It’s a home mobile robot with stereo cams and a pair of arms, capable of navigating around via the stereo cams and pciking up and carrying small objects.
The robot is controlled from a laptop running Windows (don’t kill me, please, Windows + Visual C++ is in my view still the easiest, most flexible environment to program in and when the team is of size 1 that matters a lot). The machine vision and navigation algorithms and software have been basically written from scratch and almost entirely implemented in OpenCL to make use of the vector processors and multiple cores in Intel CPUs, and I am proud to say it takes less than 50% of the processing power of the laptop which is an old Dell XPS M1210, running a 2GHz Core 2 Duo. The drive system is the RD02 - 12v Robot drive from Robot Electronics
robot-electronics.co.uk/acatalog/Drive_Systems.html
The servos in the arms are all modified with individual PIC microcontrollers to provide some smart feed-back functions such as measurement of the actual position and the current going through their motors (proportional to the force on the particular joint - this is how the robot detects when I grab the beer can), and also flashing of LEDs to detect the exact spatial position of the joint.
The drive system and servos are controlled through a single I2C bus which is connected to one of the laptop USB ports through a USB-to-I2C bridge robot-electronics.co.uk/acatalog/USB_I2C.html which simplifies the programming a lot.
In the video clip most of the software is running on the laptop, the top level logic however is running on a desktop and communicating with the laptop through WiFi, which is done to simplify programming and debugging of the top level logic, which can get quite complex. Except for the initial pressing of a start button there was no manual intervention in the robot control throughout the whole video clip.
We need details!
Thanks, I’ve added some details
Really, really cool. You did a fantastic job. You deserve the beer.