Biomechanical Design of the Berkeley Lower Extremity Exoskeleton(BLEEX)
Adam B.Zoss,H.Kazerooni,Member,IEEE,and Andrew Chu
Abstract—Wheeled vehicles are often incapable of transporting heavy materials over rough terrain or up staircases.Lower extrem-ity exoskeletons supplement human intelligence with the strength and endurance of a pair of wearable robotic legs that support a payload.This paper summarizes the design and analysis of the Berkeley lower extremity exoskeleton(BLEEX).The anthropo-morphically based BLEEX has7DOF per leg,four of which are powered by linear hydraulic actuators.The selection of the DOF, critical hardware design aspects,and initial performance measure-ments of BLEEX are discussed.
Index Terms—Biomimetics,exoskeletons,mechatronics, robotics.
I.I NTRODUCTION
H EA VY objects are typically transported by wheeled vehi-
cles.However,many environments,such as rocky slopes and staircases,pose significant challenges to wheeled vehicles. Within these settings,legged locomotion becomes an attrac-tive method of transpo
rtation since legs can adapt to a wide range of extreme terrains.Berkeley’s lower extremity exoskele-ton(BLEEX)is thefirstfield-operational robotic system that can be worn by its operator and provides the ability to carry sig-nificant loads with minimal effort over any type of terrain.This is accomplished through four critical features:a novel control scheme,high-powered compact power supplies,special com-munication protocol and electronics,and a design architecture to decrease the complexity and power consumption.This paper focuses on the design architecture.
BLEEX comprises two powered anthropomorphic legs,a power supply and a backpack-like frame on which a variety of heavy payloads can be mounted(Fig.1).BLEEX provides load-carrying capability through legged locomotion guided by human interaction;but instead of actively“driving”the vehicle, BLEEX shadows the operator’s movement as he/she“wears”it like a pair of artificial legs.By combining the strength capabili-ties of robotics with the navigational intelligence and adaptabil-ity of humans,BLEEX allows heavy loads to be carried over rough,unstructured,and uncertain terrains.Possible applica-tions include helping soldiers,disaster relief workers,wildfire fighters,and other emergency personnel to carry major loads without the strain typically associated with demanding labor.
Manuscript received September15,2005;revised November30,2005.Rec-ommended by Guest Editor
s S.Agarwal and V.Krovi.This work was supported in part by Defense Advanced Research Projects Agency(DARPA)under Grant DAAD19-01-1-509.
A.B.Zoss and H.Kazerooni are with the Mechanical Engineering De-partment,University of California,Berkeley,CA94720USA(e-mail:azoss@ me.berkeley.edu;exo@berkeley.edu).
A.Chu is with Siemens,Concord,CA94520USA(e-mail:archmage.chu@ gmail).
Digital Object Identifier
10.1109/TMECH.2006.871087
Fig.1.Conceptual sketch of a lower extremity exoskeleton.Proper actuation
of the robotic legs removes the payload weight from the wearer,while allowing
the wearer to effortlessly control and balance the machine.
II.B ACKGROUND
In the late1960s,thefirst active exoskeletons were devel-
oped almost simultaneously at General Electric(GE)[1]and
the Mihajlo Pupin Institute in Belgrade[2].Safety concerns and
complexity prevented the Hardiman project at GE from walking
and the Belgrade exoskeleton followed only preprogrammed
walking motions.Both projects were tethered to a stationary
power source.
More recently,a lightweight power assist device,HAL,has
been developed at the Tsukuba University[3].Although the
device successfully walks and carries its own power supply,
it is designed to only assist the wearer’s muscles;it cannot
carry an external load.The Kanagawa Institute of Technology
has developed a full-body“wearable power suit,”powered by
unique pneumatic actuators[4].It has been demonstrated in
limited applications without a portable power supply.
Still in development are several other lower extremity ex-
oskeletons designed to aid disabled people[5]–[7].Similar to
exoskeletons,a variety of powered orthoses are being developed
for the knee[8],[9],lower back[10],and ankle[11].
The BLEEX project is an energetically autonomous exoskele-
ton capable of carrying its own weight plus an external payload.
All previous exoskeletons are either tethered to afixed power
supply or not strong enough to carry an external load.Unlike
orthoses and braces,BLEEX transfers the payload forces to the
ground,instead of the wearer.
III.E XOSKELETON C ONTROL
The BLEEX control algorithm ensures that the exoskeleton
shadows the operator with minimal interaction forces between 1083-4435/$20.00©2006IEEE
TABLE I
BLEEX E LECTRONICS R
replaceableEQUIREMENTS
the two.The highlight of the control scheme is that it is solely based on measurements from the exoskeleton;there are no direct measurements from the operator or from where the operator contacts the exoskeleton (e.g.,no force sensors between the two)[12].This eliminates the problems associated with measuring interaction forces or human muscle activity.
Through the control algorithm,BLEEX instantaneously shad-ows the wearer’s voluntary and involuntary movements.This re-quires the controller to be very sensitive to all forces and torques the operator imposes on the exoskeleton.To achieve this,the BLEEX control increases the closed loop system sensitivity to the operator’s forces and torques [12].
The BLEEX control scheme uses a full dynamic model of the exoskeleton and utilizes a significant number of sensors to solve the dynamic model and control its actuators;thus,BLEEX con-tains a large amount of electronics.Each actuated joint contains an encoder and a pair of linear accelerometers to determine the joint’s angle,angular velocity,and angular acceleration.An in-clinometer gives the overall orientation relative to gravity.Servo valves and single-axis force sensors provide control and feed-back of the actuation forces.Foot switches determine when the exoskeleton feet touch the ground and load distribution sensors determine how the operator distributes their own weight be-tween their two feet in double stance.All the sensors connect to a distributed network of electronic boards (remote I/O mod-ules or RIOMs),which then connect to a centralized controlling computer [13].Table I summarizes the amount of electronics required for the BLEEX control algorithm.
IV .D EGREES OF F REEDOM
To ensure maximum safety and minimum collisions with the environment and operator,the BLEEX architecture is almost anthropomorphic.This means the BLEEX leg is kinematically similar to a human’s,but does not include all of the degrees of freedom (DOF)of human legs.Additionally,the BLEEX degrees of freedom are all purely rotary joints.While the details of these joints differ from hum
an joints,BLEEX has hip,knee,and ankle joints,like in a person.Overall,BLEEX has seven distinct DOF per leg:a)3DOF at the hip;
b)1DOF at the knee (pure rotation in the sagittal plane);c)3DOF at the ankle.
It is natural to design a 3-DOF exoskeleton hip joint such that all three axes of rotation pass through the human ball and socket hip joint.However,going through the design of
several
Fig.2.BLEEX hip DOF (back view).Only the rotation axis does not pass through the human hip ball and socket joint.The adjustment bracket is replace-able to accommodate various sized
operators.
Fig.3.BLEEX ankle DOF.Only the flexion/extension axis passes through the human’s ankle joint.Abduction/adduction and rotation axes are not powered,but are equipped with appropriate impedances.
mock-ups and experiments we learned that these designs have limited ranges of motion and result in
singularities at some human hip postures.Therefore,the rotation joint was moved so that it does not align with the human’s hip joint.Initially the rotation joint was placed directly above each exoskeleton leg (labeled “alternate rotation”in Fig.2).This worked well for the lightweight plastic mock-up,but created problems in the full-scale prototype because the high mass of the torso and payload created a large moment about the unactuated rotation joint.Therefore,the current hip rotation joint for both legs was chosen to be a single axis of rotation directly behind the person and under the torso (labeled “current rotation”in Fig.2).The current rotation joint is typically a spring loaded toward its illustrated position using sheets of spring steel.
Similar to the human’s ankle,the BLEEX ankle also has 3DOF.The flexion/extension axis coincides with the human ankle joint.For design simplification,the abduction/adduction and rotation axes on the BLEEX ankle do not pass through the human’s leg and form a plane outside of the human’s foot (Fig.3).To take load off of the human’s ankle,the BLEEX ankle abduction/adduction joint is sprung toward vertical,but the rotation joint is completely free.
Additionally,the front of the exoskeleton foot,under the op-erator’s toes,is compliant to allow the exoskeleton foot to flex with the human’s foot (see Section IX-A).Since the human and
Fig.4.Human walking cycle[18].The walking cycle begins with the start of stance phase(foot on the ground)at heel strike followed by toe-off and swing phase (foot off the ground)beginning at∼60%of the cycle.
exoskeleton leg kinematics are not exactly the same(merely similar),the human and exoskeleton are only rigidly connected at the extremities(feet and torso).
V.C LINICAL G AIT A NAL YSIS(CGA)D A TA
A.Design by Biological Analogy
Each BLEEX leg has7DOF(besides the toeflexibility), but actuating all of them creates unnecessarily high-power con-sumption and control complexity.Instead,only joints that re-quire substantial power should be actuated.Since we intended to design an anthropomorphic exoskeleton with limb masses and inertias same as in a human,the required joint torques and power for the exoskeleton to perform a given motion were approximated as that required by a similarly sized human per-forming the same motion.Additionally,since the primary goal of a lower-extremity exoskeleton is locomotion,the joint power requirements for the BLEEX were determined by analyzing the walking cycle shown in Fig.4.
Human joint angles and torques for a typical walking cy-cle were obtained in the form of independently collected CGA data.CGA angle data is typically collected via human video motion capture.CGA torque data is calculated by estimating limb masses and inertias and applying dynamic equations to the motion data.Given the variations in individual gait and measur-ing methods,three independent sources of CGA data[14]–[16] were utilized for the analysis and design of the BLEEX.This data was further modified to yield estimates of exoskeleton ac-tuation requirements.The modifications included1)scaling the joint torques to a75-kg person(the projected weight of the BLEEX and its payload not including its wearer)and2)sum-ming the pelvic tilt angle(or lower back angle depending
on Fig.5.Angle and torque sign conventions.Each joint angle is measured as the positive counterclockwise displacement of the distal link from the proximal link(zero in the standing position).
data available)and the hip angle to yield a single angle between the torso and the thigh as shown in Fig.5.This accounted for the reduced DOF of the exoskeleton.The following sections describe the use of CGA data for exoskeleton design.The sign conventions used are shown in Fig.5.
B.Ankle
Fig.6shows the CGA ankle angle data for a75-kg human walking onflat ground at approximately1.3m/s vs.time.As can be seen in all the plots,heel strike occurs at∼0%of the step cycle and toe-off occurs at∼60%of the step cycle.
Fig.7shows the adjusted CGA data of the ankleflex-ion/extension torque.The ankle torque is almost entirely nega-tive,making unidirectional actuators an ideal actuation choice. This asymmetry also implies a preferred mounting orienta-tion for asymmetric actuators(one-sided hydraulic cylinders).
Fig.6.Adjusted CGA data of the ankle flexion/extension
angle.
Fig.7.Adjusted CGA data of the ankle flexion/extension torque.Peak negative torque is very large (−120N ·m)and occurs in late stance phase.The ankle torque during swing is quite small.
Conversely,if symmetric bidirectional actuators are considered,spring-loading would allow the use of low torque producing ac-tuators.Although the ankle torque is large during stance,it is negligible during swing.
The instantaneous ankle mechanical power (shown in Fig.8)is calculated by multiplying the joint angular velocity (derived from Fig.6)and the instantaneous joint torque (Fig.7).The ankle absorbs energy during the first half of the stance phase and releases energy just before toe-off.The average ankle power is positive,indicating that power production is required at the ankle.C.Knee
The knee angle in Fig.9is characterized by knee flexion to create a horizontal hip trajectory.The knee buckles momentarily in early stance to absorb the impact of heel strike then
undergoes
Fig.8.Adjusted CGA data of the ankle flexion/extension instantaneous me-chanical power.The average ankle power is positive,indicating the ankle does positive work and requires
actuation.
Fig.9.Adjusted CGA data of the knee flexion/extension angle.The minimum angle is −60◦,occurring in early-mid swing phase to enable the foot to clear the ground.
a large flexion during swing.This knee flexion decreases the effective leg length,allowing the foot to clear the ground when swinging forward.
The CGA-based knee actuation torque is shown in Fig.10.The required knee torque has both positive and negative com-ponents,indicating the need for a bidirectional actuator.The highest peak torque is extension in early stance (∼60N ·m);hence,asymmetric actuators should be biased to provide greater extension torque.
Fig.11shows the CGA knee power.Since the average power is negative,many prosthetic devices use power dissipative de-vices (i.e.,dampers)to mimic knee dynamics.However,the knee requires a large amount of positive power whenever the human is walking up an incline,or climbing stairs,so the BLEEX knee joint is actuated [17].
Fig.10.Adjusted CGA data of the knee flexion/extension torque.An initial −35N ·m flexion torque is required at heel strike,followed by large extension torques (∼60N ·m)to keep knee from buckling in stance
phase.
Fig.11.Adjusted CGA data of the knee flexion/extension instantaneous me-chanical power.The negative average indicates power dissipation.
D.Hip
Fig.12details the hip angle while walking.The thigh moves in a sinusoidal pattern with the thigh flexed upward at heel strike to allow foot–ground contact in front of the person.This is followed by an extension of the hip through most of stance phase and a flexion through swing.
The hip torque in Fig.13is relatively symmetric (−80to +60N ·m);hence,a bidirectional hip actuator is required.Negative extension torque is required in early stance as the hip supports the load on the stance leg.Hip torque is positive in late stance and early swing as the hip propels the leg forward during swing.In late swing,the torque goes negative as the hip decelerates the leg prior to heel strike.
Fig.14shows the instantaneous hip mechanical power.The hip absorbs energy during stance phase and injects it during
toe-
Fig.12.Adjusted CGA data of the hip flexion/extension angle.The hip has an approximately sinusoidal
behavior.
Fig.13.Adjusted CGA data of the hip flexion/extension torque.The hip torque is bidirectional.
off to propel the torso forward.The average power is positive,implying the need for active actuation.
Besides the flexion/extension joints at the ankle,knee,and hip,the other DOF in the human leg require substantially less mechanical power while walking [18].Therefore,only the three flexion/extension joints were initially actuated on BLEEX.Later,actuation was added to the hip abduction/adduction joint to improve lateral balance.CGA data shows that the hip ab-duction/adduction joint requires the most power for nonflex-ion/extension joints [18].
VI.R ANGE OF M OTION
The BLEEX kinematics are close to human leg kinematics,and therefore,the BLEEX joint ranges of motion are determined by examining human joint ranges of motion.At the very least,
Fig.14.Adjusted CGA data of the hip flexion/extension instantaneous me-chanical power.The average hip power is positive,indicating an actuator is required at the hip.
TABLE II
BLEEX J OINT R ANGES OF M
OTION
the BLEEX joint range of motion should be equal to the human range of motion during walking (see Table II),which can be found by examining CGA data [14]–[16].Safety dictates that the BLEEX range of motion should not be more than the operator’s range of motion (see Table II)[19].For each DOF,Table II also lists the BLEEX range of motion which is,in general,larger than the human range of motion during walking and less than the maximum range of human motion.
The most maneuverable exoskeleton should ideally have ranges of motion slightly less than the human’s maximum range of motion.However,BLEEX uses linear actuators (see Section VII)and so some of the joint ranges of motion are reduced to prevent the actuators’axes of motion from passing through the joint center.If this had not been prevented,the joint could reach a configuration where the actuator would be unable to produce a torque about its joint.Additionally,all the joint ranges of motion were tested and revised during prototype testing.For example,mock-up testing determined that the BLEEX ankle flexion/extension range of motion needs to be greater than the human ankle range of motion to accommodate the human foot’s smaller DOF not modeled in the BLEEX
foot.
Fig.15.Triangular configuration of a linear actuator about a rotary joint.The moment arm R is perpendicular to the actuator and varies with joint angle θ.
VII.A CTUA TOR S ELECTION AND S IZING
Hydraulic actuators have high specific power (ratio of actua-tor power to actuator weight),and thus are the smallest actuation option available.Also,hydraulic fluid is generally incompress-ible,leading to a relatively high-control bandwidth.BLEEX uses double-acting linear hydraulic actuators because of their compact size,low weight,and high-force capabilities.Rotary hydraulic actuators were not selected because they usually have either internal leakage or considerable friction.
In order to analytically determine the appropriate size of ac-tuators,an equation for the actuator’s torque capabilities must be determined.The maximum static pushing and pulling forces that a linear hydraulic actuator cylinder can supply is its cross-sectional area multiplied by the supply pressure (P S ).When a linear actuator is used to produce a torque about a rotary joint,its moment arm R changes as a function of the joint angle θ(Fig.15).The torque the actuator can produce is its peak force times that moment arm
T max push =P S
πD 2bore
4
R (θ)(1)T max pull =P S π D 2
bore −D 2
rod
4
R (θ)(2)
where D bore is the actuator bore diameter and D rod is the actu-ator rod diameter.
The problem of actuation design is to find a combination of actuator cross section,actuator endpoints,and supply pressure that provides the necessary torque but minimizes the hydraulic consumption.Since the BLEEX dynamics are close to human leg dynamics,the actuators must produce torque greater than the CGA data in order to walk.Additionally,the actuator must reach the desired ranges of motion while always being able to produce a minimum nominal torque.In general,there is not a unique solution,but many feasible possibilities.
To find a feasible actuator configuration,an actuator size (cross section,minimum length,and stroke),s
upply pressure,and one of the end-point positions were chosen for each joint.Assuming that the longest and shortest actuator lengths occur at the extremes of the joint’s range of motion,the second end-point position of the actuator can be calculated (Fig.16).The available actuator torque [(1)and (2)]can then be compared with the
版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。
发表评论