Biorobotics and Locomotion Lab Home     |     Andy Ruina
cornell_logo      Cornell Tik-Tok
Efficient, robust, and nimble
open-source legged robot

In progress

Updated: April 19 2017

     nsf_logo

Overview
The Tik-Tok project aims to build an open-source biped robot platform with the dynamic balance and robustness of Boston Dynamics's New ATLAS robot, but at a tenth the cost and energy use. Think of it as a "Segway with legs."
Its Cost of Transport (CoT) in simulation is 0.25 (≈ human). Tik-Tok uses high-efficiency chain drives and powerful brushless motors to create a highly nimble robot platform. If pushed or tripped, it can reposition its foot by up to a full step in under a quarter of a second (also ≈ human). Tik-Tok is 1.5 m tall (with head, as shown), weighs 30 kg, has 12 actuated joints, has a peak joint power of 2 kW, and is expected to be able to walk about 15 km on a single charge of its 300 W-hr 2 kg battery.

Tik-Tok CAD rendering
Tik-Tok CAD rendering

   Design goals

  • Suitable for reliable locomotion in environments designed for humans.
  • Low energy with CoT ≈ 0.25 (better than all other robot bipeds).
  • Robust balance, based on high-speed, high-accuracy foot placement for balance correction. Should match the robustness of other successful walking robots (Petman, New ATLAS, Cassie).
  • Leg swing time for foot placement, 1 radian in < 300 ms (≈ human).
  • Squat, sit down, stand up, climb steps and curbs.
  • Jog, dance, skip, hop, etc. (optional, but the physical capability will likely follow from the other requirements).

   General Details

  • 1.5 m tall (full robot, as at left)
  • 30 kg mass.
  • 0.8 m leg length (below, left).
  • 12 actuated joints: 4 arm, 4 hip , 2 knee, 2 ankle.

   Leg actuators

  • RoboDrive ILM70x18HS brushless stators and rotors with custom cases.
  • Three-stage chain drives, with walking-optimized reduction ratios of 51:1 to 62:1.
  • Overload slip clutches to protect agains collision damage.
  • 200 N m joint torque, peak.
  • 10 rad/sec joint rotational speed, peak.
  • 2 kW peak joint power.

   Power, sensing and computing

  • 60 volt 80 amp mini motor drives designed for minimal idle power.
  • 14-bit Hall Effect angle sensors on all motors and joints.
  • High-accuracy MEMS rate gyros and accelerometers at each joint.
  • Joint torque sensors.
  • Ground contact pressure sensors in feet.
  • 59 volt 5 amp-hour (295 W-hr) lithium polymer batteries.
  • Atmel 300 MHz SAM E70 microcontroller motor control and sensor boards networked via CAN bus.
  • BeagleBone Black 1 GHz top-level control board and Robot Operating System (ROS) interface. 


Pictures and Video

Demonstration of Leg Range of Motion
    
Overload clutch sprockets
Motor controller board
Hip joint
Calf with chain drive
Hip joint and thigh
Motor stator w/ wiring board + terminals

Photos

Papers and Posters


   People:

Principle Investigator: Andy Ruina
Engineer: Jason Cortell
Current Graduate Student: Matt Sheen
Previous Graduate Students: Matt Kelly, Petr Zaytsev, Anoop Grewal
Previous Post-doctoral fellow:        Javad Hasaneini
Undergraduates Peter Bagaley, Olav Imsdahl