Five years ago, Toyota announced a $1 billion commitment to its Research Institute (TRI) and hired hundreds of engineers to work on forward-thinking robotics and AI-driven products, such as self-driving cars. This week, TRI gave journalists a chance to attend a virtual open-house (part of which you can watch in the video embedded below) to see what the labs have been building.
The bulk of the prototypes shown during the event are meant to help people out around their homes. The idea of a helper robot has been around for decades, but it has already begun to gain more real-world traction, especially in countries such as Japan where an aging population could cause shortages in support staff.
TRI’s primary home robot design is built to hang from a special set of tracks built into a home’s ceiling. We typically imagine robot butlers rolling or actually walking around our spaces (due at least in some part to the awful robot butler in Rocky IV), but TRI says mounting a bot on the ceiling brings considerable advantages.
First, the robot has a simpler time learning the layout of the home—it doesn’t need to learn complex paths around obstacles to get from one place to another because there’s a kitchen island or a new recliner to get in its way. That also means the robot won’t get under foot as someone tries to navigate around their own home. TRI’s bot can fold up tightly to the ceiling when it’s not in use, so it effectively takes up zero usable space.
TRI also claims that the top-down viewpoint gives the machine a better point of view for observing its own actions and the relative position of different objects it might have to manipulate. Toyota is big on robots learning from limited sets of parameters. So, if you wanted the robot to learn to wipe down the countertops, a human could perform that action in VR and the robot would understand the human’s actions and mimic them in the real world while monitoring variables. If you wanted to teach the robot to clean your countertops, you could do it in VR and it would understand going forward. Also, through what’s known as fleet learning, a human could teach one bot, and other networked machines would get the same lessons.
Click here to read more.
Source: Popular Science