2018; Humanoid Raman (4i Labs)

Introduction

TL; DR: https://github.com/otoshuki/Raman-TheVisionBot

We, the Humanoid team at 4i Labs, IITG are developing Raman, a humanoid robot with the aim to explore and research the field of robot social interaction and cooperation.

Current progress

Since development of a complete humanoid robot is a complex task, the entire team has been divided into sub teams working on different problem statements. The entire humanoid team has been divided into design, AI, electronics and biped teams.

The humanoid has been divided into 4 different parts – Head, Torso, Arms and Biped.

While all the parts are still in development, we have a working version of the head on which we test our AI and electronics team’s current work.

Current version of the humanoid can perform visual interaction with the user, maintaining eye contact while performing verbal communication.

Work

I worked on the speech subsystem and integration of the humanoid, including the integration of Speech-to-Text system with response generation model (developed separately by the AI team) and the Text-to-Speech system, completing the current version required for verbal communication.

The humanoid responds to the trigger word “Raman” which was introduced in the system later using Porcupine library. The system currently uses Google Cloud Speech API for STT and espeak for TTS. I also worked on the Respeaker Core V2 platform for speaker localization using TDOA technique.

I was the Control and Localization Team Head and worked on human pose replication for the forelimbs and development of ROS packages for integration.