The Fourteenth Scandinavian International Conference on Fluid Power, May 20-22, 2015, Tampere, Finland
DESIGN OVERVIEW OF THE HYDRAULIC QUADRUPED ROBOTS
HyQ2MAX AND HyQ2CENTAUR
Claudio Semini, Jake Goldsmith, Bilal Ur Rehman, Marco Frigerio, Victor Barasuol,
Michele Focchi, Darwin G. Caldwell
Dept. of Advanced Robotics,
Istituto Italiano di Tecnologia (IIT),
via Morego, 30, 16163 Genova, Italy
Legged robots have not yet demonstrated the desired versatility and higher mobility that would justify their
more complicated design with respect to wheeled or tracked vehicles. To make these robots ready for real
world applications -- for example as assistants to humans in dangerous areas -- important challenges must
be solved first, such as dynamic locomotion over rough terrain, dynamic balancing after disturbances,
structural robustness to falls, self-righting (to get back up on the feet after falling), active or passive
compliance in the legs, state estimation, perception and optional dexterous manipulation. In this paper we
will focus on the robustness, self-righting and manipulation aspects. We will give an overview of the design
of two new hydraulic robots: HyQ2Max, an improved, robust version of our hydraulic quadruped HyQ, and
HyQ2Centaur, a centaur-style robot that combines the HyQ2Max locomotion platform with a pair of new
hydraulic manipulator arms. We will focus on the self-righting ability of the quadruped robot and present the
results of rigid-body dynamics simulations. Next, we will focus on the mechanical design concept of the new
compact hydraulic arms and discuss the hydraulic actuation system. To the authors’ best knowledge this is
the first time the design of a fully hydraulically actuated centaur robot is presented.
KEYWORDS: hydraulic actuation, hydraulic centaur, quadruped, legged robot, mechanical design
Research into legged robots is expected to result in vehicles that are able to navigate with agility on rough
terrain, exceeding the mobility of wheeled and tracked vehicles. However, despite the efforts of several
decades of research into legged robots, the current state of the art is still far from reaching this goal.
Recently a class of medium-sized, hydraulically actuated and torque-controlled quadrupedal robots (e.g.
Boston Dynamics' LS3 and BigDog  and IIT's HyQ ) have shown promising results of agile navigation
over flat and rough terrain and in presence of lateral disturbances . Such robots are expected to assist
humans for practical applications such as search and rescue, fire-fighting, forestry and
inspection/maintenance tasks in dangerous areas or where automation is required in unstructured
environments. Fundamental capabilities that such robots will need to have, are the following: dynamic
locomotion over rough terrain, dynamic balancing after disturbances, structural robustness to falls, self-
righting (to get back up on the feet after falling), active or passive compliance in the legs, state estimation
The HyQ project of the Istituto Italiano di Tecnologia (IIT) started in 2007 and resulted in the first version of
HyQ in 2011, a fully hydraulic, torque-controlled quadruped robot . Since then, HyQ has demonstrated a
wide repertoire of static and dynamic motions ranging from walking trot over flat, inclined and rough terrain
(indoors and outdoors), balancing under disturbances , flying trot , squat jumps, step reflexes ,
perception-enhanced trotting and crawling , to an optimized crawl gait for walking on stairs and stepping
stones . A summary video of these results is available online .
Figure 1. Pictures of the first version of HyQ robot and leg. Left: HyQ robot on outdoor test track (2013);
Right: first prototype of HyQ leg on vertical slider test bench (2008).
Based on our experiences with HyQ and earlier leg prototypes (Figure 1), we have been developing a
second version of the robot that improves upon the weaknesses of HyQ. It is our goal to develop a versatile
machine that can be used for real-world applications. We expect that there will be situations in which the
robot loses balance and falls. To allow the robot to continue with its operation, it is fundamental that it is
robust to such impacts and that it can self-right after a fall. We therefore entirely redesigned the legs and the
torso to increase the robot’s robustness, extend its joint range of motion and increase its joint torque. This
resulted in a new robot called HyQ2Max. Furthermore, we believe that a versatile quadruped robot used for
real-world applications needs to have the option to mount a pair of dexterous arms to allow it to perform
manipulation tasks. Therefore, we have been developing compact hydraulic manipulator arms that can be
mounted on HyQ2Max, turning the quadruped robot into a centaur-like machine called HyQ2Centaur. A
quadruped locomotion platform with two arms combines the stability and agility of four legs with the dexterity
and functionality of a two-arm system.
This paper gives an overview of the design and hydraulic system of HyQ2Max and HyQ2Centaur. We will
focus on the self-righting ability of the new quadruped robot and present the results of rigid-body dynamics
simulations. To the authors’ best knowledge this is the first time the design of a fully hydraulically actuated
centaur robot is presented.
This paper first discusses the state of the art in the field of hydraulic quadruped machines and centaur-style
robots. Section 3 then introduces the new hydraulic quadruped robot HyQ2Max, presenting the concept of its
mechanical design, and the results of a simulated self-righting motion. Section 4 gives an overview of the
design concept of HyQ2Centaur, focussing on the design of the new hydraulic arms. The hydraulic system of
the robots is presented in Section 5. Finally, Section 6 discusses open problems and concludes the paper
with final remarks.
2. RELATED WORK
This section discusses the state of the art in the field of hydraulically actuated quadruped robots and
2.1. Hydraulically actuated quadruped robots
Robotics research has resulted in a big variety of quadruped robots, most of them actuated by electric
motors. A much smaller number is powered by hydraulic actuators. This section presents the most important
In the 1960s, General Electric developed a four-legged walking truck that weighed over 1300kg. It was
hydraulically actuated and controlled by a human operator. Each limb of the operator was controlling one of
the robot’s four legs through an interface with force feedback. After about 20 hours of training an operator
was able to control the machine to walk, climb a stack of railroad ties and push a jeep out of the mud .
In the 1980s, Marc Raibert and colleagues constructed several hydraulically actuated legged robots, among
which also a quadruped robot. The robot had four prismatic legs with 3 hydraulic joints each and a
pneumatic spring at the end of the leg. It was able to trot, pace and bound on flat ground . More
recently, Raibert and his team at Boston Dynamics constructed several other hydraulic quadruped robots:
BigDog , LS3, cheetah and wildcat. These robots clearly raised the bar of what is possible. However, very
little information on the robot hardware, hydraulics and control has been published.
Shigeo Hirose’s Titan XI is a large size hydraulically actuated quadruped robot. The 7000kg robot is
designed for construction work on slopes. . Statically stable walking on flat and inclined terrain has been
IIT’s HyQ robot is an 80kg hydraulic quadruped that was first presented in 2010 in Claudio Semini’s PhD
thesis . Since 2011 the robot is fully torque controlled and has demonstrated a wide repertoire of motions
ranging from highly dynamic motions to carefully planned navigation over rough terrain. For a more detailed
introduction on HyQ see Section 1.
A number of hydraulic quadruped robots have been developed in Korea and China in the last years. For
example, the P2 robot  and Jinpoong developed by KITECH, SCalf by Shandong University  and
BabyElephant by SJTU .
2.2. Centaur-style robots
While humanoid and quadruped robots are very popular among researchers in the field of robotics, a
combination of the two has rarely been investigated. This section presents the state of the art in the field of
The first known centaur robot was developed by a Japanese consortium of industry and universities from
1984-1993 as part of the ART project. The project was focussing on the development of several types of
nuclear inspection machines, including an electric centaur-style robot . A few years later, KIST presented
their centaur robot with hydraulic legs and electric upper body . The robot stood 1.8 meters tall and
weighed 150 kg. More recently, Tsuda et al. presented a few papers on a small centaur robot that is
actuated by electric RC servomotors . Several other centaur-style robots were constructed with wheels at
the end of their legs (e.g. WorkPartner , NASA centaur 2 ). Even though not a full centaur, it is worth
mentioning that in 2013 a video of BigDog with one manipulator arm throwing a cinder block was published
3. HyQ2Max ROBOT DESIGN
This section introduces the design concept of the HyQ2Max robot, shows the results of a study on self-
righting and presents an overview of possible future application scenarios of this robot.
3.1. HyQ2Max Design Concept
The HyQ2Max robot (Figure 2) is an improved version of the hydraulic quadruped robot HyQ . The main
improvements are increased reliability and robustness of the robot’s hardware, larger joint range of motion
and higher joint output torque, as explained next.
Reliability and robustness against impacts and dirt are fundamental requirements for a legged vehicle
performing real-world tasks. HyQ2Max is designed to be robust against impacts and dirt. All sensitive parts
like sensors, valves, actuators and electronics are protected inside the structure. The torso is constructed
with a frame made of a strong aerospace-grade aluminium alloy (7000 series), tubular roll frames in the front
and back, and light-weight glass fibre/Kevlar covers that protect the onboard computer and hydraulics. The
four legs are built of the same aluminium alloy as the torso. The upper leg consists of two rugged halves
forming a shell that acts as protection and structural element. The lower leg is made of a light-weight yet
robust aluminium tube.
Figure 2. CAD of HyQ2Max robot and photo of single leg. Left: CAD model of the HyQ2Max robot with
explanation. The three leg joints are labelled HAA (hip abduction/adduction), HFE (hip flexion extension) and
KFE (knee flexion/extension); Right: photo of leg prototype of HyQ2Max attached to a vertical slider test
bench for experiments.
The joints’ range of motion of a legged robot determines the size of the workspace of its feet. The larger this
workspace, the more versatile motions can be implemented on the robot. A large workspace is especially
important for self-righting motions as explained in Section 3.2. Table 1 compares the joint range of motion of
HyQ and HyQ2Max and Figure 3 confronts the two different workspaces in the leg’s X-Z plane.
Table 1. Comparison of the main specifications of HyQ and HyQ2Max
Description HyQ HyQ2Max
Number of actuated joints 12 12
Joint range of motion (HAA, HFE, KFE) 90°, 120°, 120° 80°, 270°, 160°
Peak joint torque (HAA, HFE, KFE) @ 20MPa 120Nm, 181Nm, 181Nm 120Nm, 245Nm, 250Nm
Upper, lower leg segment lengths 0.35m, 0.35m 0.36m, 0.38m
Robot weight (offboard power supply) 80kg 80kg
It can be clearly seen, that HyQ2Max has a larger foot workspace than HyQ, leading to (1) faster running
since the step length can be increased, (2) self-righting ability since the leg can be moved completely up
above the robot’s center of mass (see Section 3.2), (3) a rest position of the robot by retracting the legs until
the bottom of the torso touches the ground and (4) an increased number of footholds for climbing motions
with foothold planning.
Figure 3. Comparison of the leg workspaces of HyQ (blue) and HyQ2Max (green line) in the X-Z plane (left).
HyQ2Max leg drawing in the X-Z plane illustrating the angle convention and leg coordinate frame (right).
Table 1 also shows that the HFE and KFE joints of the new robot have a higher joint output torque. This is
important for self-righting (see Section 3.2), carrying payload and for more agile motions. The HAA joint is
actuated by a double-vane rotary actuator, the HFE joint by a single-vane rotary actuator and the KFE joint
by a cylinder connected to a four-bar linkage.
3.2. Self-Righting study
As mentioned in the introduction, the self-righting capability is fundamental for a real-world legged robot,
since it is unavoidable that the robot falls during its operation on challenging terrain. We therefore
implemented a self-righting sequence (see Figure 4) and simulated it inside our rigid body dynamics
simulator SL . All kinematics and dynamics calculations are implemented with efficient C++ code,
automatically generated by the robot code generator RobCoGen .
Figure 4. Self-righting sequence of HyQ2Max shown with CAD renderings: from top left to bottom right.
The joint angle and torque plots of this simulation are shown in Figure 5. The different steps of the self-
righting sequence are illustrated with different colours. The thin black lines show the limits of joint angle and
torques as specified in Table 1. Note that the torque limits of the KFE joint depends on the KFE joint angle
since the four-bar linkage creates a nonlinear torque output profile For a detailed discussion on such output
profiles, refer to . The figure shows that all values stay inside their limits during the entire motion.
Figure 5. Simulation results of self-righting motion showing joint angles and torques for the left front (LF) and
right front (RF) leg. The different colours indicate the different steps of the self-righting sequence. The black
dashed line shows the joint angle and torque limits. Left: joint angle vs. time plots of the hip flexion/extension
(HFE) and knee flexion/extension (KFE) joints. Right: joint torques vs. time plots of the same joints.
3.3. Possible future application concepts
HyQ2Max is designed to be the light-weight, high-performance version of this new four-legged vehicle. In the
future, the robot’s hardware and configuration can be customized to match the requirements of the desired
application. Figure 6 shows the concept of the robot applied to a range of possible future tasks. Task-specific
features range from radiation-hardened hardware (e.g. nuclear decommissioning) to specific onboard
sensing (e.g. inspection) and manipulation capability (e.g. maintenance, decommissioning). The next section
will discuss our current efforts to add manipulation capability to HyQ2Max.
Figure 6. HyQ2Max application scenarios. From left to right: construction, fire and rescue, forestry industry,
inspection and maintenance, nuclear decommissioning.
4. HyQ2CENTAUR ROBOT DESIGN
Future quadruped robots operating in real-world applications will most likely need to manipulate objects in
the environment at some point, e.g. through a pair of dexterous arms. A centaur-style robot consists of a
quadruped locomotion platform and a pair of arms. It thus combines the advantages of a stable four-legged
base with the dexterity of a two arm system.
This section presents the design of HyQ2Centaur, which is a combination of HyQ2Max and a pair of arms.
We will first give an overview of the design of a pair of custom-built, light-weight hydraulic arms that can be
mounted onto HyQ2Max. Then we will present the concept of the centaur robot design and possible future
application scenarios of the centaur robot.
4.1. Hydraulic arm design
The most important requirements for a dual arms system mounted onto a quadruped robot are (1) low total
weight of arms including controllers, (2) compactness, (3) torque controllability, and (4) high joint speed and
torque. Commercially available solutions are either too bulky because of their heavy base and controller
units (Barrett’s WAM arm, KUKA’s lightweight arm), not torque controlled (Universal Robots UR5) and/or too
slow (HDT Robotics’ MK1).
Due to this lack of commercial solutions, we have developed a compact arm with 6 hydraulic, torque
controllable joints (Figure 7, left). The arm including all electronics and valves weighs around 13kg. The 6
degrees of freedom (DOF) are constructed with a combination of light-weight cylinders and rotary vane
actuators. Table 2 lists the actuator type and properties of the arm’s 6 joints, according to the definition
shown in Figure 7 on the right.
Figure 7. CAD rendering and kinematics of the new arms. Left: CAD rendering of the new pair of hydraulic 6-
DOF arms. Right: kinematics of the arm with the names of the joints: shoulder abduction/adduction (SAA),
shoulder flexion/extension (SFE), humerus rotation (HR), elbow flexion/extension (EFE), wrist rotation (WR)
and wrist flexion/extension (WFE).
Table 2. List of the actuator type and properties of the arm’s 6 hydraulic joints
Joint Name Actuator type Joint range max. Joint torque/force
SAA single vane rotary 210° 126Nm
SFE double vane rotary 90° 120Nm
HR double vane rotary 98° 120Nm
EFE cylinder 130° 4kN
WR single vane rotary 210° 60Nm
WFE cylinder 120° 4kN
Each joint’s position is measured with high resolution absolute encoders (19Bit). While the rotary actuators’
torque output is measured with strain-gauge torque sensors, the cylinder force is obtained with load cells in
series to the piston rod. All actuators are controlled by MOOG E024 servo valves. Distributed electronics on
the arm read the sensors and create the output signal for the valve amplifiers. An EtherCAT bus connects
the arm to the robot. For more detailed information on the arm design and components, see .
4.2. HyQ2Centaur Design Concept
HyQ2Centaur is a combination of HyQ2Max (see Section 3.1) and a pair of the new hydraulic arms (Section
4.1) as illustrated in Figure 8. The modular design of the arms allows easy mounting and removing from the
robot’s torso. The hydraulic interface consists of two quick release couplings for the arm’s pressure and
return lines. The communication interface is a single EtherCAT cable that also provides the electric power to
the arm’s electronic boards and valve amplifiers.
Figure 8. CAD renderings of HyQ2Centaur that consists of a HyQ2Max four-legged base and a pair of the
new arms. Left: centaur with extended arms; Right: centaur with stowed-away arms.
The additional weight of the two arms is around 26kg. This payload is not located in an optimal position with
respect to the locomotion stability. The most conservative stability criterion is static stability, where the
projection of the robot’s center of mass onto the ground needs to stay inside the support polygon (created by
the feet in contact with the ground). Other stability criteria consider simplified dynamics of the robot to create
stable locomotion (e.g. Zero Moment Point).
Since all joints of the robot can be controlled in torque , the robot‘s whole body dynamics model can be
used to obtain joint torque profiles that optimise the force distribution of the four feet (and other contact
points, e.g. with the arms). We have recently presented our first results with optimized joint torques that
allowed the robot to climb inside a V-shaped groove . The same approach allows the centaur robot in the
future to optimise joint torques during manipulation tasks.
4.3. Possible Application Scenarios
Manipulation capability allows a legged robot to perform various tasks in real-world applications. Figure 9
illustrates a few of these tasks performed by HyQ2Centaur. During an inspection or rescue operation it might
be necessary to open doors or to navigate over challenging terrain to open/close a valve. Other important
tasks will be the remote handling of hazardous objects for example for nuclear decommissioning.
Figure 9. HyQ2Centaur task scenarios. From left to right: opening doors, full-body motion to turn valve and
remote handling of hazardous materials.
5. HYDRAULIC ACTUATION SYSTEM
This section discusses the hydraulic system of HyQ2Max and HyQ2Centaur. In the current configuration
hydraulic power is supplied to the robot through two highly flexible hoses. An onboard power pack is
currently under development.
Figure 10 shows the schematic of HyQ2Centaur’s hydraulic actuation system. For simplicity only the details
of one leg are shown. The torso of the robot carries the following hydraulic system components: An
accumulator to smooth out pressure ripples and provide extra flow during fast variations in hydraulic flow
demand; a pressure relief valve to protect the system; a normally-open, solenoid-operated vent valve
connecting the pressure supply to tank in case of an emergency and two pressure transducers. The robot’s
leg and arm joints are moved by cylinders and rotary vane actuators. Each joint is controlled by a servovalve.
Figure 10. Schematics of the hydraulic actuation circuit of the HyQ2Centaur robot. For simplicity only the
details of one leg are shown. The two arms and other legs are built up with the same actuators and valves.
6. DISCUSSION AND CONCLUSION
Open challenges in the field of hydraulic legged robots are primarily the energy efficiency of the hydraulic
actuation system, the hose routing and the large size of commercially available components. Energy
efficiency is especially poor in torque controlled hydraulic robots because of the internal leakage of the high-
bandwidth servovalves needed for proper torque control . Digital hydraulics (see  for a recent review)
and variable pressure systems are some of the possible solutions that are currently being investigated by the
research community. A neat routing of hoses across moving joints is tricky especially if the joint range of
motion is large. Slip rings, custom connectors and highly flexible hoses are possible solutions. Another
challenge for a hydraulic robot designer is the generally large size of commercially available components.
Since the market for small scale components is (still) small, custom-made parts or expensive niche products
are often the only solution.
This paper presented the design concepts of the hydraulic quadruped robot HyQ2Max and the centaur-style
robot HyQ2Centaur. HyQ2Max is an evolution of IIT’s hydraulic quadruped HyQ, a robot that since 2011
demonstrated various types of agile locomotion and carefully planned navigation over rough terrain. The
second version has improved robustness, larger joint ranges and higher joint torques. We presented an
overview of the robot’s design and demonstrated the robot’s self-righting ability with a rigid body dynamics
simulation. Next, we showed the design of a pair of new light-weight hydraulic manipulator arms that can be
mounted onto the HyQ2Max platform to turn the robot into the centaur-style machine HyQ2Centaur.
Furthermore, we presented the robots’ hydraulic actuation systems and discussed open research problems
in the field of hydraulic legged robots. To the authors’ best knowledge this is the first time the design of a
fully hydraulically actuated centaur robot is presented.
This research has been funded by the Fondazione Istituto Italiano di Tecnologia. The authors would like to
thank also the colleagues that collaborated for the success of this project: Hamza Khan, Ioannis Havoutis,
Stephane Bazeille, Jesus Ortiz, Marco Camurri, Carlos Mastalli and our team of technicians. Furthermore,
we would like to thank Jonas Buchli and Thiago Boaventura of ETH Zurich and Satoshi Kitano of the Tokyo
Institute of Technology for their input and help.
 M. Raibert, K. Blankespoor, G. Nelson, R. Playter, and the BigDog Team, “Bigdog, the rough-terrain
quadruped robot,” in Proceedings of the 17th World Congress of the International Federation of
Automatic Control (IFAC), 2008.
 C. Semini, N. G. Tsagarakis, E. Guglielmino, M. Focchi, F. Cannella, and D. G. Caldwell, “Design of
HyQ - a hydraulically and electrically actuated quadruped robot,” Journal of Systems and Control
Engineering, Vol. 225(6), pp. 831-849, 2011.
 V. Barasuol, J. Buchli, C. Semini, M. Frigerio, E. R. De Pieri, and D. G. Caldwell, “A reactive controller
framework for quadrupedal locomotion on challenging terrain,” in Proceedings of the IEEE
International Conference on Robotics and Automation (ICRA), 2013.
 C. Semini, V. Barasuol, T. Boaventura, M. Frigerio, M. Focchi, D. G. Caldwell, J. Buchli. "Towards
versatile legged robots through active impedance control," International Journal of Robotics Research,
 M. Focchi, V. Barasuol, I. Havoutis, J. Buchli, C. Semini, and D. G. Caldwell, “Local reflex generation
for obstacle negotiation in quadrupedal locomotion,” in Proceedings of the International Conference on
Climbing and Walking Robots (CLAWAR), 2013.
 I. Havoutis, J. Ortiz, S. Bazeille, V. Barasuol, C. Semini, and D. G. Caldwell, “Onboard perception-
based trotting and crawling with the hydraulic quadruped robot (HyQ),” in Proceedings of the
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2013.
 S. Bazeille, V. Barasuol, M. Focchi, M. Frigerio, I. Havoutis, D. G. Caldwell, and C. Semini,
“Quadruped robot trotting over irregular terrain assisted by stereo-vision,” Journal of Intelligent Service
Robotics, Vol. 7(2), pp. 67-77, 2014.
 A. Winkler, I. Havoutis, S. Bazeille, J. Ortiz, M. F. Focchi, R. Dillmann, D. G. Caldwell, and C. Semini,
“Path planning with force-based foothold adaptation and virtual model control for torque controlled
quadruped robots,” in Proceedings of the IEEE International Conference on Robotics and Automation
 Online video of the locomotion skills and performance of IIT’s HyQ robot, March 2014,
 R. Liston, and R. Mosher, "Development of a quadruped walking machine," Mechanical Engineering,
Vol. 89(10), 1967.
 M. Raibert, “Legged Robots That Balance,” The MIT Press, 1986.
 M. Raibert, “Trotting, pacing and bounding by a quadruped robot,” Journal of Biomechanics, Vol. 23,
pp. 79-98, 1990.
 R. Hodoshima, T. Doi, Y. Fukuda, S. Hirose, T. Okamoto, and J. Mori, “Development of quadruped
walking robot titan XI for steep slopes - slope map generation and map information application,”
Journal of Robotics and Mechatronics, Vol. 19, pp. 13-26, 2007.
 C. Semini, “HyQ – design and development of a hydraulically actuated quadruped robot,” Ph.D.
dissertation, Istituto Italiano di Tecnologia (IIT) and University of Genova, 2010.
 T.-J. Kim, B. So, O. Kwon, and S. Park, “The energy minimization algorithm using foot rotation for
hydraulic actuated quadruped walking robot with redundancy,” ISR/Robotik, 2010.
 X. Rong, Y. Li, J. Ruan, B. Li, “Design and simulation for a hydraulic actuated quadruped robot,”
Journal of Mechanical Science and Technology, Vol. 26(4), pp. 1171-1177, 2012.
 X. Chen, F. Gao, C. Qi, and X. Zhao, “Spring parameters design to increase the loading capability of a
hydraulic quadruped robot,” in Proceedings of the International Conference on Advanced Mechatronic
Systems (ICAMechS), 2013.
 M. Kim, S. Kang, S. Lee, W. Chung, K. Cho, and C.-W. Lee, “Development of a humanoid robot
CENTAUR-design, human interface, planning and control of its upper-body,” in Proceedings of the
International Conference on Systems, Man, and Cybernetics, 1999.
 S. Tsuda, K. Shinozaki, and R. Nakatsu, “Development and Evaluation of a Centaur Robot,” Service
Robotics and Mechatronics, Springer, 2010.
 A. Halme, K. Koskinen, V.-P. Aarnio, S. Salmi, I. Leppänen, and S. Ylönen, “WorkPartner - Future
Interactive servicerobot,” In Proceedings of the 9th Finnish Artificial Intelligence Conference, 2000.
 Online video of Dynamic Robot Manipulation by Boston Dynamics’ BigDog robot, February 2013,
 S. Schaal, “The SL simulation and real-time control software package,” CLMC lab, University of
Southern California, Tech. Rep., 2009.
 M. Frigerio, J. Buchli and D. G. Caldwell, “Code Generation of Algebraic Quantities for Robot
Controllers”, in Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and
Systems (IROS), 2012.
 B. U. Rehman, M. Focchi, M. Frigerio, J. Goldsmith, D. G. Caldwell and C. Semini, “Design of a
Hydraulically Actuated Arm for a Quadruped Robot,” in Proceedings of the International Conference
on Climbing and Walking Robots (CLAWAR), 2015 (under review).
 T. Boaventura, C. Semini, J. Buchli, M. Frigerio, M. Focchi, and D. G. Caldwell, “Dynamic torque
control of a hydraulic quadruped robot,” in Proceedings of the IEEE International Conference on
Robotics and Automation (ICRA), 2012.
 M. Focchi, A. Del Prete, I. Havoutis, C. Semini, R. Featherstone and D. G. Caldwell, “High-slope
Terrain Locomotion for Torque-Controlled Quadruped Robots,” Workshop on Whole-Body Control for
Robots in the Real World, IEEE/RSJ International Conference on Intelligent Robots and Systems
 M. Linjama, “Digital Fluid Power – State of the Art,” in Proceedings of the Twelfth Scandinavian
International Conference on Fluid Power, 2011.