FOOTHOLD POSITION CONTROL SYSTEM AND METHOD FOR BIPED ROBOT

Abstract

A foothold position control system and method for a biped robot are provided. 1) A feasible collision-free path is planned by using a path planning algorithm; 2) an available foothold area of a swing foot is determined according to step-length constraints, movement capabilities, foot sizes, and center offsets of a biped robot; and 3) fuzzy processing is performed to determine a specific foothold position of the biped robot. Selection of suitable foothold positions on both sides of a path when a biped robot executes specific walking actions after finishing path planning is realized. The foothold position control system and method has the advantages of being simple and easy to implement, having low computational load and high speed, being capable of exerting extreme movement capabilities of different biped robots, enabling more flexible movement of the biped robots, and so on.

Description

TECHNICAL FIELD

The present disclosure belongs to the technical field of biped robots, and specifically, relates to a foothold position control system and method for a biped robot.

BACKGROUND

With the development of science and technology, there are more and more kinds of robots which have been more and more widely used. A biped robot is a comprehensive electromechanical integration platform integrating multiple disciplines such as mechanics, sensing, electronics, control, artificial intelligence, and bionics. The biped robot has a humanoid structure and is easy to implement most human actions, and is even more suitable for accomplishing some tasks in the areas of disaster relief, family services, aerospace, and so on. Although the biped robot has achieved remarkable results in walking, running, jumping, and so on, and has made great progress in straight walking, balance adjustment, modular management, and other control aspects, there is still a big gap in intelligent and autonomous movement compared with other types of robots or the human beings the biped robot imitates. Especially when a path from a start location to a task location is known in the environment, the selection of a foothold position after the biped robot specifically takes a step is particularly important and directly affects the performance of the robot such as the capability of following the path, flexibility, walking balance, and stability.

In the prior art, 10 to 50 available foothold positions of a swing foot with respect to a support foot are established in advance by permutation and combination according to structural constraints and movement capabilities of the biped robot, and are combined into one to three sets. In this way, other steps the biped robot may take at a certain time and under certain circumstances are severely restricted in the x-coordinate direction, y-coordinate direction, and angular direction, and the foothold position is limited to these fixed options, which is unfavorable for flexible movement of the biped robot. Besides, the non-autonomous and non-intelligent characteristics of the biped robot have been decided in selecting the foothold position from the fixed sets, and the foothold position is actually determined as soon as a planned path from a start position to a target position is determined in some methods, in which there is no process of selecting variable foothold positions.

SUMMARY

In view of the deficiencies in the prior art, the present disclosure provides a foothold position control system and method for a biped robot, where after path planning, fuzzy processing is performed to determine a specific foothold position of a robot, and the foothold position is variable, thereby ensuring flexibility and stability when the robot lands.

The present disclosure achieves the aforementioned technical objective by the following technical means.

A foothold position control system for a biped robot includes a laser radar or vision sensor, a gyroscope, a six-component force sensor, and an industrial computer, where a fuzzy controller in the industrial computer separately uses x_sd, x_en, y_sd, y_en, θ_sd, and θ_pt as inputs and performs fuzzy processing to obtain a specific foothold position of a swing foot in an available foothold area, where x_sd, y_sd is a real-time position of a support foot of a biped robot obtained by the six-component force sensor, x_en, y_en is a position of an obstacle in a robot body coordinate system obtained by the laser radar or vision sensor, θ_sd is an initial yaw angle of the biped robot obtained by the gyroscope, and θ_pt is an angle between a straight line formed by a current path node and an associated path node in a next step and an x axis of a world coordinate system.

In the aforementioned technical solution, the

${\theta }_{pt}={\tan }^{-1}\frac{y_{pt+1}-y_{pt}}{x_{pt+1}-x_{pt}},\left(x_{pt+1},y_{pt+1}\right)=\underset{1\le i\le n}{\min }\sqrt{{\left(x_{\mathrm{next}}-x_i\right)}^2+{\left(y_{\mathrm{next}}-y_i\right)}^2};$

where x_pt, y_pt is a path node closest to a position of a current support foot, x_pt+1, y_pt+1 is a path node closest to a foothold position after a current swing foot lands, x_next, y_next is a specific foothold position of the swing foot in an available foothold range, and x_i, y_i is a path node.

A foothold position control method for a biped robot includes planning a passing path; determining an available foothold area of a swing foot on the passing path according to step-length constraints, movement capabilities, foot sizes, and center offsets of a biped robot; and performing fuzzy processing to determine a specific foothold position of the robot.

Further, the fuzzy processing includes fuzzification, fuzzy rule reasoning, and defuzzification.

Furthermore, a process of fuzzy processing on x_en and x_sd is: inputting x_pt−x_pt−1 and x_en, and outputting Δx_next;

(1) fuzzification

fuzzifying x_pt−x_pt−1 as backward, small step forward, medium step forward, and big step forward, with corresponding ranges being respectively

$\left[-D_{\min },0\right],\left[0,\frac{1}{2}{D}_{\max }\right],\left[\frac{1}{3}{D}_{\max },\frac{2}{3}{D}_{\max }\right],\mathrm{and}\left[\frac{1}{2}{D}_{\max },D_{\max }\right];$

fuzzifying x_en as very near, near, medium, far, and very far, with corresponding ranges being respectively

$\left[\frac{1}{2}{D}_{\max },D_{\max }\right],\left[\frac{3}{4}{D}_{\max },2D_{\max }\right],\left[\frac{3}{2}{D}_{\max },3D_{\max }\right],\left[\frac{5}{2}{D}_{\max },4D_{\max }\right],\mathrm{and}\left[4D_{\max },5D_{\max }\right];$

fuzzifying Δx_next as backward, short, medium, and long, with corresponding ranges being respectively

$\left[-D_{\min },0\right],\left[0,\frac{1}{2}{D}_{\max }\right],\left[\frac{1}{3}{D}_{\max },\frac{2}{3}{D}_{\max }\right],\mathrm{and}\left[\frac{1}{2}{D}_{\max },D_{\max }\right],$

where D_max is a maximum forward movable distance of the swing foot, and D_min is a maximum backward movable distance of the swing foot;

(2) fuzzy rule reasoning

if x_pt−x_pt−1=backward, then x_next=backward;

if x_pt−x_pt−1=small step forward, and x_en=very near; then x_next=short;

if x_pt−x_pt−1=small step forward, and x_en=near; then x_next=short;

if x_pt−x_pt−1=small step forward, and x_en=medium; then x_next=medium;

if x_pt−x_pt−1=small step forward, and x_en=far; then x_next=long;

if x_pt−x_pt−1=small step forward, and x_en=very far; then x_next=long;

if x_pt−x_pt−1=medium step forward, and x_en=very near; then x_next=short;

if x_pt−x_pt−1=medium step forward, and x_en=near; then x_next=short;

if x_pt−x_pt−1=medium step forward, and x_en=medium; then x_next=medium;

if x_pt−x_pt−1=medium step forward, and x_en=far; then x_next=long;

if x_pt−x_pt−1=medium step forward, and x_en=very far; then x_next=long;

if x_pt−x_pt−1=big step forward, and x_en=very near; then x_next=short;

if x_pt−x_pt−1=big step forward, and x_en=near; then x_next=short;

if x_pt−x_pt−1=big step forward, and x_en=medium; then x_next=medium;

if x_pt−x_pt−1=big step forward, and x_en=far; then x_next=medium;

if x_pt−x_pt−1=big step forward, and x_en=very far; then x_next=long; and

(3) defuzzification using a centroid method

$\Delta x_{next}=\frac{\left(x_{pt}-x_{pt-1}\right){\mu }_{x_{pt}-x_{pt-1}}+x_{en}{\mu }_{x_{en}}}{{\mu }_{x_{pt}-x_{pt-1}}+{\mu }_{x_{en}}}$

where u is a membership value.

Furthermore, a process of fuzzy processing on y_en and y_sd is: inputting y_pt−y_pt−1 and y_en, and outputting Δy_next;

(1) fuzzification

fuzzifying y_pt−y_pt−1 as inward, small swing outward, medium swing outward, and big swing outward, with corresponding ranges being respectively

$\left[L_{\min },B\right],\left[B,B+\frac{1}{2}\left(L_{\max }-B\right)\right],\left[B+\frac{1}{3}\left(L_{\max }-B\right),B+\frac{2}{3}\left(L_{\max }-B\right)\right],\mathrm{and}\left[B+\frac{1}{2}\left(L_{\max }-B\right),L_{\max }\right];$

fuzzifying y_en as very near, near, medium, far, and very far, with corresponding ranges being respectively

$$\begin{gathered} \left[L_{\max },2L_{\max }\right],\left[\frac{3}{2}{L}_{\max },\frac{5}{2}{L}_{\max }\right],\text{ \\ [2⁢Lmax,3⁢Lmax],[52⁢Lmax,72⁢Lmax],and[3⁢Lmax,4⁢Lmax];} \end{gathered}$$

fuzzifying Δy_next as inward, small, medium, and large, with corresponding ranges being respectively

$$\begin{gathered} \left[L_{\min },B\right],\left[B,B+\frac{1}{2}\left(L_{\max }-B\right)\right],\text{ \\ [B+13⁢(Lmax-B),B+23⁢(Lmax-B)],and[B+12⁢(Lmax-B),Lmax],} \end{gathered}$$

where L_max is a maximum outward movable distance, L_min is a maximum inward movable distance, and B is a distance between projections of centers of gravity of two legs when the biped robot stands upright;

(2) fuzzy rule reasoning

if y_pt−y_pt−1=inward, then Δy_next=inward;

if y_pt−y_pt−1=small swing outward, and y_en=very near; then Δy_next=small;

if y_pt−y_pt−1=small swing outward, and y_en=near; then Δy_next=small;

if y_pt−y_pt−1=small swing outward, and y_en=medium; then Δy_next=medium;

if y_pt−y_pt−1=small swing outward, and y_en=far; then Δy_next=large;

if y_pt−y_pt−1=small swing outward, and y_en=very far; then Δy_next=large;

if y_pt−y_pt−1=medium swing outward, and y_en=very near; then Δy_next=small;

if y_pt−y_pt−1=medium swing outward, and y_en=near; then Δy_next=small;

if y_pt−y_pt−1=medium swing outward, and y_en=medium; then Δy_next=medium;

if y_pt−y_pt−1=medium swing outward, and y_en=far; then Δy_next=large;

if y_pt−y_pt−1=medium swing outward, and y_en=very far; then Δy_next=large;

if y_pt−y_pt−1=big swing outward, and y_en=very near; then Δy_next=small;

if y_pt−y_pt−1=big swing outward, and y_en=near; then Δy_next=small;

if y_pt−y_pt−1=big swing outward, and y_en=medium; then Δy_next=medium;

if y_pt−y_pt−1=big swing outward, and y_en=far; then Δy_next=large;

if y_pt−y_pt−1=big swing outward, and y_en=very far; then Δy_next=large; and

(3) defuzzification using a centroid method

$\Delta y_{\mathrm{next}}=\frac{\left(y_{pt}-y_{pt-1}\right){\mu }_{y_{pt}-y_{pt-1}}+y_{en}{\mu }_{y_{en}}}{{\mu }_{y_{pt}-y_{pt-1}}+{\mu }_{y_{en}}}$

where u is a membership value.

Furthermore, a process of fuzzy processing on θ_sd and θ_pt is: inputting θ_sd and θ_pt, and outputting Δθ_next;

(1) fuzzification

fuzzifying θ_sd as inward yaw, small-amplitude outward yaw, medium-amplitude outward yaw, and large-amplitude outward yaw, with corresponding ranges being respectively

$\left[-{\theta }_{\min },0\right],\left[0,\frac{1}{2}{\theta }_{\max }\right],\left[\frac{1}{3}{\theta }_{\max },\frac{2}{3}{\theta }_{\max }\right],\mathrm{and}\left[\frac{1}{2}{\theta }_{\max },{\theta }_{\max }\right];$

fuzzifying θ_pt as negative big orientation, negative small orientation, medium, positive small orientation, and positive big orientation, with corresponding ranges being respectively

$\left[-\frac{\pi }{2},-\frac{\pi }{4}\right],\left[-\frac{\pi }{3},-\frac{\pi }{6}\right],\left[-\frac{\pi }{4},\frac{\pi }{4}\right],\left[\frac{\pi }{6},\frac{\pi }{3}\right],\mathrm{and}\left[\frac{\pi }{4},\frac{\pi }{2}\right];$

fuzzifying Δθ_next as inward yaw, small-amplitude outward yaw, medium-amplitude outward yaw, and large-amplitude outward yaw, with corresponding ranges being respectively

$\left[-{\theta }_{\min },0\right],\left[0,\frac{1}{2}{\theta }_{\max }\right],\left[\frac{1}{3}{\theta }_{\max },\frac{2}{3}{\theta }_{\max }\right],\mathrm{and}\left[\frac{1}{2}{\theta }_{\max },{\theta }_{\max }\right],$

where θ_min is a maximum inward yaw angle of the swing foot, and θ_max is a maximum outward yaw angle of the swing foot;

(2) fuzzy rule reasoning

when a support foot is a left foot:

if θ_sd=inward yaw, and θ_pt=negative big orientation; then Δθ_next=large-amplitude outward yaw;

if θ_sd=inward yaw, and θ_pt=negative small orientation; then Δθ_next=medium-amplitude outward yaw;

if θ_sd=inward yaw, and θ_pt=medium; then Δθ_next=small-amplitude outward yaw;

if θ_sd=inward yaw, and θ_pt=positive small orientation; then Δθ_next=inward yaw;

if θ_sd=inward yaw, and θ_pt=positive big orientation; then Δθ_next=inward yaw;

if θ_sd=small-amplitude outward yaw, and θ_pt=negative big orientation; then Δθ_next=large-amplitude outward yaw;

if θ_sd=small-amplitude outward yaw, and θ_pt=negative small orientation; then Δθ_next=medium-amplitude outward yaw;

if θ_sd=small-amplitude outward yaw, and θ_pt=medium; then Δθ_next=small-amplitude outward yaw;

if θ_sd=small-amplitude outward yaw, and θ_pt=positive small orientation; then Δθ_next=inward yaw;

if θ_sd=small-amplitude outward yaw, and θ_pt=positive big orientation; then Δθ_next=inward yaw;

if θ_sd=medium-amplitude outward yaw, and θ_pt=negative big orientation; then Δθ_next=large-amplitude outward yaw;

if θ_sd=medium-amplitude outward yaw, and θ_pt=negative small orientation; then Δθ_next=medium-amplitude outward yaw;

if θ_sd=medium-amplitude outward yaw, and θ_pt=medium; then Δθ_next=inward yaw;

if θ_sd=medium-amplitude outward yaw, and θ_pt=positive small orientation; then Δθ_next=inward yaw;

if θ_sd=medium-amplitude outward yaw, and θ_pt=positive big orientation; then Δθ_next=inward yaw;

if θ_sd=large-amplitude outward yaw, and θ_pt=negative big orientation; then Δθ_next=medium-amplitude outward yaw;

if θ_sd=large-amplitude outward yaw, and θ_pt=negative small orientation; then Δθ_next=small-amplitude outward yaw;

if θ_sd=large-amplitude outward yaw, and θ_pt=medium; then Δθ_next=inward yaw;

if θ_sd=large-amplitude outward yaw, and θ_pt=positive small orientation; then Δθ_next=inward yaw;

if θ_sd=large-amplitude outward yaw, and θ_pt=positive big orientation; then Δθ_next=inward yaw;

when the support foot is a right foot:

if θ_sd=inward yaw, and θ_pt=negative big orientation; then Δθ_next=inward yaw;

if θ_sd=inward yaw, and θ_pt=negative small orientation; then Δθ_next=inward yaw;

if θ_sd=inward yaw, and θ_pt=medium; then Δθ_next=small-amplitude outward yaw;

if θ_sd=inward yaw, and θ_pt=positive small orientation; then Δθ_next=medium-amplitude outward yaw;

if θ_sd=inward yaw, and θ_pt=positive big orientation; then Δθ_next=large-amplitude outward yaw;

if θ_sd=small-amplitude outward yaw, and θ_pt=negative big orientation; then Δθ_next=inward yaw;

if θ_sd=small-amplitude outward yaw, and θ_pt=negative small orientation; then Δθ_next=inward yaw;

if θ_sd=small-amplitude outward yaw, and θ_pt=medium; then Δθ_next=small-amplitude outward yaw;

if θ_sd=small-amplitude outward yaw, and θ_pt=positive small orientation; then Δθ_next=medium-amplitude outward yaw;

if θ_sd=small-amplitude outward yaw, and θ_pt=positive big orientation; then Δθ_next=large-amplitude outward yaw;

if θ_sd=medium-amplitude outward yaw, and θ_pt=negative big orientation; then Δθ_next=inward yaw;

if θ_sd=medium-amplitude outward yaw, and θ_pt=negative small orientation; then Δθ_next=inward yaw;

if θ_sd=medium-amplitude outward yaw, and θ_pt=medium; then Δθ_next=inward yaw;

if θ_sd=medium-amplitude outward yaw, and θ_pt=positive small orientation; then Δθ_next=small-amplitude outward yaw;

if θ_sd=medium-amplitude outward yaw, and θ_pt=positive big orientation; then Δθ_next=medium-amplitude inward yaw;

if θ_sd=large-amplitude outward yaw, and θ_pt=negative big orientation; then Δθ_next=inward yaw;

if θ_sd=large-amplitude outward yaw, and θ_pt=negative small orientation; then Δθ_next=inward yaw;

if θ_sd=large-amplitude outward yaw, and θ_pt=medium; then Δθ_next=inward yaw;

if θ_sd=large-amplitude outward yaw, and θ_pt=positive small orientation; then Δθ_next=small-amplitude outward yaw;

if θ_sd=large-amplitude outward yaw, and θ_pt=positive big orientation; then Δθ_next=medium-amplitude inward yaw; and

(3) defuzzification

defuzzification using a centroid method:

$\Delta {\theta }_{\mathrm{next}}=\frac{{\theta }_{\mathrm{pt}}{\mu }_{{\theta }_{\mathrm{pt}}}+{\theta }_{sd}{\mu }_{{\theta }_{\mathrm{sd}}}}{{\mu }_{{\theta }_{\mathrm{pt}}}+{\mu }_{{\theta }_{\mathrm{sd}}}}$

where u is a membership value.

Furthermore, the specific foothold position of the robot is (x_next, y_next, θ_next), and x_next=x_sd+Δx_next, y_next=y_sd+Δy_next, and θ_next=θ_sd+Δθ_next.

The beneficial effects of the present disclosure are as follows:

(1) In the present disclosure, x_sd, x_en, y_sd, y_en, θ_sd, and θ_pt are separately used as inputs of a fuzzy controller, fuzzy processing is performed, fuzzy processing including fuzzification, fuzzy rule reasoning, and defuzzification, finally x_next, y_next, and θ_next are output, and a specific foothold position of a swing foot of a robot in an available foothold area is determined; and the planning of the foothold position is determining the available foothold area according to step-length constraints, movement capabilities, foot sizes, and center offsets of the biped robot, instead of setting foothold sets in advance and selecting few limited foothold positions in the foothold sets; it is unnecessary to design a limited number of foothold sets in advance, the flexibility of the biped robot is improved, and movement capabilities of the biped robot are fully exerted, so that the foothold of the biped robot after path planning can be selected in an autonomous and intelligent way.

(2) The fuzzy control process in the present disclosure considers the position of a closest path node, the position of an obstacle, and the position of a current support foot, so that the fuzzy control process is more precise and comprehensive, and the selection of the foothold position is more reasonable.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic structural view of a foothold position control system for a biped robot in the present disclosure.

FIG. 2 is a flowchart of a foothold position control method for a biped robot in the present disclosure.

FIG. 3 is an overall effect diagram of a passing path in the present disclosure.

FIG. 4 is a schematic diagram of an available foothold area of a swing foot in the present disclosure.

FIG. 5 is a schematic diagram of a fuzzy control process in the present disclosure.

FIG. 6 is a schematic diagram of θ_pt in the present disclosure.

DETAILED DESCRIPTION OF THE EMBODIMENTS

The present disclosure is further illustrated below with reference to the accompanying drawings and specific embodiments, but the protection scope of the present disclosure is not limited thereto.

As shown in FIG. 1, a foothold position control system for a biped robot includes a laser radar or vision sensor 1, a gyroscope 3, a six-component force sensor 6, and an industrial computer; the laser radar or vision sensor 1 is disposed on the top of a biped robot body 2, and is used for obtaining environmental information of a biped robot, and position coordinates (x_en, y_en) of an obstacle 8 in the environmental information in a robot body coordinate system; the gyroscope 3 is disposed in the upper part of the biped robot body 2, and is used for obtaining an initial value of a yaw angle θ_sd of the biped robot; the six-component force sensor 6 is disposed in a sole of the biped robot body 2, and is used for obtaining a real-time position (x_sd, y_sd) of a support foot of the biped robot; the laser radar or vision sensor 1, the gyroscope 3, and the six-component force sensor 6 all perform signal transmission with the industrial computer, a fuzzy controller is disposed in the industrial computer, and the fuzzy controller converts the real-time position (x_sd, y_sd) of the support foot in a world coordinate system into a path node (x_pt, y_pt) closest to the position of the current support foot (see equation (1)); x_sd, x_en, y_sd, y_en, θ_sd and θ_pt are separately used as inputs of the fuzzy controller, and fuzzification, fuzzy rule reasoning, and defuzzification are performed to obtain a specific foothold position (x_next, y_next, θ_next) of the swing foot in an available foothold range, and the swing foot is controlled to land. The θ_pt is an angle between a straight line formed by a current path node and an associated path node in a next step and an x axis of the world coordinate system (see FIG. 6) (see equations (4) and (5) for the obtaining process). The x-axis direction in the figure is the advancing direction of the robot, the y-axis direction is perpendicular to the advancing direction of the robot, and they form the robot body coordinate system; 4 is a lower limb of the robot, and 7 is a foothold position of the robot.

As shown in FIG. 2, a foothold position control method for a biped robot specifically includes the following steps:

Step (1): Path Planning

The laser radar or vision sensor 1 is mounted on the biped robot to obtain an environmental map, grid processing is performed on the environmental map, a start point (x_start, y_start) and a task goal (x_goal, y_goal) are set in the grid map, and an optimal collision-free path (namely, the path 5 in FIG. 3) from the start to the goal is planned by using an A* algorithm, where each path point is called a path node and represented as (x_i, y_i) (i=1 . . . n); that is, passable path nodes form a passable path Path=(x_start, y_start)∪(x₁, y₁) . . . ∪(x₂, y₂) . . . ∪(x_N, y_N)∪(x_goal, y_goal), as shown in

FIG. 3, where 9 is the start, and 10 is the goal.

Step (2): Determining of Foothold Range

An available foothold area of the swing foot is determined according to step-length constraints, movement capabilities, foot sizes, and center offsets of a specific biped robot.

When the advancing direction of the biped robot is the x-axis direction, x_max=x_sd+D_max, x_min=x_sd−D_min, y_max=y_sd+L_max, y_min=y_sd+L_min;

when the advancing direction of the biped robot is along the y axis, y_max=y_sd+D_max, y_min=y_sd−D_min, x_max=x_stand+L_max, x_min=x_sd+L_min;

when the angle between the advancing direction of the biped robot and the x axis is θ,

$x_{\max }=x_{sd}+\frac{D_{\max }}{\cos \left(\theta \right)},x_{\min }=x_{sd}-\frac{D_{\min }}{\cos \left(\theta \right)},y_{\max }=y_{sd}+\frac{L_{\max }}{\cos \left(\theta \right)},y_{\min }=y_{sd}+\frac{L_{\min }}{\cos \left(\theta \right)}.$

Thus it can be determined that the available foothold area of the swing foot of the biped robot is [x_min x_max]∪[y_min y_max]∪[θ_min θ_max], as shown in FIG. 4, where the area in the dashed box is the available foothold area of the swing foot, where θ_max is a maximum outward yaw angle of the swing foot, θ_min is a maximum inward yaw angle of the swing foot, D_max is a maximum forward movable distance, D_min is a maximum backward movable distance, L_max is a maximum outward movable distance, L_min is a maximum inward movable distance, and B is a distance between projections of centers of gravity of two legs when the biped robot stands upright.

Step (3): Determining of Specific Foothold Position

Any (x, y, θ) combination in the available foothold area can represent a possible foothold position of the actual swing foot.

According to the position (x_sd, y_sd, θ_sd) of the support foot, the position (x_pt, y_pt, θ_pt) of the path node closest to the position of the support foot, and the position (x_en, y_en) of the obstacle in the environment, x, y, and θ are separately subjected to fuzzification processing in their respective ranges, subjected to fuzzy rule reasoning, and then subjected to defuzzification to finally obtain the specific foothold position (x_next, y_next, θ_next) of the swing foot in the available foothold range.

x_sd, x_en, y_sd, y_en, θ_sd, and θ_pt are separately used as inputs of fuzzy control, and the current support foot of the biped robot is the swing foot in the previous step; thus the value of the yaw angle θ_sd of the current support foot of the biped robot is the value of θ_next−1 of the swing foot in the previous step.

As shown in FIG. 5, the process of fuzzy processing on x_en and x_sd is as follows:

{circle around (1)} The position of the path node closest to the position of the support foot is obtained through x_sd and y_sd:

$\begin{array}{cc}\left(x_{\mathrm{pt}},y_{\mathrm{pt}}\right)=\underset{1\le i\le n}{\min }\sqrt{{\left(x_{\mathrm{sd}}-x_i\right)}^2+{\left(y_{\mathrm{sd}}-y_i\right)}^2}& \left(1\right)\end{array}$

A difference x_pt−x_pt−1 between x_pt calculated in this step and x_pt−1 in the last step is used as an input, and a step length of the next step is predicted through a step length of the last step; meanwhile, x_en is used as another input, and the step length of the next step is predicted through the distance from the obstacle; Δx_next is output according to fuzzy rules and reasoning.

{circle around (2)} Fuzzification processing: the range of x_pt−x_pt−1 is determined approximately as the range [−D_min, D_max] of one step length, and the range is fuzzified and divided into four stages: backward, small step forward, medium step forward, and big step forward, with corresponding ranges being respectively

$\left[-D_{\min },0\right],\left[0,\frac{1}{2}{D}_{\max }\right],\left[\frac{1}{3}{D}_{\max },\frac{2}{3}{D}_{\max }\right],\mathrm{and}\left[\frac{1}{2}{D}_{\max },D_{\max }\right];$

a value range of x_en is approximately represented as

$\left[\frac{1}{2}{D}_{\max },5D_{\max }\right]$

and is fuzzified as very near, near, medium, far, and very far, with corresponding ranges being respectively

$\left[\frac{1}{2}{D}_{\max },D_{\max }\right],\left[\frac{3}{4}{D}_{\max },2D_{\max }\right],\left[\frac{3}{2}{D}_{\max },3D_{\max }\right],\left[\frac{5}{2}{D}_{\max },4D_{\max }\right],\mathrm{and}\left[4D_{\max },5D_{\max }\right];$

the output Δx_next is fuzzified as backward, short, medium, and long, with corresponding ranges being respectively

$\left[-D_{\min },0\right],\left[0,\frac{1}{2}{D}_{\max }\right],\left[\frac{1}{3}{D}_{\max },\frac{2}{3}{D}_{\max }\right],\mathrm{and}\left[\frac{1}{2}{D}_{\max },D_{\max }\right].$

All the inputs use a triangle method as a membership function, and have respective membership values being μ_x.

{circle around (3)} Fuzzy rules and reasoning:

IF x_ptx_pt−1=back THEN x_next=backward;

IF x_pt−x_pt−1=small step forward AND x_en=very near, THEN x_next=short;

IF x_pt−x_pt−1=small step forward AND x_en=near, THEN x_next=short;

IF x_pt−x_pt−1=small step forward AND x_en=medium, THEN x_next=medium;

IF x_pt−x_pt−1=small step forward AND x_en=far, THEN x_next=long;

IF x_pt−x_pt−1=small step forward AND x_en=very far, THEN x_next=long;

IF x_pt−x_pt−1=medium step forward AND x_en=very near, THEN x_next=short;

IF x_pt−x_pt−1=medium step forward AND x_en=near, THEN x_next=short;

IF x_pt−x_pt−1=medium step forward AND x_en=medium, THEN x_next=medium;

IF x_pt−x_pt−1=medium step forward AND x_en=far, THEN x_next=long;

IF x_pt−x_pt−1=medium step forward AND x_en=very far, THEN x_next=long;

IF x_pt−x_pt−1=big step forward AND x_en=very near, THEN x_next=short;

IF x_pt−x_pt−1=big step forward AND x_en=near, THEN x_next=short;

IF x_pt−x_pt−1=big step forward AND x_en=medium, THEN x_next=medium;

IF x_pt−x_pt−1=big step forward AND x_en=far, THEN x_next=medium;

IF x_pt−x_pt−1=big step forward AND x_en=very far, THEN x_next=long.

{circle around (4)} Defuzzification

Defuzzification is performed using a centroid method:

$\begin{array}{cc}\Delta x_{next}=\frac{\left(x_{pt}-x_{pt-1}\right){\mu }_{x_{\mathrm{pt}}-x_{\mathrm{pt}-1}}+x_{en}{\mu }_{x_{en}}}{{\mu }_{x_{\mathrm{pt}}-x_{\mathrm{pt}-1}}+{\mu }_{x_{en}}}& \left(2\right)\end{array}$

The process of fuzzy processing on y_en and y_sd is as follows:

{circle around (1)} The position (x_pt, y_pt) of the path node closest to the position of the support foot is obtained from equation (1).

A difference y_pt−y_pt−1 between y_pt calculated in this step and y_pt−1 in the last step is used as an input, and a step length of the next step is predicted through a step length of the last step; meanwhile, y_en is used as another input, and the step length of the next step is predicted through the distance from the obstacle; y_next is output according to fuzzy rules and reasoning.

{circle around (2)} Fuzzification processing: the range of y_pt−y_pt−1 is determined approximately as the range [L_min, L_max] of one step swing, and the range is fuzzified and divided into four stages: inward, small swing outward, medium swing outward, and big swing outward, with corresponding ranges being respectively

$\left[L_{\min },B\right],\left[B,B+\frac{1}{2}\left(L_{\max }-B\right)\right],\left[B+\frac{1}{3}\left(L_{\max }-B\right),B+\frac{2}{3}\left(L_{\max }-B\right)\right],\mathrm{and}\left[B+\frac{1}{2}\left(L_{\max }-B\right),L_{\max }\right];$

a value range of y_en is approximately represented as [L_max, 4L_max], and fuzzified as very near, near, medium, far, and very far, with corresponding ranges being respectively

$\left[L_{\max },2L_{\max }\right],\left[\frac{3}{2}{L}_{\max },\frac{5}{2}{L}_{\max }\right],\left[2L_{\max },3L_{\max }\right],\left[\frac{5}{2}{L}_{\max },\frac{7}{2}{L}_{\max }\right],\mathrm{and}\left[3L_{\max },4L_{\max }\right];$

the output Δy_next is fuzzified as inward, small, medium, and large, with corresponding ranges being respectively

$\left[L_{\min },B\right],\left[B,B+\frac{1}{2}\left(L_{\max }-B\right)\right],\left[B+\frac{1}{3}\left(L_{\max }-B\right),B+\frac{2}{3}\left(L_{\max }-B\right)\right],\mathrm{and}\left[B+\frac{1}{2}\left(L_{\max }-B\right),L_{\max }\right].$

All the inputs use a triangle method as a membership function, and have respective membership values being μ_y.

{circle around (3)} Fuzzy rules and reasoning

IF y_pt−y_pt−1=inward THEN Δy_next=inward;

IF y_pt−y_pt−1=small swing outward AND y_en=very near, THEN Δy_next=small;

IF y_pt−y_pt−1=small swing outward AND y_en=near, THEN Δy_next=small;

IF y_pt−y_pt−1=small swing outward AND y_en=medium, THEN Δy_next=medium;

IF y_pt−y_pt−1=small swing outward AND y_en=far, THEN Δy_next=large;

IF y_pt−y_pt−1=small swing outward AND y_en=very far, THEN Δy_next=large;

IF y_pt−y_pt−1=medium swing outward AND y_en=very near, THEN Δy_next=small;

IF y_pt−y_pt−1=medium swing outward AND y_en=near, THEN Δy_next=small;

IF y_pt−y_pt−1=medium swing outward AND y_en=medium, THEN Δy_next=medium;

IF y_pt−y_pt−1=medium swing outward AND y_en=far, THEN Δy_next=large;

IF y_pt−y_pt−1=medium swing outward AND y_en=very far, THEN Δy_next=large;

IF y_pt−y_pt−1=big swing outward AND y_en=very near, THEN Δy_next=small;

IF y_pt−y_pt−1=big swing outward AND y_en=near, THEN Δy_next=small;

IF y_pt−y_pt−1=big swing outward AND y_en=medium, THEN Δy_next=medium;

IF y_pt−y_pt−1=big swing outward AND y_en=far, THEN Δy_next=large;

IF y_pt−y_pt−1=big swing outward AND y_en=very far, THEN Δy_next=large.

{circle around (4)} Defuzzification

Defuzzification is performed using a centroid method:

$\begin{array}{cc}\Delta y_{\mathrm{next}}=\frac{\left(y_{\mathrm{pt}}-y_{\mathrm{pt}-1}\right){\mu }_{y_{pt}-y_{\mathrm{pt}-1}}+y_{en}{\mu }_{y_{en}}}{{\mu }_{y_{\mathrm{pt}}-y_{\mathrm{pt}-1}}+{\mu }_{y_{en}}}& \left(3\right)\end{array}$

The process of fuzzy processing on θ_sd and θ_pt is as follows:

{circle around (1)} The current support foot of the biped robot is the swing foot in the previous step, and thus the value of the yaw angle θ_sd of the support foot of the biped robot is the value θ_next−1 of the swing foot in the previous step, and is used as an input, and a yaw angle in the next step is predicted through a yaw angle in the previous step; θ_pt is obtained from equations (4) and (5) and is used as another input, and the yaw angle in the next step is predicted through the yaw angle in the previous step.

$\begin{array}{cc}\left(x_{\mathrm{pt}+1},y_{\mathrm{pt}+1}\right)=\underset{1\le i\le n}{\min }\sqrt{{\left(x_{\mathrm{next}}-x_i\right)}^2+{\left(y_{\mathrm{next}}-y_i\right)}^2}& \left(4\right)\end{array}$ $\begin{array}{cc}{\theta }_{\mathrm{pt}}={\tan }^{-1}\frac{y_{pt+1}-y_{pt}}{x_{pt+1}-x_{pt}}& \left(5\right)\end{array}$

where x_pt+1, y_pt+1 is a path node closest to a foothold position after a current swing foot lands.

{circle around (2)} Fuzzification processing

A variation range of θ_sd is [−θ_min, θ_max], and the range is fuzzified and divided into four stages: inward yaw, small-amplitude outward yaw, medium-amplitude outward yaw, and large-amplitude outward yaw, with corresponding ranges being respectively

$\left[-{\theta }_{\min },0\right],\left[0,\frac{1}{2}{\theta }_{\max }\right],\left[\frac{1}{3}{\theta }_{\max },\frac{2}{3}{\theta }_{\max }\right],{\mathrm{and}\left[\frac{1}{2}{\theta }_{\max },{\theta }_{\max }\right]}_{;}$

the range of θ_pt is approximately represented as

$\left[-\frac{\pi }{2},\frac{\pi }{2}\right],$

and is fuzzified as negative big orientation, negative small orientation, medium, positive small orientation, and positive big orientation, with corresponding ranges being respectively

$\left[-\frac{\pi }{2},-\frac{\pi }{4}\right],\left[-\frac{\pi }{3},-\frac{\pi }{6}\right],\left[-\frac{\pi }{4},\frac{\pi }{4}\right],\left[\frac{\pi }{6},\frac{\pi }{3}\right],{\mathrm{and}\left[\frac{\pi }{4},\frac{\pi }{2}\right]}_{;}$

the output Δθ_next is fuzzified as inward yaw, small-amplitude outward yaw, medium-amplitude outward yaw, and large-amplitude outward yaw, with corresponding ranges being respectively

$\left[-{\theta }_{\min },0\right],\left[0,\frac{1}{2}{\theta }_{\max }\right],\left[\frac{1}{3}{\theta }_{\max },\frac{2}{3}{\theta }_{\max }\right],\mathrm{and}\left[\frac{1}{2}{\theta }_{\max },{\theta }_{\max }\right].$

All the inputs use a triangle method as a membership function, and have respective membership values being μ_θ.

{circle around (3)} Fuzzy rules and reasoning

When the support foot is a left foot:

IF θ_sd=inward yaw AND θ_pt=negative big orientation, THEN Δθ_next=large-amplitude outward yaw;

IF θ_sd=inward yaw AND θ_pt=negative small orientation, THEN Δθ_next=medium-amplitude outward yaw;

IF θ_sd=inward yaw AND θ_pt=medium, THEN Δθ_next=small-amplitude outward yaw;

IF θ_sd=inward yaw AND θ_pt=positive small orientation, THEN Δθ_next=inward yaw;

IF θ_sd=inward yaw AND θ_pt=positive big orientation, THEN Δθ_next=inward yaw;

IF θ_sd=small-amplitude outward yaw AND θ_pt=negative big orientation, THEN Δθ_next=large-amplitude outward yaw;

IF θ_sd=small-amplitude outward yaw AND θ_pt=negative small orientation, THEN Δθ_next=medium-amplitude outward yaw;

IF θ_sd=small-amplitude outward yaw AND θ_pt=medium, THEN Δθ_next=small-amplitude outward yaw;

IF θ_sd=small-amplitude outward yaw AND θ_pt=positive small orientation, THEN Δθ_next=inward yaw;

IF θ_sd=small-amplitude outward yaw AND θ_pt=positive big orientation, THEN Δθ_next=inward yaw;

IF θ_sd=medium-amplitude outward yaw AND θ_pt=negative big orientation, THEN Δθ_next=large-amplitude outward yaw;

IF θ_sd=medium-amplitude outward yaw AND θ_pt=negative small orientation, THEN Δθ_next=medium-amplitude outward yaw;

IF θ_sd=medium-amplitude outward yaw AND θ_pt=medium, THEN Δθ_next=inward yaw;

IF θ_sd=medium-amplitude outward yaw AND θ_pt=positive small orientation, THEN Δθ_next=inward yaw;

IF θ_sd=medium-amplitude outward yaw AND θ_pt=positive big orientation, THEN Δθ_next=inward yaw;

IF θ_sd=large-amplitude outward yaw AND θ_pt=negative big orientation, THEN Δθ_next=medium-amplitude outward yaw;

IF θ_sd=large-amplitude outward yaw AND θ_pt=negative small orientation, THEN Δθ_next=small-amplitude outward yaw;

IF θ_sd=large-amplitude outward yaw AND θ_pt=medium, THEN Δθ_next=inward yaw;

IF θ_sd=large-amplitude outward yaw AND θ_pt=positive small orientation, THEN Δθ_next=inward yaw;

IF θ_sd=large-amplitude outward yaw AND θ_pt=positive big orientation, THEN Δθ_next=inward yaw;

When the support foot is a right foot:

IF θ_sd=inward yaw AND θ_pt=negative big orientation, THEN Δθ_next=inward yaw;

IF θ_sd=inward yaw AND θ_pt=negative small orientation, THEN Δθ_next=inward yaw;

IF θ_sd=inward yaw AND θ_pt=medium, THEN Δθ_next=small-amplitude outward yaw;

IF θ_sd=inward yaw AND θ_pt=positive small orientation, THEN Δθ_next=medium-amplitude outward yaw;

IF θ_sd=inward yaw AND θ_pt=positive big orientation, THEN Δθ_next=large-amplitude outward yaw;

IF θ_sd=small-amplitude outward yaw AND θ_pt=negative big orientation, THEN Δθ_next=inward yaw;

IF θ_sd=small-amplitude outward yaw AND θ_pt=negative small orientation, THEN Δθ_next=inward yaw;

IF θ_sd=small-amplitude outward yaw AND θ_pt=medium, THEN Δθ_next=small-amplitude outward yaw;

IF θ_sd=small-amplitude outward yaw AND θ_pt=positive small orientation, THEN Δθ_next=medium-amplitude outward yaw;

IF θ_sd=small-amplitude outward yaw AND θ_pt=positive big orientation, THEN Δθ_next=large-amplitude outward yaw;

IF θ_sd=medium-amplitude outward yaw AND θ_pt=negative big orientation, THEN Δθ_next=inward yaw;

IF θ_sd=medium-amplitude outward yaw AND θ_pt=negative small orientation, THEN Δθ_next=inward yaw;

IF θ_sd=medium-amplitude outward yaw AND θ_pt=medium, THEN Δθ_next=inward yaw;

IF θ_sd=medium-amplitude outward yaw AND θ_pt=positive small orientation, THEN Δθ_next=small-amplitude outward yaw;

IF θ_sd=medium-amplitude outward yaw AND θ_pt=positive big orientation, THEN Δθ_next=medium-amplitude inward yaw;

IF θ_sd=large-amplitude outward yaw AND θ_pt=negative big orientation, THEN Δθ_next=inward yaw;

IF θ_sd=large-amplitude outward yaw AND θ_pt=negative small orientation, THEN Δθ_next=inward yaw;

IF θ_sd=large-amplitude outward yaw AND θ_pt=medium, THEN Δθ_next=inward yaw;

IF θ_sd=large-amplitude outward yaw AND θ_pt=positive small orientation, THEN Δθ_next=small-amplitude outward yaw;

IF θ_sd=large-amplitude outward yaw AND θ_pt=positive big orientation, THEN Δθ_next=medium-amplitude inward yaw.

{circle around (4)} Defuzzification

Defuzzification is performed using a centroid method:

$\begin{array}{cc}\Delta {\theta }_{\mathrm{next}}=\frac{{\theta }_{\mathrm{pt}}{\mu }_{{\theta }_{\mathrm{pt}}}+{\theta }_{sd}{\mu }_{{\theta }_{\mathrm{sd}}}}{{\mu }_{{\theta }_{\mathrm{pt}}}+{\mu }_{{\theta }_{\mathrm{sd}}}}& \left(6\right)\end{array}$

Finally, the three outputs (x_next, y_next, θ_next) of the fuzzy controller are:

• • x_next=x_sd+Δx_next, y_next=y_sd+Δy_next, θ_next=θ_sd+Δθ_next

Thus, the specific foothold position of the biped robot in the world coordinate system can be determined.

The described embodiments are preferred embodiments of the present disclosure, but the present disclosure is not limited to the aforementioned embodiments. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present disclosure shall fall within the protection scope of the present disclosure.

Claims

1. A foothold position control system for a biped robot, comprising a laser radar or vision sensor, a gyroscope, a six-component force sensor, and an industrial computer, wherein a fuzzy controller in the industrial computer separately uses xsd, xen, ysd, yen, θsd, and θpt as inputs and performs fuzzy processing to obtain a specific foothold position of a swing foot in an available foothold area, wherein xsd, ysd is a real-time position of a support foot of a biped robot obtained by the six-component force sensor, xen, yen is a position of an obstacle in a robot body coordinate system obtained by the laser radar or vision sensor, θsd is an initial yaw angle of the biped robot obtained by the gyroscope, and θpt is an angle between a straight line formed by a current path node and an associated path node in a next step and an x axis of a world coordinate system.

2. The foothold position control system for the biped robot according to claim 1, wherein the

3. A control method of the foothold position control system for the biped robot according to claim 1, comprising: planning a passing path; determining the available foothold area of the swing foot on the passing path according to step-length constraints, movement capabilities, foot sizes, and center offsets of the biped robot; and performing fuzzy processing to determine a specific foothold position of the biped robot.

4. The control method according to claim 3, wherein the fuzzy processing comprises fuzzification, fuzzy rule reasoning, and defuzzification.

5. The control method according to claim 4, wherein a process of the fuzzy processing on xen and xsd is: inputting xpt−xpt−1 and xen, and outputting Δxnext; (1) fuzzificationfuzzifying xpt−xpt−1 as backward, small step forward, medium step forward, and big step forward, with corresponding ranges being respectively

6. The control method according to claim 4, wherein a process of the fuzzy processing on yen and ysd is: inputting ypt−ypt−1 and yen, and outputting Δynext; (1) fuzzificationfuzzifying ypt−ypt−1 as inward, small swing outward, medium swing outward, and big swing outward, with corresponding ranges being respectively

7. The control method according to claim 4, wherein a process of the fuzzy processing on θsd and θpt is: inputting θsd and θpt, and outputting Δθnext; (1) fuzzificationfuzzifying θsd as inward yaw, small-amplitude outward yaw, medium-amplitude outward yaw, and large-amplitude outward yaw, with corresponding ranges being respectively

8. The control method according to claim 5, wherein the specific foothold position of the biped robot is (xnext, ynext, θnext), wherein xnext=xsd+Δxnext, ynext=ysd+Δynext, and θnext=θsd+Δθnext.

9. The control method according to claim 6, wherein the specific foothold position of the biped robot is (xnext, ynext, θnext), wherein xnext=xsd+Δxnext, ynext=ysd+Δynext, and θnext=θsd+Δθnext.

10. The control method according to claim 7, wherein the specific foothold position of the biped robot is (xnext, ynext, θnext), wherein xnext=xsd+Δxnext, ynext=ysd+Δynext, and θnext=θsd+Δθnext.

11. A control method of the foothold position control system for the biped robot according to claim 2, comprising: planning a passing path; determining the available foothold area of the swing foot on the passing path according to step-length constraints, movement capabilities, foot sizes, and center offsets of the biped robot; and performing fuzzy processing to determine a specific foothold position of the biped robot.

12. The control method according to claim 11, wherein the fuzzy processing comprises fuzzification, fuzzy rule reasoning, and defuzzification.

13. The control method according to claim 12, wherein a process of the fuzzy processing on xen and xsd is: inputting xpt−xpt−1 and xen, and outputting Δxnext; (1) fuzzificationfuzzifying xpt−xpt−1 as backward, small step forward, medium step forward, and big step forward, with corresponding ranges being respectively

14. The control method according to claim 12, wherein a process of the fuzzy processing on yen and ysd is: inputting ypt−ypt−1 and yen, and outputting Δynext; (1) fuzzificationfuzzifying ypt−ypt−1 as inward, small swing outward, medium swing outward, and big swing outward, with corresponding ranges being respectively

15. The control method according to claim 12, wherein a process of the fuzzy processing on θsd and θpt is: inputting θsd and θpt, and outputting Δθnext; (1) fuzzificationfuzzifying θsd as inward yaw, small-amplitude outward yaw, medium-amplitude outward yaw, and large-amplitude outward yaw, with corresponding ranges being respectively

16. The control method according to claim 13, wherein the specific foothold position of the biped robot is (xnext, ynext, θnext), wherein xnext=xsd+Δxnext, ynext=ysd+Δynext, and θnext=θsd+Δθnext.

17. The control method according to claim 14, wherein the specific foothold position of the biped robot is (xnext, ynext, θnext), wherein xnext=xsd+Δxnext, ynext=ysd+Δynext, and θnext=θsd+Δθnext.

18. The control method according to claim 15, wherein the specific foothold position of the biped robot is (xnext, ynext, θnext), wherein xnext=xsd+Δxnext, ynext=ysd+Δynext, and θnext=θsd+Δθnext.