Development of an Interoperable and Robust Bipedal Robot Vision Platform with Image Stabilization

William Rittase, Mathew Kandler, Scott Bevan, Danielle Renzi, Steven Shooter, and Keith Buffinton

Keywords

Robot Vision, Image Stabilization, Robot Head

Abstract

The M2V2 humanoid robot had been used by the Institute of Human and Machine Cognition (IHMC) and Bucknell University over the past several years to develop algorithms for more efficient walking and push recovery. However, the biped has only consisted of a torso and legs. To facilitate operations in an urban environment, an interoperable vision platform (or head) was developed for M2V2 to provide an operator with a comfortable, user friendly, accurate, and dynamic means of perceiving of the biped’s environment. In addition, the head required hardware capabilities for future development of object tracking and avoidance. A list of specifications was created to quantify subjective requirements. The result was a two degree of freedom (DOF) system which is actuated by DC motors and uses harmonic drives and a belt-pulley system to transmit torque with minimal backlash. The head uses two individual cameras to create 3D vision for a set of goggles worn by the operator. The goggle system also has a gyroscope to track movements of the operator’s head and move the vision platform to provide the operator with the perception that the robot’s environment is his own, providing easier use. An active stabilization system was used to stabilize the image and provide more comfort to the operator. Image movement was reduced on average in the video by 55% with this system. To meet additional future requirements of such a platform, a binocular camera was provided for object tracking and avoidance.

Important Links:



Go Back