Toyota Research Institute believes that as the world’s population ages, robotic capabilities will become an important avenue for people to age in place longer, with a higher quality of life.
For robots to learn fast enough to take up this mantle, TRI says, something called “fleet learning” has to occur. That would involve teaching one robot to learn to perform a task, either from a person or a simulation. The robot would then share this knowledge with other robots, allowing them to perform the task in new situations and achieving an exponential spike in robotic capabilities.
Robots, like automated cars, continuously monitor their surroundings to gauge a path forward. Still, even teaching them to make a pot of coffee has its complexities, especially if someone moves the coffee or puts the filters in a different drawer.
TRI proposes to teach robots with an immersive telepresence system, in which there is a model of the robot, mirroring what the robot is doing. A human teacher sees what the robot is seeing live, in 3D, from the robot’s sensors. The teacher then picks different behaviors to instruct and then annotate the 3D scene, such as associating parts of the scene to a behavior, specifying how to grasp a handle, or drawing the line that defines the axis of rotation of a cabinet door.
A human’s creativity to, say, use the robot’s hands to perform a specific task, makes leveraging and using different tools easy, allowing humans to quickly transfer their knowledge to the robot for specific situations.
Toyota says its home-care robots will do their learning in actual homes.
Currently, its system can successfully perform a relatively complex human-level task about 85 percent of the time. This includes letting the robot automatically try again if it recognizes that it has failed at a specific behavior. Tasks are typically made up of about 45 independent behaviors, which means that every individual behavior results in success, or recoverable failure 99.6 percent of the time.