Upper Body Design of a Humanoid Robot for the DARPA Robotics Challenge
Humanoid robots have captured the imagination of authors and researchers for years. Development of the bipedal walking necessary for humanoid robots began in earnest in the late 60's with research in Europe and Japan. The the unique challenges of a bipedal locomotion led to initial robots keeping power, computation, and perception systems off-board while developing the actuators and algorithms to enable locomotion. As technology has improved humanoid and exoskeleton systems have finally incorporated all the various subsytems to build a full independent system. Many of the groups building these platforms have developed them based on knowledge acquired through decades of prior development. For groups developing new humanoid systems little guidance on the pitfalls and challenges of humanoid design exist.
Virginia Tech's robot ESCHER, developed for the DARPA Robotics Challenge (DRC), is the 4th generation full sized humanoid developed at the University. This paper attempts to quantify the design trades and techniques used to predict performance of ESCHER and how these trades specifically affected the design of the upper body. The development of ESCHER became necessary when it became obvious that the original design assumptions behind the previous robot THOR left it incapable of completing the DRC course and the necessary upgrades would require an almost complete redesign. Using the methods described in this paper ESCHER was designed manufactured and began initial testing within 10 months. One and a half months later ESCHER became the first humanoid to walk the 60 m course at the DRC.
The methods described in this paper provide guidance on the decision making process behind the various subsystems on ESCHER. In addition the methodology of developing a dynamic simulation to predict performance before development of the platform helped provide design requirements that ensured the performance of the system. By setting design requirements ESCHER met or exceeded the goals of the team and remains a valuable development platform that can provide utility well beyond the DRC.