There were several Jetson TK1 based robotic vehicles at the GPU Technology Conference (GTC) 2015 in San Jose, California. Looky here:
Officine Robotiche – Jetson Champs
In the NVIDIA booth, there were two members of Officine Robotiche: Raffaello Bonghi and Walter “Myzhar” Lucetti both from Italy. Their robots are based on the Jetson TK1, and run an open source Robot Operating System (ROS) stack which includes depth and RGB vision from an ASUS Xtion camera. Here are a couple of closeup shots (click to enlarge)
Though dissimilar in appearance, both robots run the same software stack! Read their blogs for more information.
Paracosm was showing off a traditional radio controlled car that they transformed into a smart-bot using their Spatial Intelligence Platform. The platform takes advantage of the Jetson TK1 computing power to capture real world environments and turn them into 3D models.
The robot was placed into a lunar themed arena with various obstacles in place. Given a destination, the robot autonomously navigated there avoiding the obstacles. Alternatively, drivers wearing virtual reality headsets could drive the vehicle around the arena.
Here’s some more information, and some pictures (click to enlarge):
eInfochips was showing a vehicle which used vision for goal setting using the Jetson TK1. As seen in the video above, the vehicle could identify a red ball, and then drive to it autonomously. There were different colored balls in the booth, the robot would identify the red ones and drive towards them. Simple, but effective, computer vision demonstration. Another demonstration in the booth showed a Jetson TK1 being used to track a vehicle from a video image in real time.
There were other vehicles at the show that were Jetson based, such as a tracked robot seen at previous shows presented by GE. The proliferation of vision based robots based on the Jetson TK1 platform starts to tell the story of the promise of what a powerful, mobile GPU SoC can deliver on a robotic platform.