Method of estimating floor reactions of bipedal walking body, and method of estimating joint moments of bipedal walking body Number:7,054,763 from the United States Patent and Trademark Office (PTO) owispatent

Title: Method of estimating floor reactions of bipedal walking body, and method of estimating joint moments of bipedal walking body

Abstract: Whether a motion state of leg bodies (2) is a single stance state or a double stance state is sequentially determined and a total reaction force (F) is estimated on the basis of an equation of motion for a center of gravity (G0) of a bipedal walking body (1). If the motion state is the single stance state, then the estimated value of the total floor reaction force (F) is directly used as an estimated value of the floor reaction force of the leg body (2). If the motion state is the double stance state, then a floor reaction force (Fr) of the leg body (2) at the rear side is determined, using measurement data of elapsed time of the double stance state and moving speed of a bipedal walking body (1) and pre-established characteristic data, and the floor reaction force (Fr) is subtracted from the total floor reaction force (F) to determine a floor reaction force (Ff) of the leg body (2) at the front side. Thus, it is possible to provide a method that allows floor reaction forces acting on leg bodies of a bipedal walking body, such as a human being, and moments acting on joints of the leg bodies to be determined in real time with high accuracy by a relatively simple technique.

Patent Number: 7,054,763 Issued on 05/30/2006 to Kawai,   et al.

---

Inventors:	Kawai; Masakazu (Wako, JP); Ikeuchi; Yasushi (Wako, JP); Katoh; Hisashi (Wako, JP)
Assignee:	Honda Giken Kogyo Kabushiki Kaisha (Tokyo, JP)
Appl. No.:	485439
Filed:	July 26, 2002
PCT Filed:	July 26, 2002
PCT NO:	PCT/JP02/07592
371 Date:	January 30, 2004
PCT PUB.NO.:	WO03/015997
PCT PUB. Date:	February 27, 2003

---

Foreign Application Priority Data

---

Aug 01, 2001 [JP]			2001-234155
Feb 27, 2002 [JP]			2002-050790

Current U.S. Class:	702/42 ; 73/760
Current International Class:	G01L 1/00 (20060101); G01L 3/00 (20060101); G01L 5/00 (20060101); G06F 19/00 (20060101)

---

References Cited [Referenced By]

---

U.S. Patent Documents

	5151859		September 1992		Yoshino et al.
	5349277		September 1994		Takahashi et al.
	5404086		April 1995		Takenaka et al.
	6289265		September 2001		Takenaka et al.

Foreign Patent Documents

	1 120 203		Aug., 2001		EP
	2000-249570		Sep., 2000		JP

Primary Examiner: Barlow; John
Assistant Examiner: Kundu; Sujoy
Attorney, Agent or Firm: Rankin, Hill, Porter & Clark LLP

---

Claims

---

The invention claimed is:

1. A method of estimating floor reaction forces acting on respective leg bodies of a bipedal walking body, comprising: a first step for determining whether a motion state of leg bodies of the bipedal walking body is a single stance state wherein only one of the leg bodies is in contact with the ground or a double stance state wherein both leg bodies are in contact with the ground; a second step for sequentially determining positions of the center of gravity of the bipedal walking body while also sequentially determining accelerations of the center of gravity in an absolute coordinate system fixed with respect to the ground, using time-series data of the positions of the center of gravity; a third step for sequentially determining estimated value of total floor reaction force on the basis of an equation of motion of the centers of gravity represented by a weight of the bipedal walking body, gravitational accelerations, the accelerations of the center of gravity, and total floor reaction force, which is a resultant force of the floor reaction forces acting on the respective leg bodies; a fourth step for sequentially measuring time elapse from a start of the double stance state until an end thereof each time the double stance state begins; a fifth step for measuring a moving speed of the bipedal walking body at least on or before each double stance state begins; a step of sequentially determining estimated values of the total floor reaction force, when the bipedal walking body is in the single stance state, as estimate values of floor reaction forces acting on a single leg body in contact with the ground; and a step of sequentially determining estimated values of floor reaction forces acting on one of the two leg bodies, when the bipedal walking body is in the double stance state, the one leg body being located at the rear side in relation to the direction of travel of the bipedal walking body, on the basis of characteristic data pre-established as indicative of characteristics of changes in floor reaction force acting on the one leg body with respect to the elapsed time of the double stance state and moving speed of the bipedal walking body, and sequentially determining estimated values of the floor reaction forces acting on the other leg body by subtracting the determined estimated values of the floor reaction forces of the one leg body from the estimated values of the total floor reaction force.

2. The method according to claim 1, wherein the characteristic data is data indicating a relationship between a ratio of a floor reaction force of the one leg body to the total floor reaction force at a start of the double stance state and a ratio of the elapsed time to a duration from a start to an end of the double stance state, a duration of the double stance state is estimated from a measured value of the moving speed on the basis of a pre-established correlation between the moving speed of the bipedal walking body and the duration of the double stance state, and estimated values of floor reaction forces acting on the one leg are sequentially determined on the basis of the estimated value of the duration of the double stance state, a measured value of the elapsed time, an estimated value of the total floor reaction force at the start of the double stance state, and the characteristic data.

3. The method according to claim 1, further comprising the steps of: measuring a tilt angle of a crus under a knee joint of each leg body of the bipedal walking body and a tilt angle of a thigh between a hip joint and a knee joint of the leg body at a start of each double stance state; calculating, at a start of each double stance state, a shift amount of a position of a bottom end portion of the crus of the leg body with respect to the hip joint of the leg body existing at the rear side in relation to the direction of travel of the bipedal walking body, the shift taking place in the direction of travel of the bipedal walking body from the start of the preceding double stance state, on the basis of measured values of tilt angles of the thigh and the crus of the leg body and pre-acquired sizes of the thigh and the crus of the leg body each time the double stance state begins; and measuring the time elapsed from a start of each double stance state to a start of the next double stance state as elapsed time for one step, wherein in the fifth step, each time the double stance state begins, the shift amount calculated at the start is divided by the one-step elapsed time measured from a start of the preceding double stance state to a start of the present double stance state so as to obtain a measured value of the moving speed.

4. The method according to claim 1, further comprising the steps of: sequentially measuring a vertical acceleration of a lower portion of a torso, which is adjacent to a hip joint, the torso being supported on the two leg bodies through the intermediary of hip joints of the leg bodies, wherein a motion state of the bipedal walking body is determined in the first step as the beginning of the double stance state and the end of the single stance state when the vertical acceleration of the lower portion of the torso increases to a predetermined threshold value or more, and as the end of the double stance state and the beginning of the single stance state when an estimated value of the floor reaction force acting on the leg body at the rear side in relation to the direction of travel of the bipedal walking body in the double stance state decreases to a predetermined threshold value or less.

5. The method according to claim 2, comprising the steps of: measuring a vertical acceleration of a lower portion of a torso, which is adjacent to a hip joint, the torso being supported on the two leg bodies through the intermediary of hip joints of the leg bodies, wherein a motion state of the bipedal walking body is determined in the first step as the beginning of the double stance state and the end of the single stance state when the vertical acceleration of the lower portion of the torso increases to a predetermined threshold value or more, and as the end of the double stance state and the beginning of the single stance state when a measured value of time elapsed from the start of the double stance state reaches an estimated value of a duration of the double stance state.

6. The method according to claim 1, comprising the steps of: sequentially measuring tilt angles of a torso supported on the two leg bodies through the intermediary of hip joints of the respective leg bodies, bending angles of at least hip joints and knee joints of the respective leg bodies, and accelerations in the absolute coordinate system of a predetermined reference point of the bipedal walking body, wherein in the second step, positions of the center of gravity of the bipedal walking body with respect to the reference point are sequentially determined on the basis of tilt angles of the torso, bending angles of the hip joints and the knee joints, a rigid body link model representing the bipedal walking body in the form of a link assembly of a plurality of rigid bodies, pre-acquired weights of individual portions of the bipedal walking body, which correspond to individual rigid bodies of the rigid body link model, and the pre-acquired positions of the centers of gravity of the portions corresponding to rigid bodies in the individual portions corresponding to rigid bodies, accelerations of the center of gravity with respect to the reference point are sequentially determined on the basis of time-series data of the positions of the center of gravity, and accelerations of the center of gravity in the absolute coordinate system are determined from the accelerations of the center of gravity with respect to the reference point and the accelerations of the reference point in the absolute coordinate system.

7. The method according to claim 6, wherein the reference point is set in the torso.

8. The method according to claim 6, wherein the torso has a waist connected to the two leg bodies through the intermediary of hip joints, and a chest located on the waist such that it can be tilted with respect to the waist, wherein tilt angles of the torso used to determine a position of the center of gravity are tilt angles of the waist and the chest, respectively.

9. The method according to claim 8, wherein the rigid body link model is a model in which cruses under knee joints of the respective leg bodies, thighs between the knee joints and the hip joints, the waist, and a body that includes the chest located on the upper side of the waist of the bipedal walking body are respectively represented in the form of rigid bodies.

10. A method of estimating moments acting on at least one joint of each leg body of the bipedal walking body by using estimated values of floor reaction forces related to each leg body that have been sequentially determined according to the method of estimating floor reactions of a bipedal walking body described in claim 1, comprising the steps of: sequentially measuring tilt angles of a torso supported on the two leg bodies through the intermediary of hip joints of the respective leg bodies, bending angles of at least hip joints and knee joints of the respective leg bodies, and accelerations of a pre-established reference point of the bipedal walking body in the absolute coordinate system; sequentially determining tilt angles of portions corresponding to rigid bodies of the bipedal walking body that are associated with respective rigid bodies of a rigid body link model on the basis of tilt angles of the torso, bending angles of the respective hip joints and the knee joints of the leg bodies, and the rigid body link model representing the bipedal walking body in the form of a link assembly of a plurality of rigid bodies; sequentially determining positions of centers of gravity of the portions corresponding to rigid bodies in relation to the reference point on the basis of tilt angles of the portions corresponding to the rigid bodies, pre-obtained weights of the portions corresponding to the rigid bodies, and pre-obtained positions of centers of gravity of the portions corresponding to the rigid bodies in the individual portions corresponding to the rigid bodies, and sequentially determining accelerations of the centers of gravity of the portions corresponding to the rigid bodies with respect to the reference point on the basis of time-series data of the positions of centers of gravity of the portions corresponding to the rigid bodies; sequentially determining accelerations of the centers of gravity of the portions corresponding to the rigid bodies in the absolute coordinate system from the accelerations of the centers of gravity of the portions corresponding to the rigid bodies in relation to the reference point and the accelerations of the reference point in the absolute coordinate system; sequentially determining angular accelerations of the portions corresponding to the rigid bodies on the basis of time-series data of the tilt angles of the portions corresponding to the rigid bodies; and sequentially determining estimated positions of points of application of floor reaction forces of the leg bodies in the bipedal walking body on the basis of at least either tilt angles of thighs of the leg bodies as the portions corresponding to the rigid bodies of the bipedal walking body or bending angles of knee joints of the leg bodies, wherein moments acting on at least one joint of the leg bodies of the bipedal walking body are estimated on the basis of an inverse dynamic model by using the estimated values of the floor reaction forces, estimated positions of points of application of floor reaction forces, the accelerations of the centers of gravity of the portions corresponding to the rigid bodies and the angular accelerations of the portions corresponding to the rigid bodies in the absolute coordinate system, the tilt angles of the portions corresponding to the rigid bodies, the pre-acquired weights and sizes of the portions corresponding to the rigid bodies, the pre-acquired positions of the centers of gravity of the portions corresponding to the rigid bodies in the respective portions corresponding to the rigid bodies, and pre-acquired inertial moments of the portions corresponding to the rigid bodies.

11. A method of estimating moments acting on at least one joint of each leg body of the bipedal walking body by using estimated values of floor reaction forces related to each leg body that have been sequentially determined according to the method of estimating floor reactions of a bipedal walking body described in claim 6, comprising the steps of: sequentially determining tilt angles in the absolute coordinate system of respective portions corresponding to rigid bodies of the bipedal walking body that are associated with respective rigid bodies of a rigid body link model on the basis of tilt angles of the torso, bending angles of the hip joints and the knee joints of the leg bodies, and the rigid body link model; sequentially determining positions of centers of gravity of the portions corresponding to rigid bodies in relation to the reference point on the basis of tilt angles of the portions corresponding to the rigid bodies, pre-obtained weights of the portions corresponding to the rigid bodies, and positions of centers of gravity of the portions corresponding to the rigid bodies in the individual portions corresponding to the rigid bodies, and sequentially determining accelerations of the centers of gravity of the portions corresponding to the rigid bodies with respect to the reference point on the basis of time-series data of the positions of centers of gravity of the portions corresponding to the rigid bodies; sequentially determining accelerations of the centers of gravity of the portions corresponding to the rigid bodies in the absolute coordinate system from the accelerations of the centers of gravity of the portions corresponding to the rigid bodies in relation to the reference point and the accelerations of the reference point in the absolute coordinate system; sequentially determining angular accelerations of the portions corresponding to the rigid bodies on the basis of time-series data of the tilt angles of the portions corresponding to the rigid bodies; and sequentially determining estimated positions of points of application of floor reaction forces of the leg bodies in the bipedal walking body on the basis of at least tilt angles of thighs of the leg bodies as the portions corresponding to the rigid bodies of the bipedal walking body or bending angles of knee joints of the leg bodies, wherein moments acting on at least one joint of the leg bodies of the bipedal walking body are estimated on the basis of an inverse dynamic model by using the estimated values of the floor reaction forces, estimated positions of points of application of floor reaction forces, the accelerations of the centers of gravity of the portions corresponding to the rigid bodies and the angular accelerations of the portions corresponding to the rigid bodies in the absolute coordinate system, the tilt angles of the portions corresponding to the rigid bodies, the pre-acquired weights and sizes of the portions corresponding to the rigid bodies, the pre-acquired positions of the centers of gravity of the portions corresponding to the rigid bodies in the respective portions corresponding to the rigid bodies, and pre-acquired inertial moments of the portions corresponding to the rigid bodies.

---

Description

---

TECHNICAL FIELD

The present invention relates to a method of estimating floor reaction forces acting on leg bodies of a bipedal walking body, such as a human being, a bipedal walking robot or the like. The present invention further relates to a method of estimating moments acting on joints of a leg body of a bipedal walking body by using the aforesaid estimated values of the floor reaction forces.

BACKGROUND ART

To control an operation of, for example, a walking assisting apparatus for aiding a human being in walking or to control traveling motions of a bipedal walking robot, it is necessary to sequentially determine floor reaction forces acting on leg bodies of the human being or the bipedal walking robot (to be more specific, the forces from a floor that act on grounding portions of the leg bodies). Determining the floor reaction forces makes it possible to acquire moments or the like acting on joints of the leg bodies of the bipedal walking body, and to decide target auxiliary forces of the walking assisting apparatus or desired drive torques or the like of joints of the bipedal walking robot on the basis of the determined moments or the like.

As a technique for determining the floor reaction forces, one disclosed in, for example, Japanese Unexamined Patent Application Publication No. 2000-249570 has been known. According to this technique, a floor reaction force of each leg body is defined as a resultant value (linear combination) of a plurality of trigonometric functions having mutually different cycles of 1/n (n=1, 2, . . . ) of a walking cycle, because time-dependent change waveforms of floor reaction forces of each leg body periodically change during steady walking of a bipedal walking body. In this case, weighting factors of the trigonometric functions for combining the plurality of trigonometric functions use predetermined values preset for each bipedal walking body or values obtained by adjusting the preset predetermined values according to topography.

However, the foregoing technique is adapted to determine floor reaction forces of the leg bodies for one step or a plurality of steps of the bipedal walking body. For this reason, it is difficult to accurately determine floor reaction forces if the gait of the bipedal walking body sequentially changes. Furthermore, to enhance the accuracy of determined floor reaction forces, the weighting factors of the trigonometric functions must be set for each bipedal walking body or adjusted according to topology or the like. This makes it difficult to accurately determine floor reaction forces by minimizing influences of environments under which bipedal walking bodies move or individual differences among bipedal walking bodies.

There has been known, for example, a bipedal walking robot having force sensors, such as six-axis force sensors, attached to ankles and foot portions of each leg body to determine floor reaction forces on the basis of outputs of the force sensors. There has been known another technique whereby a bipedal walking body is walked on a force plate installed on a floor to determine floor reaction forces from the outputs of the force plate.

According to the technology using force sensors, however, in order to determine the floor reaction forces of human leg bodies, in particular, force sensors have to be attached to ankles and foot portions, so that the force sensors inconveniently interfere with walking in normal living environments. The technology using force plates allows floor reaction forces to be determined only in environments wherein the force plates are installed.

The present invention has been made with a view of the aforementioned background, and it is an object thereof to provide a method of estimating floor reactions that permits floor reaction forces to be accurately determined in real time by a relatively simple technique and that is ideally suited for determining floor reaction forces of human beings, in particular, as bipedal walking bodies.

It is a further object of the present invention to provide a method of estimating joint moments of a bipedal walking body that allows moments acting on joints, such knee joints, of leg bodies to be accurately determined in real time by using an estimated value of a floor reaction force thereof.

DISCLOSURE OF INVENTION

First, a basic concept of the method of estimating floor reaction forces of a bipedal walking body in accordance with the present invention will be explained.

Motion states of leg bodies of a bipedal walking body (motion states of the leg bodies during walking) include a single stance state in which only one leg body 2 (the front leg body in relation to the direction of travel in the figure) of both leg bodies 2 and 2 of a bipedal walking body 1 is in contact with the ground, as illustrated in FIG. 1(a), and a double stance state in which both leg bodies 2 and 2 are in contact with the ground, as shown in FIG. 1(b).

If a total floor reaction force acting on the two leg bodies 2 and 2 from a floor A is denoted by F, then the total floor reaction force F is equal to the floor reaction force acting on the leg body 2 in contact with the ground in the single stance state shown in FIG. 1(a), while it is a resultant force of floor reaction forces Ff and Fr acting on the two leg bodies 2 and 2, respectively, in the double

If a total floor reaction force acting on the two leg bodies 2 and 2 from a floor A is denoted by F, then the total floor reaction force F is equal to the floor reaction force acting on the leg body 2 in contact with the ground in the single stance state shown in FIG. 1(a), while it is a resultant force of floor reaction forces Ff and Fr acting on the two leg bodies 2 and 2, respectively, in the double stance state shown in FIG. 1(b). If components in an X-axis direction (horizontal direction in relation to the direction of travel of the bipedal walking body 1) and in a Z-axis direction (vertical direction) of an acceleration a of a center of gravity G0 of the bipedal walking body 1 in an absolute coordinate system Cf fixed with respect to the ground are denoted by ax and az, respectively, and components in the X-axis direction and the Z-axis direction of the total floor reaction force F acting on the bipedal walking body 1 are denoted by Fx and Fz, respectively, then an equation of motion of the center of gravity G0 (more specifically, an equation of motion related to translational motion of the center of gravity G0) is represented by an expression (1) below: .sup.T(Fx, Fz-Mg)=M.sup.T(ax, az) (1)

(where M: Weight of the bipedal walking body; g: Gravitational acceleration)

The parenthesized part .sup.T(, ) of both sides in expression (1) means a two-component vector. In the present description, the symbol .sup.T(, ) denotes a vector.

In other words, the equation of motion of the center of gravity G0 is a relational expression indicating that a product of the acceleration a of a center of gravity G0 and the weight M of the bipedal walking body 1 is equal to a resultant force of a gravity acting on the center of gravity G0 (=Mg) and the total floor reaction force F.

Therefore, if the acceleration a=.sup.T(ax, az) of the center of gravity G0 of the bipedal walking body 1 is determined, then an estimated value of the total floor reaction force F=.sup.T(Fx, Fz) can be obtained by expression (2) given below, using the aforesaid acceleration a, the value of the weight M of the bipedal walking body 1, and the value of the gravitational acceleration g: .sup.T(Fx, Fz)=M.sup.T(ax, az+g) (2)

In the single stance state shown in FIG. 1(a), the floor reaction force acting on the single leg body 2 in contact with the ground is equal to the foregoing total floor reaction force F, so that an estimated value of the floor reaction force F acting on the single leg body 2 will be obtained by expression (2).

In this case, the weight M necessary to obtain an estimated value of the floor reaction force F can be known in advance by measurement or the like. As will be discussed in detail hereinafter, the position of the center of gravity G0 and the acceleration a can be sequentially determined by a publicly known techniques or the like, using outputs of sensors, such as a sensor for detecting a bending angle (rotational angle) of each joint of the bipedal walking body 1, an accelerometer, a gyro sensor, and the like.

Meanwhile, regarding the double stance state shown in FIG. 1(b), a floor reaction force acting on the leg body 2 at the rear in relation to the direction of travel of the bipedal walking body 1 is expressed as Fr=.sup.T(Frx, Frz), while a floor reaction force acting on the leg body 2 at the front is expressed as Ff=.sup.T(Ffx, Ffz). At this time, as will be discussed in detail later, according to the knowledge of the inventors of the present application, the components Frx and Frz of the floor reaction force Fr acting on the rear leg body 2, in particular, exhibit characteristic changes having marked correlation with respect to the time elapsed from the start of the double stance state and the moving speed of the bipedal walking body 1. Therefore, preparing characteristic data representing such characteristics (maps, data tables, approximate function expressions, or the like) by performing various experiments, simulations and the like makes it possible to estimate the floor reaction force Fr acting on the rear leg body 2 on the basis of the characteristic data. The resulting force of the floor reaction force Fr acting on the rear leg body 2 and the floor reaction force Ff acting on the front leg body 2 provides the total floor reaction force F obtained by the foregoing expression (2). Hence, as shown by expression (3) below, an estimated value of the floor reaction force Ff acting on the front leg body 2 can be determined by subtracting the estimated value of the floor reaction force Fr acting on the rear leg body 2 determined on the basis of the characteristic data described above from the estimated value of the total floor reaction force F (vector subtraction). Ff=F-Fr=.sup.T(Fx-Frx, Fz-Frz) (3)

Accordingly, in the double stance state, if the moving speed of the bipedal walking body 1 and the time elapsed from the start of the double stance state are determined, then estimated values of the floor reaction forces Fr and Ff of the two leg bodies 2 can be determined, using the above measured values and an estimated value of the total floor reaction force F. In this case, the moving speed of the bipedal walking body 1 can be determined by, for example, using measurement data or the like of tilt angles of a thigh and a crus of each leg body 2, as will be described in detail hereinafter.

Based on the explanation above, the present invention will now be explained. To fulfill the foregoing objects, a method of estimating floor reaction of a bipedal walking body, that is, a method of estimating floor reaction forces acting on individual leg bodies of the bipedal walking body, includes a first step for determining whether a motion state of leg bodies of the bipedal walking body is a single stance state wherein only one of the leg bodies is in contact with the ground or a double stance state wherein both leg bodies are in contact with the ground; a second step for sequentially determining positions of the center of gravity of the bipedal walking body while also sequentially determining accelerations of the centers of gravity in an absolute coordinate system fixed with respect to the ground, using time-series data of the positions of the centers of gravity; a third step for sequentially determining estimated values of total floor reaction force on the basis of an equation of motion of the centers of gravity represented by a weight of the bipedal walking body, a gravitational acceleration, the accelerations of the centers of gravity, and a total floor reaction force, which is a resultant force of the floor reaction forces acting on the respective leg bodies; a fourth step for sequentially measuring time elapses from a start of the double stance state until an end thereof each time the double stance state begins; and a fifth step for measuring a moving speed of the bipedal walking body at least before each double stance state is begun. Furthermore, a method of the present invention comprises the steps of sequentially determining estimated values of the total floor reaction force as estimated values of floor reaction forces acting on a single leg body in contact with the ground when the bipedal walking body is in the single stance state, and sequentially determining estimated values of floor reaction forces acting on one of the two leg bodies, the one leg body being located at the rear side in relation to the direction of travel of the bipedal walking body, on the basis of characteristic data pre-established as indicative of characteristics of changes in floor reaction force acting on the above one leg body with respect to the elapsed time of the double stance state and moving speed of the bipedal walking body, and sequentially determine estimated values of the floor reaction forces acting on the other leg body when the bipedal walking body is in the double stance state by subtracting the determined estimated values of the floor reaction forces of the one leg body from the estimated values of the total floor reaction force.

According to the present invention, in the third step, estimated values of the total floor reaction force are sequentially determined (refer to expression (2)) according to the aforementioned equation of motion of the center of gravity of the bipedal walking body (refer to expression (1)) on the basis of values of the weight, the gravitational acceleration and the center of gravity acceleration of the bipedal walking body. Then, in the first step, it is determined whether the motion states of the leg bodies of the bipedal walking body are in the single stance state or the double stance state, and estimated values of the floor reaction forces are determined by techniques based on the respective support states. More specifically, in the single stance state of the bipedal walking body, an estimated value of the total floor reaction force directly provides an estimated value of the floor reaction force acting on a single leg body in contact with the ground. At this time, the floor reaction force acting on the leg body not in contact with the ground (the free leg) is "zero."

In the double stance state of the bipedal walking body, an estimated value of a floor reaction force acting on one leg body located at the rear side in relation to the direction of travel of the bipedal walking body is determined on the basis of the aforementioned characteristic data, using measurement data of time elapsed from a start of the double stance state and measurement data of the moving speed of the bipedal walking body. Furthermore, an estimated value of the floor reaction force acting on the other leg body is determined by subtracting the estimated value of the floor reaction force of the foregoing one leg body from an estimated value of the total floor reaction force determined in the third step (performing vector subtraction) (refer to expression (3)).

According to the present invention, an estimated value of the total floor reaction force is used to determine a floor reaction force acting on each leg body in both the single stance state and the double stance state. The weight of the bipedal walking body necessary to determine an estimated value of the total floor reaction force may be obtained beforehand by measurement or the like. The positions of center of gravity and the accelerations of the bipedal walking body can be determined in real time, using data of outputs of relatively small sensors that can be easily attached to a bipedal walking body, such as a sensor (potentiometer or the like) for detecting bending angles (rotational angles) of each joint of the bipedal walking body, and an accelerometer, a gyro sensor, and the like.

Thus, the method of estimating floor reactions in accordance with the present invention makes it possible to determine floor reaction forces in real time by a relatively simple technique without the need for attaching force sensors to ankles and foot portions of a bipedal walking body or using a force plate.

In the method for estimating floor reaction forces in accordance with the present invention, the characteristic data is, for example, the data indicating a relationship between a ratio of a floor reaction force of the one leg body to the total floor reaction force at a start of the double stance state and a ratio of the elapsed time to a duration from a start to an end of the double stance state, a duration of the double stance state is estimated from a measured value of the moving speed on the basis of a pre-established correlation between the moving speed of the bipedal walking body and the duration of the double stance state, and an estimated value of a floor reaction force acting on the one leg is sequentially determined on the basis of the estimated value of the duration of the double stance state, a measured value of the elapsed time, an estimated value of the total floor reaction force at the start of the double stance state, and the characteristic data.

More specifically, according to the knowledge of the inventors of the present application, if attentions are focused on the relationship between a ratio of a floor reaction force of the one leg body (the leg body at the rear side in relation to the direction of travel of the bipedal walking body) with respect to the total floor reaction force at the start of the double stance state and a ratio of the elapsed time with respect to a duration from a start to an end of the double stance state, then the ratio of the floor reaction force exhibits a substantially constant change with respect to the ratio of the elapsed time, regardless of the moving speed or the like of the bipedal walking body. Hence, characteristic data can be easily set by defining the characteristic data as the data representing the relationship between the ratio of floor reaction forces and the ratio of elapsed time described above.

In this case, the duration of the double stance state, which provides a reference of the ratio of elapsed time, changes with the moving speed of the bipedal walking body. If the moving speed remains constant, then the duration of the double stance state will be substantially constant. Accordingly, establishing beforehand the correlation between the moving speed and the duration of the double stance state makes it possible to properly estimate the duration of the double stance state on the basis of the correlation from a measured value of the moving speed. The total floor reaction force, which provides a reference of the ratio of floor reaction force acting on the one leg body, is estimated on the basis of an equation of motion of the center of gravity of the bipedal walking body.

Therefore, to estimate the floor reaction forces of the respective leg bodies in the double stance state, as described above, the characteristic data is established as the correlation data between a ratio of the floor reaction forces of the one leg body based on the total floor reaction force at the start of the double stance state and a ratio of the elapsed time based on the duration of the double stance state, and an estimated value of the duration of the double stance state, which is determined from a measured value of the moving speed, a measured value of the elapsed time of the double stance state, and an estimated value of the total floor reaction force at the start of the double stance state are applied with the aforesaid characteristic data. This makes it possible to properly determine with high accuracy an estimated value of the floor reaction force acting on the one leg body.

Floor reaction forces are vector quantities. To be more specific, therefore, a ratio of the floor reaction force of the one leg body to the total floor reaction force is a ratio of each component of the floor reaction force in a certain appropriate coordinate system (e.g., a coordinate system fixed with respect to a floor on which a bipedal walking body moves).

A method of estimating floor reactions in accordance with the present invention includes, to measure a moving speed of a bipedal walking body in the fifth step, the steps of measuring a tilt angle of a crus under a knee joint of each leg body of the bipedal walking body and a tilt angle of a thigh between a hip joint and a knee joint of the leg body at a start of each double stance state, a step for calculating, at a start of each double stance state, a shift amount of a position of a bottom end portion of the crus of the leg body with respect to the hip joint of the leg body existing at the rear side in relation to the direction of travel of the bipedal walking body, the shift taking place in the direction of travel of the bipedal walking body from the start of the preceding double stance state, on the basis of measured values of tilt angles of the thigh and the crus of the leg body and predetermined sizes of the thigh and the crus of the leg body, and measuring the time elapsed from a start of each double stance state to a start of the next double stance state as elapsed time for one step. In the fifth step, each time the double stance state begins, the shift amount calculated at the start is divided by the one-step elapsed time measured from a start of the preceding double stance state to a start of the present double stance state so as to determine a measured value of the moving speed.

More specifically, an amount of a shift in the position of a bottom end portion of the crus of the leg body with respect to the hip joint of the leg body (this leg body being the one in contact with the ground in the single stance state immediately before each double stance state begins) existing at the rear side in relation to the direction of travel of the bipedal walking body at a start of each double stance state, the shift taking place in the direction of travel of the bipedal walking body from the start of the preceding double stance state will be the amount of distance over which the hip joint of the leg body (a proximal end portion of the leg body) has moved per step from a start of the preceding double stance state to a start of the present double stance state. Accordingly, an average moving speed of the bipedal walking body during the one-step elapsed time will be determined by dividing the aforesaid shift amount by the aforesaid one-step elapsed time, which is the time elapses from the start of the preceding double stance state to the start of the present double stance state. In this case, the shift amount can be determined by geometric computation using measured values of tilt angles of the thigh and the crus, respectively, of the leg body, which exists at the rear side when the double stance state begins, the tilt angles being measured at the starts of the present and the preceding double stance states, and data of the sizes (to be more specific, the lengths) of the thigh and the crus. Furthermore, the tilt angles of the thigh and the crus of each leg body can be measured using a sensor for detecting bending angles of joints of leg bodies or other sensors, such as an accelerometer, and a gyro sensor.

Thus, the moving speed of the bipedal walking body that is required for determining an estimated value of a floor reaction force acting on each leg body in the double stance state can be easily measured without using a large sensor or the like. Moreover, the moving speed measured is the moving speed immediately before a start of each double stance state, so that the reliability can be improved, eventually permitting higher accuracy of estimated values of floor reaction forces in the double stance state to be achieved.

Furthermore, the method of estimating floor reactions in accordance with the present invention comprises the steps of sequentially measuring vertical accelerations of a lower portion of a torso, which is adjacent to a hip joint, the torso being supported on the two leg bodies through the intermediary of hip joints of the leg bodies, wherein a motion state of the bipedal walking body is determined in the first step as follows. It is determined as the beginning of the double stance state and the end of the single stance state when the vertical acceleration of the lower portion of the torso increases to a predetermined threshold value or more, whereas it is determined as the end of the double stance state and the beginning of the single stance state when an estimated value of the floor reaction force acting on the leg body located at the rear side in relation to the direction of travel of the bipedal walking body in the double stance state decreases to a predetermined threshold value or less.

Alternatively, especially regarding the determination of an end of the double stance state (a start of the single stance state), according to the present invention for estimating the duration of the double stance state, as described above, a motion state of the bipedal walking body may be determined that the double stance state ends and the single stance state begins when a measured value of elapsed time from a start of the double stance state reaches an estimated value of a duration of the double stance state.

More specifically, while the bipedal walking body is moving (walking), when a motion state of the leg body switches from the single stance state to the double stance state, a vertical acceleration (upward acceleration) of the lower portion of the torso temporarily increases by a great extent when a leg body on a free leg side touches the ground. This phenomenon usually does not take place in other motion states of the leg bodies. Hence, comparing the acceleration with a predetermined threshold value (relatively large threshold value) makes it possible to accurately determine an end of the single stance state and a start of the double stance state.

When the motion states of leg bodies switch from the double stance state to the single stance state, the moment one leg body leaves a floor, the floor reaction force acting on that leg body reduces to zero. Hence, comparing the floor reaction force acting on the leg body with a predetermined threshold value (e.g., a threshold value slightly larger than zero) makes it possible to properly determine an end of the double stance state and a start of the single stance state. Especially when estimating a duration of the double stance state, the time elapsed from a start of the double stance state reaches an estimated value of the duration when a motion state of the leg body switches from the double stance state to the single stance state, thus allowing the end of the double stance state and the start of the single stance state to be accurately determined.

As discussed above, determining a motion state of leg bodies makes it possible to accurately determine whether the motion state is the single stance state or the double stance state. As a result, it is possible to switch, at a precise timing, between methods for calculating estimated values of floor reaction forces that differ between the single stance state and the double stance state, leading to higher accuracy of the estimated values of floor reaction forces. A vertical acceleration of a lower portion of a torso that is necessary to determine motion states of the leg bodies can be easily known from outputs of, for example, an accelerometer attached to the lower portion of the torso.

In the case of a human being or the like, in which a torso has a waist connected to both leg bodies through the intermediary of hip joints and a chest located above the waist such that it can be inclined with respect to the waist, the vertical acceleration of the lower portion of the torso to be measured is preferably a vertical acceleration of the waist.

In a method for estimating floor reactions in accordance with the present invention, various methods are conceivable and various publicly known methods can be used for determining a position of the center of gravity and an acceleration of the center of gravity of a bipedal walking body in the aforesaid second step. It is preferred, however, to determine the position of the center of gravity and the acceleration of the center of gravity according to the following method.

The method comprises the steps of sequentially measuring tilt angles of a torso supported on the two leg bodies through the intermediary of hip joints of the respective leg bodies, bending angles of at least hip joints and knee joints of the respective leg bodies, and accelerations of a predetermined reference point of the bipedal walking body in the absolute coordinate system, wherein in the second step, positions of the center of gravity of the bipedal walking body with respect to the reference point are sequentially determined on the basis of a tilt angle of the torso, bending angles of the hip joints and the knee joints, a rigid body link model representing the bipedal walking body in the form of a link assembly of a plurality of rigid bodies, pre-acquired weights of individual portions of the bipedal walking body, which correspond to individual rigid bodies of the rigid body link model, and the pre-acquired positions of the centers of gravity of the portions corresponding to rigid bodies in the respective portions corresponding to rigid bodies, accelerations of the center of gravity with respect to the reference point are sequentially determined on the basis of time-series data of the positions of center of gravity, and accelerations of the center of gravity in the absolute coordinate system are determined from the accelerations of the center of gravity with respect to the reference point and the accelerations of the reference point in the absolute coordinate system.

When a reference point is arbitrarily set in the bipedal walking body, a position of center of gravity of the bipedal walking body with respect to the reference point is generally decided by a mutual relationship among postures of a torso, thighs from hip joints to knee joints of respective leg bodies, and cruses under the knee joints. The posture relationship can be known from measurement data obtained by measuring tilt angles of the torso and bending angles of the hip joints and the knee joints. As will be discussed in detail hereinafter, if the aforesaid rigid body link model (e.g., a model assuming that an upper portion (including a torso) of the hip joints of the two leg bodies of the bipedal walking body, the thighs of the respective leg bodies, and the cruses as rigid bodies) is assumed, then a position of the center of gravity of the bipedal walking body with respect to the reference point can be determined on the basis of the weights of the portions corresponding to the rigid bodies of the bipedal walking body, the positions of the centers of gravity of the portions corresponding to the rigid bodies in the portions corresponding to the rigid bodies (to be more specific, the positions of the portions corresponding to the rigid bodies in a coordinate system fixed for the portions corresponding to the rigid bodies), and the aforementioned posture relationship. Furthermore, the accelerations of the center of gravity with respect to the reference point can be determined as two-level differential values of the positions of center of gravity obtained from the time-series data of the positions of center of gravity. Hence, by measuring the accelerations of the reference point in the absolute coordinate system beforehand, the accelerations of the center of gravity of the bipedal walking body in the absolute coordinate system are determined in terms of resultant accelerations of the accelerations of the center of gravity with respect to the reference point and the accelerations of the reference point.

In this case, the tilt angles of the torso necessary to determine the accelerations of the bipedal walking body can be determined from outputs of an accelerometer and a gyro sensor attached to the torso and another sensor, such as an inclinometer, as described above, and the bending angles of the hip joints and knee joints of the respective leg bodies can be determined from outputs of a sensor, such as a potentiometer, attached to the locations of the respective joints. Furthermore, the accelerations of the reference point in the absolute coordinate system can be obtained from outputs of a sensor, such as an accelerometer, attached to a portion integral with the reference point. In addition, the weights of the portions corresponding to individual rigid bodies of the bipedal walking body and the positions of the center of gravity of the portions corresponding to rigid bodies in the portions corresponding to individual rigid bodies can be obtained by measurement or the like in advance.

Thus, the positions of center of gravity and accelerations of a bipedal walking body can be easily determined in real time without the need for attaching a relatively large sensor or the like to the bipedal walking body.

Preferably, the reference point is set in the torso when determining the positions of center of gravity and accelerations of a bipedal walking body, as described above. This makes it possible to attach a sensor, such as an accelerometer, for measuring accelerations of the reference point in the absolute coordinate system to the torso, so that the number of sensors attached to leg bodies can be reduced, making it possible to prevent the sensors from interfering with a walking motion of the bipedal walking body.

In the case of a human being or the like having a waist connected to the two leg bodies through the intermediary of hip joints, and a chest located on the waist such that it may be tilted with respect to the waist, tilt angles of the torso used to determine the position of center of gravity are preferably tilt angles of the waist and the chest, respectively. Especially in this case, the rigid body link model is preferably a model that represents cruses under knee joints of the leg bodies of the bipedal walking body, thighs between the knee joints and the hip joints, the waist, and a body including the chest located on the waist, as respective rigid bodies.

With this arrangement, if a bipedal walking body is a human being, positions of center of gravity and accelerations thereof can be accurately determined, thus permitting higher accuracy of estimated values of floor reaction forces to be achieved.

A method of estimating joint moments of a bipedal walking body in accordance with the present invention is a method for estimating a moment acting on at least one joint of respective leg bodies of the bipedal walking body by using estimated values of floor reaction forces regarding the leg bodies sequentially determined by the method of estimating floor reactions in accordance with the present invention described above. The method of estimating joint moments in accordance with the present invention comprises the steps of sequentially measuring tilt angles of a torso supported on the two leg bodies through the intermediary of hip joints of the respective leg bodies, bending angles of at least hip joints and knee joints of the respective leg bodies, and an acceleration in the absolute coordinate system of a predetermined reference point of the bipedal walking body, sequentially determining tilt angles of portions corresponding to rigid bodies of the bipedal walking body that are associated with respective rigid bodies of a rigid body link model on the basis of the tilt angles of the torso, the bending angles of the hip joints and the knee joints of the leg bodies, and the rigid body link model representing the bipedal walking body in the form of a link assembly of a plurality of rigid bodies, sequentially determining positions of centers of gravity of the portions corresponding to rigid bodies in relation to the reference point on the basis of the tilt angles of the portions corresponding to the rigid bodies, pre-acquired weights of individual portions of the bipedal walking body, which correspond to individual rigid bodies of the rigid body link model, and pre-acquired positions of the centers of gravity of the portions corresponding to rigid bodies in the individual portions corresponding to rigid bodies, and for sequentially determining accelerations of the centers of gravity of the portions corresponding to rigid bodies in relation to the reference point on the basis of time-series data of the positions of centers of gravity of the portions corresponding to the rigid bodies, sequentially determining accelerations of the centers of gravity of the portions corresponding to the rigid bodies in the absolute coordinate system from the accelerations of the centers of gravity of the portions corresponding to the rigid bodies in relation to the reference point and the accelerations of the reference point in the absolute coordinate system, sequentially determining angular accelerations of the portions corresponding to the rigid bodies on the basis of time-series data of the tilt angles of the portions corresponding to the rigid bodies, and sequentially determining estimated positions of points of application of floor reaction forces of the leg bodies in the bipedal walking body on the basis of at least either tilt angles of thighs of the leg bodies as the portions corresponding to the rigid bodies of the bipedal walking body or bending angles of knee joints of the leg bodies, wherein a moment acting on at least one joint of the leg bodies of the bipedal walking body is estimated on the basis of an inverse dynamic model by using the estimated values of the floor reaction forces, estimated positions of points of application of floor reaction forces, the accelerations of the centers of gravity of the portions corresponding to the rigid bodies and the angular accelerations of the portions corresponding to the rigid bodies in the absolute coordinate system, the tilt angles of the portions corresponding to the rigid bodies, pre-acquired weights and sizes of the portions corresponding to the rigid bodies, pre-acquired positions of the centers of gravity of the portions corresponding to the rigid bodies in the respective portions corresponding to the rigid bodies, and pre-acquired inertial moments of the portions corresponding to the rigid bodies.

If, as described above, the second step in the method of estimating floor reactions includes a step for measuring a tilt angle of a torso, bending angles of hip joints and knee joints, respectively, of leg bodies, and an acceleration of a reference point of the bipedal walking body in the absolute coordinate system so as to determine a position of center of gravity, etc., of the bipedal walking body in relation to the reference point, then there is no need to measure them anew. As the rigid link model, the same rigid link model as that used for determining the position of the center of gravity, etc., of the bipedal walking body may be used.

According to the method of estimating joint moments in accordance with the present invention, measurement data obtained by measuring tilt angles of a torso and bending angles of hip joints and knee joints makes it possible to determine tilt angles of the portions corresponding to the rigid bodies of the bipedal walking body, including the torso, thighs and cruses (representing a mutual posture relationship among the portions corresponding to the rigid bodies). Positions of the centers of gravity of the portions corresponding to the rigid bodies in relation to the reference point can be determined on the basis of the weights of the portions corresponding to the rigid bodies, the positions of the centers of gravity of the portions corresponding to the rigid bodies in the respective portions corresponding to the rigid bodies (more specifically, positions of the portions corresponding to the rigid bodies in a coordinate system fixed with respect to the portions corresponding to the rigid bodies), and tilt angles of the portions corresponding to the rigid bodies. Furthermore, the accelerations of the center of gravity of the respective portions corresponding to the rigid bodies with respect to the reference point can be determined as two-level differential values of the positions of center of gravity obtained from the time-series data of the positions of center of gravity. Hence, by measuring the accelerations of the reference point in the absolute coordinate system beforehand, the accelerations of the centers of gravity of the portions corresponding to the rigid bodies of the bipedal walking body in the absolute coordinate system are determined in terms of resultant accelerations of the accelerations of the centers of gravity with respect to the reference point and the accelerations of the reference point (acceleration in the absolute coordinate system).

Furthermore, angular accelerations of the portions corresponding to the rigid bodies are determined as two-level differential values of tilt angles obtained from time-series data of tilt angles of the portions corresponding to the rigid bodies.

According to the knowledge of the inventors of the present application, positions of application points of floor reaction forces of leg bodies in a bipedal walking body, e.g., positions of application points of floor reaction forces of the leg bodies with respect to ankles of the leg bodies, are closely correlated to tilt angles of thighs of the leg bodies as the portions corresponding to the rigid bodies of the bipedal walking body and bending angles of knee joints of the leg bodies. Therefore, estimated positions of the application points of floor reaction forces in the bipedal walking body can be determined on the basis of at least either the tilt angles of the thighs of the leg bodies or the bending angles of the knee joints.

As described above, when the accelerations of the centers of gravity of the portions corresponding to the rigid bodies of the bipedal walking body, the angular accelerations of the portions corresponding to the rigid bodies, and estimated positions of the application points of the floor reaction forces have been determined, the data, including these values and estimated values of the floor reaction forces determined by the method of estimating floor reactions, can be used to estimate moments acting on knee joints and hip joints of the leg bodies on the basis of a publicly known so-called inverse dynamic model. In short, according to a method based on the inverse dynamic model, moments acting on the joints of a bipedal walking body that correspond to joints of a rigid body link model are determined in order, beginning with one closest to a floor reaction force application point, using an equation of motion related to translational