Demonstration of Locomotion with the Powered ProsthesisAMPRO
utilizing Online Optimization-Based Control
Huihua ZhaoMechanical EngineeringTexas A&M UniversityCollege
Station, USA
[email protected]
Jake ReherMechanical EngineeringTexas A&M UniversityCollege
Station, [email protected]
Jonathan HornMechanical EngineeringTexas A&M
UniversityCollege Station, [email protected]
Victor ParedesMechanical EngineeringTexas A&M
UniversityCollege Station, USA
[email protected]
Aaron D. Ames∗
Mechanical EngineeringTexas A&M UniversityCollege Station,
[email protected]
ABSTRACTThis demonstration presents an unimpaired subject
walk-ing with a custom built self-contained powered
transfemoralprosthesis: AMPRO, which is controlled by a novel
nonlin-ear real-time optimization based controller. To achieve
thebehaviors that will be demonstrated, controllers that havebeen
successfully implemented on bipedal walking robots aretranslated to
the prosthesis with the goal of achieving natu-ral human-like
walking while minimizing power consump-tion. To achieve this goal,
we begin by collecting refer-ence human locomotion data via
Inertial measurement Units(IMUs). This data forms the basis for an
optimization prob-lem that generates virtual constraints for the
prosthesis thatprovably yields walking in simulation. Utilizing
methodsthat have proven successful in generating stable robotic
loco-motion, control Lyapunov function (CLF) based
QuadraticPrograms (QPs) are utilized to optimally track the
result-ing desired trajectories. The parameterization of the
tra-jectories is determined through a combination of
on-boardsensing on the prosthesis together with IMU data,
therebycoupling the actions of the user with the controller.
Fi-nally, impedance control is integrated into the QP yieldingan
optimization based control law that displays remarkabletracking and
robustness, outperforming traditional PD andimpedance control
strategies.
1. BACKGROUNDThere are approximately 222, 000 people in the
United Statesalone that are transfemoral amputees [2]. Despite this
largeamputee population, the current market for commercial
trans-
∗Prof. Ames is with department of Mechanical Engineeringand
Electrical Engineering of Texas A&M University.
Permission to make digital or hard copies of part or all of this
work forpersonal or classroom use is granted without fee provided
that copies arenot made or distributed for profit or commercial
advantage and that copiesbear this notice and the full citation on
the first page. Copyrights for third-party components of this work
must be honored. For all other uses, contactthe Owner/Author.HSCC
’15 April 14 - 16, 2015, Seattle, WA, USACopyright is held by the
owner/author(s).ACM
978-1-4503-3433-4/15/04.http://dx.doi.org/10.1145/2728606.2728638
Figure 1: Transfemoral prosthetic: AMPRO.
femoral prosthesis remains largely limited to
energeticallypassive prosthetic devices, therefore, limiting the
day-to-day life of amputees with increased metabolic cost and
con-strained locomotion capabilities [5]. As one of the most
im-portant applications of bipedal robotic research,
poweredlower-limb prosthesis capable of providing net power in
con-junction with various prosthesis controllers have been
devel-oped in recent decades [3, 4]. However, despite the
improve-ments that these smart controllers have achieved, there
arestill limitations related to the optimality of the
controllersand the need for exhaustive clinical testing to
determine con-trol parameters. These issues motivate the main
objectivesof this work.
2. OBJECTIVEThe objectives of this work are twofold: a) to
propose themethod of using bipedal robots to test prosthetic
controllers.A nominal walking gait is found for the robot
platform
Figure 2: General procedure to obtain robust pros-thetic
walking.
which displays qualitatively human-like walking, and pros-thetic
controllers are tested on a “leg” of the robot. Throughthis method,
we are able to present and test a novel onlineoptimization-based
transfemoral prosthesis control method:control Lyapunov function
(CLF) based quadratic programs(QPs) coupled with variable impedance
control; b) havingverified the controllers on the robot platform,
we take thenext step to translate the complete methodology from
real-izing human-like robotic walking to achieve stable
prostheticwalking on a custom-built self-contained transfemoral
pros-thesis device: AMPRO. The primary goal of this demo is toshow
that the proposed optimal controller will yield stablehuman-like
prosthetic walking.
3. APPROACHMotivated by disadvantages of impedance control, a
novelprosthetic controller that combines the rapidly
exponentiallystabilizing control Lyapunov functions (RES-CLFs) [1]
withimpedance control is proposed with the goal of achievingbetter
tracking and improved energy efficiency on prosthe-sis. This
controller was first verified in simulation [6] thentested on a
bipedal robot platform: AMBER, which wasshown to be able to achieve
stable “prosthetic” walking [7].The controller is then realized on
a custom-built prostheticdevice: AMPRO. We begin with utilizing a
custom motioncapture system with Inertial Measurement Units (IMUs)
tocollect human locomotion trajectories. With the collecteddata, a
human-inspired optimization problem is then lever-aged to obtain a
stable and robust gait for a specific testsubject. IMUs are used to
estimate human movements dur-ing walking thus providing human
sensory feedback. Finally,the proposed controller is realized on
the prosthetic deviceAMPRO. An illustration of the entire process
demonstratedin this demo can be seen in Fig. 2.
4. RESULTSThrough the systematic methodology for translating
human-inspired robotic walking to prosthesis, stable robust
human-like prosthetic walking in both the laboratory and in
real-world environments is achieved. The resulted walking is
alsorobust to unknown obstacles. Gait tiles of a user walkingover
uneven terrain with obstacles is shown in Fig. 3. Theprosthesis has
been used to realize assisted walking contin-uously for 1 mile and
is able to walk 3 hours with a singlecharge. The proposed
controller also outperforms other ex-
Figure 3: Gait tiles of walking over obstacle.
isting controllers (such as PD) w.r.t. both tracking
(23%improvement) and power consumption (25% reduction) [8].
5. CONCLUSIONThe online optimization-based controller was
realized on theprosthetic device AMPRO experimentally with
improvedtracking and power consumption performance. The proce-dure
for testing this controller both in simulation and onthe bipedal
robot helped to predict and resolve many imple-mentation issues
before attempting to realize walking witha test subject. The
presented procedure, therefore, has thepotential to reduce the cost
of clinical testing of prosthesisthrough the efficient development
and testing of controllers.
6. REFERENCES[1] A. D. Ames, K. Galloway, and J. Grizzle.
Control
lyapunov functions and hybrid zero dynamics. InDecision and
Control (CDC), 2012 IEEE 51st AnnualConference on, pages
6837–42.
[2] T. Dillingham. Limb amputation and limb
deficiency:Epidemiology and recent trends in the united
states.Southern Medical Journal, 2002.
[3] J. Hitt, A. M. Oymagil, T. Sugar, K. Hollander,A. Boehler,
and J. Fleeger. Dynamically controlledankle-foot orthosis (dco)
with regenerative kinetics:incrementally attaining user
portability. In Roboticsand Automation, 2007 IEEE International
Conferenceon, pages 1541–1546. IEEE, 2007.
[4] F. Sup, A. Bohara, and M. Goldfarb. Design andControl of a
Powered Transfemoral Prosthesis. TheInternational journal of
robotics research,27(2):263–273, Feb. 2008.
[5] D. A. Winter. Biomechanics and motor control ofhuman gait:
normal, elderly and pathological. 1991.
[6] H. Zhao and A. Ames. Quadratic program based controlof
fully-actuated transfemoral prosthesis for flat-groundand up-slope
locomotion. In American ControlConference (ACC), 2014, pages
4101–4107, June 2014.
[7] H. Zhao, S. Kolathaya, and A. D. Ames. Quadraticprogramming
and impedance control for transfemoralprosthesis. In International
Conference on Robotics andAutomation (ICRA) 2014, June.
[8] H. Zhao, J. Reher, J. Horn, V. Paredes, and A. D.Ames.
Realization of nonlinear real-time optimizationbased controllers on
self-contained transfemoralprosthesis. In 6th International
Conference on CyberPhysics System, Seattle, WA, 2015.