Mission Statement

The mission of Yobotics is to advance the state-of-the art in legged robots and powered leg orthotics while filling the needs of various legged robot markets.  Our goal is to grow alongside the fields of biomimetic robotics and powered prosthetics as we lead the development of new technologies with healthcare, academic, military, and entertainment applications. 

1    Who We Are

Yobotics, Inc was founded in January 2000 by four graduates of the M.I.T. Leg Laboratory, a subsidiary of the M.I.T.’s world renowned Artificial Intelligence Lab.  We draw considerable experience and technical expertise from nearly 20 years of successfully designing, building, and controlling biologically inspired running and walking robots.   Our robots have been featured in television programs on the Discovery Channel, TLC, and PBS and have been photographed for National Geographic, Robo Sapiens, Playboy (Japanese Version) and the front cover of Wired magazine (August 2000).

2       What We Do

Yobotics is a cutting-edge robotic design, consulting, and research firm specializing in biomimetic robots, powered leg orthotics, and force-controllable actuators.   We believe that in the future, robots will play a role in our everyday lives.  Our vision includes robots that not only perform everyday tasks for people, but also provide a new form of entertainment.  For this to be possible, robots need to transcend the stereotype of clunky, arthritic devices.  We intend to work towards eliminating this stereotype by building dexterous, agile, dynamic machines capable of operating in human environments.

2.1      Health Care Applications

Currently, we are working on a powered exoskeletal device which will provide the disabled (particularly those with lower extremity weakness) a higher degree of mobility.  The device will determine the users intent (to speed up, slow down, turn left, turn right, climb stairs etc.) and amplify his/her motions to achieve greater speed and endurance in everyday activities.  The ultimate goal of this project is to allow the severely disabled, such as paraplegics and quadriplegics, to walk and balance with the assistance of the exoskeleton.  

2.2 Entertainment Applications  

There is great potential for walking and running robots in various entertainment venues.  Potential applications include robotic events such as robot competitions, advertising, trade shows, robot toys, legged  robots in museums and theme parks, etc.   We are actively pursuing teaming arrangements in these fields.  Our expertise is the design, construction, and control of walking and running robots.  We are looking for partners whose expertise is marketing, event planning, advertising, and promotion.

2.3  Military Applications

While at the M.I.T. Leg Laboratory, all four of Yobotics’ co-founders were involved in basic research for the Department of Defense.  We are continuing to pursue funding from the DoD for various applications of legged robotic devices which would assist soldiers and keep them from harms way.

2.4  Academic Applications

There is currently a growing community of legged robotics laboratories throughout the world.  However, no common bipedal robotic platform exists.  Due to the complexity of designing and building a bipedal robot, the laboratories typically choose to stick purely to simulations, or venture forth on a long and expensive journey to design and build their own custom robot.  Unfortunately, many of these robots never work.  Since there is no common robot platform or simulation environment, it is difficult for labs to reproduce or validate each other’s work, or to collaborate on a common project.  Reproduction, validation, and collaboration are the hallmarks of science and the reason for its rapid advancement.  Walking robot research has made very slow progress since its inception in the late 1960’s, largely due to the lack of such measures.  

One of Yobotics’ initial thrusts was to help synergize bipedal walking robot research by providing a common platform and establishing a protocol to encourage collaboration and validation between groups already doing such research.  We intended to produce and distribute a common research platform, consisting of both a simulation software package and a mechanical robot, to achieve these goals.  This has progressed much slower than we had hoped, but we are currently communicating with various laboratories which are interested in using either our Spring Flamingo or M2 bipedal walking robots in their research.  We have also developed a simulation package for legged robots and currently have an evaluation version available to the robotic community. 

3       Unique Approach

3.1       A Comparative Study

During the past 30 years, many different schemes to control walking and running robots have been implemented, with vary degrees of success. Among the most common are trajectory tracking methods, which rely on deriving and solving the dynamic equations of motion and using the solution to determine a desired trajectory.  High gain position controllers, often employing adaptive elements, are used to follow the predefined trajectory.

Trajectory tracking methods have two primary drawbacks.   First, they are computationally intensive.  The dynamic equations for a robot with many degrees-of-freedom can be very difficult to solve, even numerically.  Second, trajectory controlled robots are not robust to disturbances or changes in the environment. For example, if the robot collides with an object in its environment or its load changes, it must recalculate its dynamics and trajectory or it will fall over. 

Our approach to controlling legged robots is unique in two ways.  First, the control algorithms we employ do not require pre-programmed joint trajectories, but instead rely on active balance.   Reaction forces required to maintain balance are calculated on the fly using Virtual Model Control, a novel control technique originated by Prof. Gill Pratt and developed by Yobotics co-founder Jerry Pratt while a graduate student at the MIT Leg Laboratory.  Virtual Model Control uses virtual springs and dampers placed at strategic locations on the robot to control the pitch, height and speed of the robot.  The virtual forces applied by the springs are mapped to physical torques at each of the robots joints.  The resulting reaction forces on the body exactly mimic the virtual forces created by the virtual elements. Active balance is effective because it can easily be adapted to many different environments and postures.

The second key feature of our control approach to robot locomotion is the exploitation of natural dynamics.  Legged animals are “designed” in such a way that the natural interaction of their limbs and gravity accomplish much of what’s needed in walking.  Muscle control is superimposed on top of the passive dynamic behaviors of the limbs.  For instance, during the swing phase of the human walking gait, the leg muscles experience a power spike to begin leg swing and remain limp throughout the rest of the swing phase. Thus, the leg swings naturally like a pendulum during this phase of gait.  Likewise, we exploit this feature in our robots by allowing the joints of the leg go limp during swing. 

Our focus on active balance and natural dynamics is superior to the playback method of control. It results in a gait that is smooth and natural looking, while robust to disturbances and deviations in terrain.

3.2      Enabling Technology

A requirement for implementing active balance and exploiting passive dynamics is the use of force controllable actuators at each joint of the robot.   To achieve force control in our robots, we use Series Elastic Actuators, a patented actuation technology invented by Gill Pratt and Matt Williamson (U.S. patent number 5650704) at the MIT Artificial Intelligence Lab.  As graduate students in the MIT Leg Laboratory, the Yobotics founders helped to further develop the actuators.  We are now licensing this technology from MIT and have produced a few commercially available versions that are being used in our robots and in several robotics labs.  

With Series Elastic Actuators, a spring is intentionally placed in series with the output of the actuator.  By measuring the spring deflection, load force can be determined and controlled.  The result is high fidelity, low impedance, low friction force output.  

When implementing active balance, force control is imperative because it allows the control engineer to impart restoring forces to the body.    For example, if a 12 degree-of-freedom biped with force-controlled actuators begins to tip to its left, the control system can simply apply a restoring force to the right.  Using Virtual Model Control, this restoring force can easily be implemented by coordinating the forces applied at each of the 12 joints. 

In contrast, if the robot’s joints are position-controllable, and not force-controllable, it is difficult to generate a restoring force.  The best one can do in this case is to reconfigure the robot’s posture by commanding a new position at each joint and hope that it prevents tipping.  This requires precise timing and coordination to successfully implement, not to mention a much greater degree of understanding of the robot’s behavior.

Force control is also necessary when exploiting the passive dynamics of the robot.  For passive dynamic behaviors to be implemented, the dynamics of the motor driving the given joint must be decoupled from the dynamics of the limb attached to that joint.  If the two remain coupled then the dynamics of the limb will be dominated by the dynamics of the motor (which have huge reflected inertias due to large gear reductions).   

Typical robots, including the Honda P3 and the Sony SDR3, utilize position controllable actuators which are good at tracking trajectories, but not necessarily good at providing smooth corrective forces on the robot or taking advantage of the robot’s passive dynamics.  The pervasive use of position control actuators in robots seems to be the underlying reason for the pervasive use of joint trajectory tracking control methods – there simply wasn’t any alternative.  With force-controllable Series Elastic Actuators, Yobotics provides a fresh approach to bipedal walking robots.

4       Yobotics Founders

Yobotics was founded by four MIT graduates, all who obtained advanced degrees while working at the MIT Leg Laboratory. We have significant expertise in the area of legged robots, actuation technologies, mechanical engineering, electrical engineering, and computer science.

Dr. Jerry Pratt:  Dr. Pratt received his Bachelors Degrees in Mechanical and Electrical Engineering, a Masters Degree in Computer Science, and a Ph.D. in Computer Science, all from the Massachusetts Institute of Technology.  As a graduate student, Jerry worked in the MIT Leg Laboratory, where he made significant contributions to the field of walking robotics.  For his Masters Thesis, Jerry developed a control tool, called Virtual Model Control, which allowed for simpler, more robust algorithms for all types of walking robots.  Virtual Model Control was successfully used in the control of two bipedal walking robots, Spring Turkey and Spring Flamingo, which Jerry designed and developed.  For his PhD thesis, Jerry developed control algorithms for Spring Flamingo which enabled graceful, human-like walking at speeds of 1.2 m/s (tied with the fastest bipedal robot in the world). Jerry also wrote control algorithms for M2, a three dimensional bipedal robot, and assisted in its design and construction. 

Mr. Ben Krupp:  Mr. Krupp earned a Bachelors Degree in Mechanical Engineering from Ohio State University and has a Masters Degree from MIT in Mechanical Engineering. At the MIT Leg Lab, Ben designed, simulated, and controlled Corndog, a planar robot for studying quadrupedal locomotion. At Yobotics, he has developed a 7 degree-of-freedom force-controllable agile robot arm.   He spent his last year at OSU working on the design and control of a five degree-of-freedom robotic arm for NSF. 

Mr. Daniel  Paluska:  Mr. Paluska earned his Bachelors and Masters Degrees in Mechanical Engineering from MIT. For his Masters Thesis Mr. Paluska designed and built a 12 degree of freedom bipedal robot, called M2, at the MIT Leg Laboratory.  For his undergraduate thesis, Mr. Paluska designed and built a hopping lamp modeled after the famed Pixar Luxo Lamp. 

Mr. Chris Morse:  Mr. Morse earned his Bachelors Degree in Mechanical Engineering from Northeastern University and received his Masters Degree in Mechanical Engineering from MIT.  For his Masters Thesis Mr. Morse designed Coco, a fifteen degree of freedom quadrupedal walking robot which mimics a small gorilla.  Additionally, Mr. Morse has several years of industrial design experience. 

5       Our Robots

Yobotics Co-founders have developed and controlled various robots and robotic actuators including: 

• M2:  A 3D, 12 degree-of-freedom, bipedal walking robot that can currently walk in place and balance on one leg.  Walking algorithms for M2 are currently being developed.
• Spring Flamingo:  A planar, 6 degree of freedom, bipedal walking robot that walks at 1.2 m/s with a natural dynamic gait.
• Spring Turkey:  A planar, 4 degree-of-freedom, bipedal walking robot.
• Corndog:  A 4 degree-of-freedom planar running robot, consisting of one side of a quadruped. 
• Coco:  A  fifteen degree-of-freedom quadrupedal walking robot which mimics a small gorilla. 
• Agile Robot Arm:  A seven degree-of-freedom force-controllable robot arm developed for human-robot interaction research.  One of the robots is currently in use by the MIT Media Lab.
• Series Elastic Actuators:  Electric, or hydraulic force-controllable actuators with high force-fidelity, low friction, and low reflected inertia.

©2000-2011 Yobotics, Inc. All Rights Reserved