Scientists at Florida Atlantic University (FAU) have used 3D printing, deep learning, artificial intelligence and robotics to create Astro, an intelligent robot dog that can see, hear, train and learn.
A product of FAU’s Machine Perception and Cognitive Robotics Laboratory (MPCR) in the Charles E. Schmidt College of Science, Astro uses the quadruped robotic system developed by American engineering and robotics design company Boston Dynamics. Building on Boston Dynamics’ work, the FAU team attached a 3D printed head, resembling a Doberman pinscher, to a quadruped robot. The head features a computer simulated brain that doesn’t operate according to pre-programmed robotic automation. Instead it utilizes a deep neural network that allows the robodog to be trained using vocal inputs and visual signals, therefore learning and behaving from experience, with the potential of performing useful tasks.
“Our Machine Perception and Cognitive Robotics laboratory team was sought out by Drone Data’s Astro Robotics group because of their extensive expertise in cognitive neuroscience, which includes behavioral, neurophysiological and embedded computational approaches to studying the brain,” commented Ata Sarajedini, Ph.D., dean of FAU’s Charles E. Schmidt College of Science.
Described as a “puppy-in-training”, currently Astro only responds to commands such as “sit,” “stand” and “lie down.” Over time, the scientists claim that Astro will learn to understand and respond to hand signals, detect different colors, comprehend many languages, coordinate his efforts with drones, distinguish human faces, and even recognize other dogs.
Astro is able to interact with its environment in real time through the integration of sensory inputs like high-tech radar imaging, cameras and a directional microphone. The devices allow Astro to detect environmental data across multiple modalities including optical, sound, gas and radar.
It is equipped with a set of Nvidia Jetson TX2 graphics processing units with a combined four teraflops of computing power. Reportedly amounting to four trillion computations a second, this provides Astro with the ability to process its sensory inputs and make autonomous behavioral decisions based on what it perceives.
Astro the robodog. Photo via FAU.
With its sensory inputs, processing and computing power, the MPCR team intend for Astro to act as an information scout. It is said that eventually the robodog will be able to traverse through rough terrains, and respond to hazardous situations, with the intention of keeping humans and animals out of danger. According to the scientists, it will potentially be able to sort through a facial recognition database, smell the air to detect foreign substances, and hear and respond to distress calls that fall outside a human’s audible hearing range.
Key missions for the robodog could include the detection of guns and explosives to assist police, the military and security personnel. Beyond armed services, Astro can also be programmed to operate as a service dog for the visually impaired, or provide medical diagnostic monitoring. The scientists are also training the robot dog to serve as a first responder for search and rescue missions like hurricane reconnaissance and military maneuvers. They are also programming Astro to contain a database of logged experiences that can be drawn upon by the robot to help its decision making process.
3D printing the path towards the singularity
3D printing has contributed towards the development of multiple artificially intelligent robots in recent years. Due to its design and manufacturing capabilities, the technology is also being researched in the field of soft robotics.
Boston Dynamics itself has used 3D printing to develop its humanoid robot, Atlas. Created for the purpose of search and rescue, the robot is able to walk, jog on uneven ground, jump over obstacles, and pick itself up from a fall. 3D printing played a significant part in the development of Atlas’ legs, integral for its agility and maneuverability.
ETH Zurich robotic engineers have also created a self-learning ice-skating robot using 3D printing named Skaterbot. The robot was exhibited at this year’s World Economic Forum in Davos, where it also took part in an ice hockey match.