KR20090025074A - A method for design humanoid robot parts and device - Google Patents

Abstract

A method for designing the parts of a humanoid robot and a design apparatus thereof are provided to design the parts systematically and efficiently in consideration of the internal dynamic load which is generated due to various operations. The rigid multi-body dynamics is interpreted(S201). By using the interpreted result, the linear dynamic analysis(for example, the transient-response analysis) is performed(S202). Equivalent static loads are calculated by using the displacement field obtained from the dynamic analysis(S203). The linear static analysis is performed by using the equivalence static load (S204). By using the result obtained through the linear static analysis, the linear optimal design is performed (S205). The size, shape and property of each part are updated through the result obtained from the linear optimal design(S206).

Description

A design method of a humanoid robot component and its design apparatus {A METHOD FOR DESIGN HUMANOID ROBOT PARTS AND DEVICE}

The present invention relates to a method for designing a robot, and more particularly, to a method for designing each part in consideration of the dynamic effects occurring during the operation of a humanoid robot.

Humanoid robots aimed at providing useful services have recently been developed and released.

In order for humanoid robots to provide useful services to humans, humanoid robots must be able to operate like humans. When a humanoid robot behaves like a human, dynamic loads are constantly generated on each part. In this case, damage to the parts due to dynamic load or severe damage may occur. Therefore, the design should take into account the dynamic load.

Feyerabend (F. Feyerabend, "Systematic optimization of a robot arm structure," The Industrial Robot, Vol. 15, No. 4, 219-222, 1998) is the static that is expected to generate the greatest loads in the design of robot parts. The robot parts were designed in consideration of the phosphorus state.

In addition, Saravanos (DA Saravanos and JS Lamancusa, "Optimum structural design of robotic manipulators with fiber reinforced composite materials," Computers and Structures, Vol. 36, No. 1, pp. 119-132, 1990) for the design of robots The robot parts were designed on the basis of several operating states deemed important and assuming that a static load acts on each state.

However, the above-mentioned studies are designed by considering only the static state of the robot, selecting some representative motions that are considered the most vulnerable among the overall motions of the robot, and considering only the static characteristics of each selected representative motion. Perform the design. However, the method of performing the design in consideration of the static characteristics as described above has a disadvantage in that the effects of the vibration characteristics and the inertia of the component cannot be considered. In addition, the above studies are mainly aimed at structures with small movements, and have not been studied for the design of large movements and repetitive mechanisms like actual mechanical devices.

On the other hand, the above studies are mainly designed using the finite element method and the optimal design method using computer simulation. However, since these designs have been developed assuming static systems, they have been developed to use a large number of static finite element analyzes (also referred to as linear static analyzes). However, dynamic finite element analysis (hereinafter, also referred to as linear dynamic analysis) for dynamic systems such as robots is very expensive and requires a lot of time, which makes it difficult to apply existing optimal design algorithms. In addition, the dynamic optimization design based on transient response analysis is a difficult method to apply to real problems. Both techniques are inadequate for the design of complex, highly dynamic structures such as humanoid components.

Therefore, there is a need for an optimal design method of a robot component that overcomes the disadvantages of these existing design methods and considers dynamic and inertial effects.

For the design of humanoid robot parts, it is necessary to develop a practical design method considering dynamic loads.

Accordingly, an object of the present invention is to develop a method for systematically and efficiently designing a part of a humanoid robot in which dynamic loads of complex shapes are generated due to various operations in consideration of dynamic loads.

In order to achieve the above object, the present invention uses the multibody dynamics technique to grasp the dynamic characteristics of each part of the humanoid, and combine the equivalent static load method to efficiently and systematically design the humanoid part . Here, the equivalent static load refers to a static load that allows a linear static analysis to obtain a response value equal to a response value such as displacement or stress obtained from a linear dynamic analysis (or transient response analysis) using the dynamic load. In other words, the result of the linear dynamic analysis using the dynamic load is the same as the result of the linear static analysis using the equivalent static load.

Specifically, the present invention extracts the "dynamic load" generated in each part of the humanoid by using a multibody dynamics technique, and converts the dynamic load into "equivalent static load" which exhibits the same characteristics as in the dynamic state. Optimal design technique is applied to design parts of humanoid parts.

More specifically, the present invention comprises the steps of: a) calculating the dynamic loads generated in each part of the robot; b) calculating dynamic characteristics of the part using the obtained dynamic load; and c) determining the overall shape design and detailed dimensions of each part by using the obtained dynamic load.

In addition, the present invention comprises the steps of: a) performing a multi-body dynamics analysis on the humanoid robot; b) performing a linear dynamic analysis using the results of the multibody dynamics analysis; c) calculating an equivalent static load using the results of the linear dynamic analysis; d) performing a linear static analysis using the calculated equivalent static load; e) by using the results of the linear static analysis, to provide a component design method of the humanoid robot comprising the step of performing an optimal design for each component of the humanoid robot.

On the other hand, the present invention and the storage device for storing the design variables of the humanoid robot; After loading the design variables in the storage device, the dynamic load generated in each part of the robot is calculated, the dynamic characteristics of the part are calculated using the obtained dynamic load, and then the overall dynamics of each part are obtained using the obtained dynamic load. Also provided is a design device comprising a processor for calculating shape and detailed dimensions.

As described above, the present invention allows the design of a robust and economical component in the actual humanoid operating state by considering the dynamic effect in the design of the humanoid component.

Hereinafter, with reference to the accompanying drawings, the present invention will be described in detail.

1 is a flowchart illustrating an optimal design method of a dynamic system using equivalent static loads.

Equivalent static load means a static load in a linear static analysis that generates a response value equal to a response value such as displacement or stress generated from the linear dynamic analysis (or transient response analysis). The equivalent static load in a linear dynamic system is calculated by multiplying the linear stiffness matrix of the finite element model by the displacement field obtained from the results of the dynamic analysis.

1) First, in order to calculate an equivalent static load, a linear dynamic analysis is performed (S101). Here, the linear dynamic analysis is an analysis process for determining the structural characteristics of the structure generated by dynamic loads (loads whose size and direction change with time) are applied to a specific structure. All loads acting on each part of the humanoid are dynamic loads. Therefore, the structural characteristics of each part of humanoid can be obtained through linear dynamic analysis using dynamic load. As a result of the linear dynamic analysis, various characteristics such as the degree of deformation of each part, the strain rate, and the magnitude of the stress can be obtained.

The equation shown is the equation of motion of the structure for linear dynamic analysis. Where M is the mass matrix, z is the displacement vector by dynamic load, K is the stiffness matrix, and f is the dynamic load for dynamic analysis. From the dynamic analysis we can obtain z (t), the displacement vector at all time steps.

는 모든 시 절점에서의 변위장을 일치시키기 위한 정적 해석을 위한 하중이다. 만약 시간 t를 n계의 시간 단계로 나누어 동적 해석을 수행하였다면 등가 정하중은 총 n개의 하중 집합을 가지게 되는 것이다. 2) Next, an equivalent static load is calculated (S102). Here, the equivalent static load means a set of static loads such that structural characteristics identical to those exhibited when the structure is subjected to dynamic loads are expressed. The equivalent static load can be calculated by multiplying the dynamic displacement field z at any time by the linear stiffness matrix K. The linear stiffness matrix K can be obtained from the finite element analyzer. So obtained

Is the load for static analysis to match the displacement field at all time nodes. If the dynamic analysis is performed by dividing time t into n time steps, the equivalent static load has a total of n load sets.

3) Next, a linear static analysis is performed using the equivalent static load (S103).

Previously, when designing a structure with dynamic loads, the design was performed by performing a linear static analysis using static loads. In this case, the static load is expressed as the maximum value or the average value of the dynamic load acting on the humanoid. In the case of designing using such static loads, there are disadvantages in that the characteristics in the dynamic state cannot be accurately known, and various characteristics generated by the dynamic effects cannot be considered at all. On the other hand, the structure can be designed using an optimal design technique using linear dynamic analysis using the dynamic load directly. However, the optimal design using linear dynamic analysis has various problems that are very difficult to apply to the real problem.

Accordingly, the present invention performs linear static analysis using equivalent static loads to overcome these disadvantages. In the case of linear static analysis using equivalent static load, a relatively large load condition is applied to one analysis, which is very short and expensive compared to the linear dynamic analysis because it is analyzed by a technique called multiple load condition. Has

The equation shown is a finite element equation for static analysis. Equivalent static load acts as an external force and what is calculated from this analysis is a linear displacement vector. This linear displacement vector exactly matches the dynamic displacement vector at all time nodes. Therefore, it can be seen that a linear static analysis using an equivalent static load yields the same displacement field as the dynamic displacement field.

4) Next, linear optimization design is performed using the equivalent static load (S104). Linear optimal design refers to the process of selecting the size or shape of the part to be changed in the roughly designed structure and determining the size and shape that best fits the purpose of the designer. Linear optimal design is a very useful and efficient design method that is applied when the load applied to the structure is a linear load. However, the linear optimization design technique has a disadvantage in that it cannot be applied when dynamic load is applied. Therefore, the present invention solves this problem by performing linear optimal design using equivalent static loads. Equivalent static loads have the same structural characteristics as in the dynamic load state as described above, so that the linear optimum design using equivalent static loads can achieve the optimal design for the structure under dynamic loads. The efficiency is excellent.

5) Finally, it is determined whether the result of the linear optimal design (S104) satisfies the constraint condition (S105). If the initial design value is changed and the dynamic analysis is performed using only the result of the linear optimal design (S104), the effect of design improvement may occur, but the designer may not satisfy the desired constraint condition initially desired. This is due to the difference between the sensitivity of dynamic analysis and that of static analysis. Sensitivity refers to the rate of change in response to changes in design variables.

Therefore, when the result of the linear optimization design (S104) does not satisfy the constraint condition, in order to overcome the above-described difference in sensitivity, the solution obtained from the linear optimization design (S104) is set as an initial value, and the entire process is performed again. To perform. Through the repetition of this process, the difference in sensitivity is gradually narrowed, and finally, all responses and sensitivity are almost identical to obtain the optimal solution satisfying the designer's intended constraint.

The repetition of the above process ends when the difference between the equivalent static load value in the previous step and the equivalent static load value in the current step is very small.

However, the process described so far is aimed at a structure in which there is almost no movement and deformation. Therefore, in order to design a component having a large movement and a deformation at the same time as a humanoid, performing the method of FIG. 2 is performed. More preferred.

2 is a flow chart showing a method for performing optimal design of a structure of a multibody dynamics system.

As can be seen with reference to Figure 2, the optimal design of the structure of the multi-body dynamics system is subjected to a similar process to the optimal design in consideration of the above-described dynamic characteristics. However, unlike the process of FIG. 1, the multibody dynamics analysis is preceded before the linear dynamic analysis. This is because the load acting on the structure must be known for the molding dynamic analysis.

1) First, a rigid multibody dynamics is analyzed (S201). Here, the multi-body dynamics analysis is an analysis method for grasping the characteristics of the mechanism formed by connecting a plurality of members and joints. At this time, the case where each member consists of rigid bodies is called rigid multibody dynamics analysis. Each member of the rigid multibody dynamics is characterized by the fact that no deformation occurs because all members are rigid. The rigid multibody dynamics system generally moves dynamically under initial conditions with a number of members connected and constrained by a number of joints, but no deformation occurs in each member.

A simple example is a car wiper, which consists of a glass wiping rubber, a plate holding the rubber, and a connecting rod connected to a motor driving the wiper. Plate, rubber and connecting rod are all connected through the joint. Rubber and plate are connected by sliding sliding joints, and connecting rod and plate are connected by rotatable joints. These are multibody dynamics systems, designed by the input by the motor, to perform the work properly under the confinement of the joint. Like the wiper, the humanoid robot is a very complex rigid multibody dynamics system consisting of various types of members and joints.

Information such as the magnitude and direction of the load generated when the rigid multibody dynamics system is operating can be obtained through the rigid multibody dynamics analysis. In other words, when the rigid multibody dynamics analysis is performed, the load generated in each joint, the change in position caused by the motion of each member, the speed, the acceleration, and the like can be obtained. More specifically, when the rigid multibody dynamics analysis is performed, various structural information such as the geometry and characteristics of the structure, the position or velocity applied to the structure, and the acceleration may be obtained. Therefore, in the present invention, in order to design the structure of the humanoid, by performing a multibody dynamics analysis, the load at each joint generated in each structure of the humanoid can be obtained.

2) Next, a linear dynamic analysis (eg, transient response analysis) is performed using the result calculated through the rigid multibody dynamic analysis (S201) (S202). Through the linear dynamic analysis (S202) as described above, the displacement field can be obtained.

3) Next, an equivalent static load is calculated using the displacement field obtained from the dynamic analysis (S203).

4) Next, a linear static analysis is performed using the equivalent static load (S205).

5) And, using the result obtained through the linear static analysis, a linear optimization design is performed (S205).

6) The design variables such as the size, shape, characteristics, etc. of the parts obtained through the linear optimization design is updated (S206).

7-8) Since the design variable has been changed, the linear dynamic analysis is performed again by reflecting the changed design variable (S207), and the equivalent static load is calculated (S208).

9) Then, it is determined whether the first condition is satisfied (S209). Here, the first condition is whether a difference between the previous equivalent static load calculation result (S203) and the current equivalent static load calculation result (S208) is smaller than a predetermined level value. Here, the constant level value is a very small value close to zero.

If the difference between the previous constant static load calculation result (S203) and the current equivalent static load calculation result (S208) is greater than the predetermined level value, the linear optimum design execution process (S205) is returned.

10. However, if the difference between the previous equivalent static load calculation result (S203) and the current equivalent static load calculation result (S208) is smaller than the predetermined level value, the rigid body multibody dynamics analysis is performed once again (S210). .

11) Then, it is determined whether the second condition is satisfied (S211). Here, the second convergence condition is whether the difference between the result of the previous rigid body multibody dynamics analysis (S201) and the current result of the rigid body multibody dynamics analysis (S210) is small. This is because when the structure of the humanoid is changed, the load generated on the structure is also changed.

If the difference between the result of the previous rigid body multi-body dynamics analysis (S201) and the current result of the rigid body multi-body dynamics analysis (S210) is large, the linear dynamic analysis process (S202) is fed back. However, if the difference is small, the process ends.

3 is another flowchart illustrating a method of designing a robot component of the present invention.

As can be seen with reference to FIG. 3, by implementing the method of FIG. 2, the phase of the humanoid component (ie, the schematic shape of the component) is optimized (ie, the phase optimal design). Then, the method of FIG. 2 is again performed as shown, thereby optimizing the size and detailed shape of the humanoid component (ie, shape optimization design).

The design method according to the present invention, which has been described so far, can be applied not only to the design of parts of a humanoid robot, but also to the design of parts of various mechanisms to which dynamic mechanisms are applied in consideration of dynamic effects.

The method according to the invention described thus far can be implemented in software, hardware, or a combination thereof. For example, the method according to the present invention may be stored in a storage medium (eg, internal memory, flash memory, hard disk, etc.) and may be executed by a processor (eg a microprocessor). It can be implemented as codes or instructions within a program.

Specifically, the design variable, design results may be stored in a storage medium, and the method shown in FIGS. 2 and 3 may be performed by a microprocessor.

Example

Hereinafter, with reference to the accompanying drawings an embodiment of a humanoid pelvic part (Pelvis) designed according to the method according to the present invention will be described in detail.

Figure 4 is an exemplary view showing the parts of the humanoid and the robot to be designed in accordance with the present invention.

The humanoid part shown in FIG. 4 is a part connecting the upper body and the lower body called Pelvis. Since the pelvis supports the upper body while walking, the pelvis is subjected to a relatively large load because it simultaneously receives the load generated from both legs. For this reason, in the present invention, Pelvis was selected as the design target. In addition, Pelvis was designed in consideration of the dynamic load generated when walking in a straight line.

The load generated between the foot and the ground and the displacement resulting from the shaking of the upper body are simultaneously transmitted to Pelvis. The load acting on Pelvis can be calculated using the multibody dynamics software using as input the information on the joint angles obtained by the humanoid simulator.

FIG. 5 illustrates an example of obtaining dynamic loads by performing a multibody dynamics analysis of the humanoid component illustrated in FIG. 4, and FIG. 6 illustrates a maximum generated in Pelvis when the dynamic loads illustrated in FIG. Illustrated diagram of load and stress distribution.

As can be seen with reference to FIGS. 5 and 6, multibody dynamics analysis and dynamic analysis were performed to confirm the initial state of Pelvis. The stress values at the stress distribution and the maximum stress are obtained. The maximum stress occurred at 4.4 seconds when the foot of the humanoid changed and it was confirmed that it had a size of 78.4 MPa.

FIG. 7 is an exemplary view showing a design area for a phase optimum design for a part of a humanoid robot, and FIG. 8 is an exemplary view showing a part designed by performing a phase optimum design according to the present invention with respect to the part of FIG. 7.

As can be seen with reference to Figure 7, Type (a) and Type (b) is in the form of a flat plate of a constant thickness, Type (c) is a material in the middle of the plate is determined to be able to reinforce the rigidity Added. The design problem is formulated as follows by applying constraint that Pelvis mass is less than or equal to the current model.

Find optimal topology

to minimize strain energy

subject to mass <massinitial

Referring to FIG. 8, the results obtained through the phase optimization design show that the type (a) has a typical shape in which materials are arranged at the portion where the upper and lower bodies are connected. In the case of type (b), the material behind the mounting part, and in the case of type (c), the material of the newly added bottom surface affects the rigidity. The mass of each model is equal to the current model mass.

FIG. 9 is an exemplary diagram illustrating a process of preparing a reference form for performing a shape optimization design using a schematic shape of a component on which the phase optimization design of FIG. 8 is performed, and FIG. Exemplary diagram showing a part designed by performing a shape optimization design according to the present invention. And, Figure 11 is an exemplary view showing the difference between the part designed according to the present invention and the part before it is designed according to the present invention.

As can be seen with reference to Figure 9, using the reference shape of Figure 8 obtained through the phase optimization design, before performing the shape optimization design to perform the shape optimization design, the intermediate process, that is, the weight of the Pelvis While minimizing, the optimal design problem was formulated as follows to ensure that the maximum stress is below 50 MPa, which is considered to be a sufficiently stable region.

Find optimal shape

to minimize mass

subject to

Based on the formulation, the shape optimization design for selecting actual shape and dimension of the actual product for the part is performed.

Referring to Figure 10, it can be seen that the difference between the formation of the part before it is designed according to the present invention and the shape of the Pelvis designed according to the present invention.

Referring to FIG. 11, the results obtained through the method of the present invention show that the weight is reduced by about 6-20% and the maximum stress is reduced by about 35-41%. Through this, it can be confirmed that the method according to the present invention can systematically and efficiently design humanoid components in a dynamic situation.

In the above described exemplary embodiments of the present invention by way of example, but the scope of the present invention is not limited only to these specific embodiments, the present invention in various forms within the scope of the spirit and claims of the present invention Can be modified, changed, or improved.

1 is a flowchart illustrating a method for optimal design of a dynamic structure using an equivalent static load of the present invention.

2 is a flowchart illustrating a method of designing a robot component of the present invention.

3 is another flowchart illustrating a method of designing a robot component of the present invention.

Figure 4 is an exemplary view showing the parts of the humanoid and the robot to be designed in accordance with the present invention.

5 is an example in which the dynamic load is obtained by performing the multibody dynamics analysis of the humanoid component shown in FIG. 4.

FIG. 6 is an exemplary view showing the dynamic load shown in FIG. 5 as the maximum load and the stress distribution. FIG.

7 is an exemplary view showing a design area for a part of the humanoid robot.

FIG. 8 is an exemplary view illustrating a component designed by performing a phase optimization design according to the present invention with respect to the component of FIG. 7.

FIG. 9 is an exemplary diagram illustrating a process of preparing a component in which the phase optimization design of FIG. 8 is performed to perform a shape optimization design.

FIG. 10 is an exemplary view showing a difference between a part designed by performing a shape optimization design according to the present invention with respect to the part shown in FIG. 8 and a part before being designed according to the present invention.

11 is an exemplary view showing a difference in structural characteristics between a part before it is designed according to the present invention and a part designed according to the present invention.

Claims (12)

a) calculating dynamic loads generated at each part of the robot; b) calculating dynamic characteristics of the part using the obtained dynamic load; and c) determining the overall shape design and detailed dimensions of each part by using the obtained dynamic load. The method of claim 1, The dynamic load is a component design method of the humanoid robot, characterized in that calculated through the multi-body dynamics analysis. The method of claim 1, wherein b) Performing a linear dynamic analysis using the dynamic load; Calculating an equivalent static load using the result of the linear dynamic analysis; And performing a linear static analysis using the calculated equivalent static load. a) performing a multibody dynamics analysis on the humanoid robot; b) performing a linear dynamic analysis using the results of the multibody dynamics analysis; c) calculating an equivalent static load using the results of the linear dynamic analysis; d) performing a linear static analysis using the calculated equivalent static load; and e) using the result of the linear static analysis, to perform an optimal design for each part of the humanoid robot. The method of claim 4, wherein the multibody dynamics analysis step The component design method of the humanoid robot, characterized in that it is possible to obtain the force, displacement, speed, acceleration, load conditions generated in each component of the humanoid robot. The method of claim 4, wherein f) updating a design variable based on a result of performing the optimal design; g) performing a linear dynamic analysis of the humanoid robot based on the updated design parameters; h) calculating an equivalent static load using the results of the linear dynamic analysis; i) determining whether the difference between the equivalent static load calculation result in step c) and the equivalent static load calculation result in step h) is equal to or less than a first predetermined value; Here, if the difference is greater than the predetermined first value, e) the component design method of the humanoid robot, characterized in that feedback to the step of performing the optimal design. The method of claim 6, If the difference is less than or equal to the predetermined first value, performing a rigid multibody dynamics analysis; Determining whether a difference between the result of the rigid multibody dynamics analysis and the result of the a) rigid multibody dynamics analysis is equal to or less than a second predetermined value; Here, the component design method of the humanoid robot, characterized in that the design is terminated when the difference is less than the second value. The method of claim 7, wherein If the difference is greater than the second value, b) the part design method of the humanoid robot, characterized in that the tracking by the linear dynamic analysis step. Component of a humanoid robot, characterized in that it is designed according to the method of any one of claims 4 to 8. A storage device for storing design variables of the humanoid robot; After loading the design variables in the storage device, the dynamic load generated in each part of the robot is calculated, the dynamic characteristics of the part are calculated using the obtained dynamic load, and then the overall dynamics of each part are obtained using the obtained dynamic load. And a processor for calculating shapes and detailed dimensions. 11. The design apparatus of claim 10, wherein the dynamic load is calculated through multibody dynamics analysis. The method of claim 10, wherein the dynamic characteristics Calculated by the processor performing a linear dynamic analysis using the dynamic load, calculating an equivalent static load using the result of the linear dynamic analysis, and performing a linear static analysis using the calculated equivalent static load. Design device characterized in that.

Images