running speed is greater than the maximum speed limit, a corresponding penalty coefficient will be assigned. i It indicates the penalty factor for the acceleration change rate violating the acceleration change restriction condition. When the acceleration amplitude change rate is greater than the specified threshold, a corresponding penalty coefficient will be assigned. i It represents the penalty factor for violating the maximum power limit. When the average power is greater than the physical limit value, a corresponding penalty coefficient will be assigned.

9. The dynamic multi-objective collaborative optimization method of a climbing operation robot according to claim 8, characterized in that: The process of obtaining the best solution set of optimization variables includes: The optimization problem is to solve the global collaborative optimization model of the system. The satisfactory solution of the optimization problem is: ψ(X p )=aΦ(u pa )+bF(s,pa)X p +cS(X p ), In the formula, ψ(X p ) represents the optimization condition of the optimization problem, u pa The collaborative evaluation index value vector representing the solution set of the optimization problem is determined by the specific optimization problem. pa ) represents the characteristic mapping function from the collaborative evaluation vector to the solution space. The calculation result is a vector or matrix. s represents the quantitative coefficient of the evaluation process. F(s,u pa ) represents the collaborative evaluation index value vector u pa The quantization function, S(X p ) represents the regularization function of the solution set of the optimization problem; Finally, the optimal solution set of the optimization problem is obtained, namely:

Images