Robots Atlas>ROBOTS ATLAS

Sophia

Shipping

Positioning and history

Sophia is a social humanoid android developed by Hanson Robotics (Hong Kong) and first publicly demonstrated in March 2016. David Hanson led its design, modelling Sophia's face on Audrey Hepburn. In 2017 Sophia became the world's first robot to receive honorary citizenship — granted by the Saudi Arabian government — and was appointed UNDP Innovation Champion for Sustainable Development.

Sophia has become a global symbol of social robotics, appearing on television programmes (60 Minutes, Tonight Show), at major conferences (CES, SXSW, World Economic Forum in Davos) and before the UN General Assembly. The robot continues to be actively developed, with Hanson Robotics regularly releasing updated versions with improved software and more capable limbs.

Construction and materials

The robot stands 167 cm tall and weighs approximately 20 kg. The head and face are made of Frubber (Hanson Robotics patented elastomer) driven by over 30 actuators (Dynamixel XM430 and Xpert servos) — totalling 36 degrees of freedom in the head and neck alone. Each servomechanism corresponds to a set of Frubber muscles enabling more than 50 facial expressions. The full body has 83 DOF: 36 in head/neck, 15 per arm/hand, 3 in torso, and 14 in the mobile base.

The skeleton uses carbon fibre, CNC aluminium, steel, Spectra fibre, Delrin thermoplastic, acrylic, polycarbonate, and 3D-printed parts. The robot can be powered from a 110/220 V mains supply or a 24 V lithium-polymer battery.

Perception and compute

Sophia has three camera systems: a wide-angle 1080p camera in the chest, an Intel RealSense depth camera, and a pair of 720p HD cameras in the eyes for face tracking. Touch sensors in the fingers, joint angle sensors, and force sensors in the arm joints complement perception. Sound is localised via an audio localisation microphone array. The compute unit is an Intel Core i7 at 3 GHz with 32 GB RAM and integrated GPU.

Software and AI

The operating system is Ubuntu Linux. Network connectivity is via Ethernet and Wi-Fi. The AI software is based on the open-source OpenCog AGI framework and dedicated Hanson Robotics modules: face recognition, visual tracking, natural language processing (NLP), and response generation. The robot has also supported integrations with external NLP engines including Google Speech and IBM Watson.