Aug-24-19
The Astro robotic dog can be taught new tricks with the help of a deep neural network that allows it to learn through trial and error.
Developed by a team from Florida Atlantic University, the Astro robo-dog steps out of the pack with its Nvidia Jetson TX2 graphics processing units, offering four teraflops of processing power. This power, along with Astro’s onboard cameras, sensors, radar imaging and directional microphone, allows the robot to learn from experience—much like training a real dog. According to the team, Astro could eventually carry out a range of functions, including bomb detection, search and rescue, and even making real-time decisions based on past experiences.
Show 1 Comment