ROS, Human-like movement, Autonomous Navigation and more!

Hardware
Rosie now has two arms and a claw (to help with picking up objects if the hands don't work at first)!

Billy learned how to weld (in a week!) and built Rosie a differential drive mobile base with 12V motors which will allow her to navigate around.

Lastly, Rosie received an upgrade to her neck which gives us additional degrees of freedom. This has enabled us to achieve a more "human-like" motion (see below for more).

Rosie's new neck (three pistons instead of one)

Rosie on her new wheels!

Robot Operating System
We decided to make the leap to Robot Operating System (ROS), which should give us more functionality and freedom in the future. Unfortunately, this meant rewriting 90% of Rosie's code base. She now runs a custom code stack written in C++ and Python and integrates with AWS and Google cloud services (like speech synthesis). I'll be the first to admit learning ROS was extremely difficult, but we persevered through numerous challenges and are beginning to see the benefits.

More Human Like Movement
We have been working with a fellow Inmoov builder from the west coast to produce more life-like movements.

Inverse Kinematics Integration
We're about 75% of the way with adding inverse kinematics to Rosie. In previous demonstrations, we were hardcoding each servo's position. With inverse kinematics, we are able to tell Rosie to move her hand to a point in three-dimensional space. Soon, we will be able to detect an object with the Xbox Kinect sensor in her chest and tell her to position her hand in such a way to grab the object. This is dynamic and will be able to adjust on the fly if the object moves. Really cool!

Below is our computer model of Rosie with the hand set to a point in 3D space. Notice how the physical bot can now move to the same positions- all with zero hardcoding!

Rosie moving to the position

Rosie's 3D model moving

 

Rosie seeing the world in 3D for the first time
through the eyes of her Xbox Kinect.

Autonomous Navigation
We've added LiDAR to Rosie - the same technology found in self-driving cars. We're in the process of building a Laser Scan map of the first floor of my house using a process called SLAM (Simultaneous Localization and Mapping). Soon (a week or two), we will be able click a location on the map and the base will travel there. It will also avoid obstacles, such as furniture or cats! Once we get the calibration right we'll add Rosie to the base.

Example of laser scan map

What's next?

• Object detection - teaching her how to correctly identify objects.
• Gripping - teaching her how to pick up objects.
• Retrieve a beer - our short term goal is for Rosie to get a beer out of the fridge and bring it to a person located in another room.