After much deliberation and with a heavy heart, the decision has been made to officially stop work on ROTH. The two main goals for ROTH were that it be able to walk and that others be able to build it for themselves. The first goal, walking, was mostly accomplished, in that ROTH walked but never with a great deal of stability or reproducibility. This was because the servos, especially at the hip and ankles, were always running near their stall torque which caused them to heat up. Once hot the servos would start to behave differently and eventually fail (went through several 5 amp fuses and 10 servos). The servos were overloaded due to ROTH's considerable weight, at around 8lbs, and its 16.5" height.
Several avenues were explored to assist the struggling servos, such as using two servos per joint and integrating an extra gear into the joint frame for higher torque at the expense of angular travel and speed. While these efforts helped they also added complexity, cost, and weight. Already a beast at over 40 different parts any additional complexity made fabrication and documentation a daunting prospect. A host of alternative servos were also evaluated ranging from the $30 Raiden DMP-105 49G to the $100 HS-7954SH. The HS-7954SH had 3 times the torque of the T-Pro MG996Rs that were in the initial design and would have allowed the robot to be completed. However at about x10 the cost adding an extra $700 to the already expensive parts list just wasn't compatible with the secondary goal of allowing others to build ROTH for themselves.
Many important lessons were learned from ROTH about size, weight, joint construction, design complexity, battery usage and static walking. Not to despair, this knowledge will be leveraged on the next prototype FOBO (FOrth BOt), which has been under tandem development for a couple of months now.