The News Wheel
No Comments

Toyota Research Institute Discusses Development of Human-Assist Robot

Decrease Font Size Increase Font Size Text Size Print This Page
TRI Kitchen Robot
Photo: Toyota

The Toyota Research Institute is hard at work on developing robots that can assist you in daily tasks, such as putting away the groceries and the dishes, folding the laundry, and tidying up before guests come over.

Dr. Russ Tedrake, TRI Vice President of Robotics Research, says that this kind of technology is important because while it might only seem like a convenience to many people, “For others, including our growing population of older people, applications like this could be the difference between living at home or in an assisted car facility.”

But getting a robot to do these tasks is much harder than popular science fiction might make it seem. Tedrake asks: What if a robot could load the dishes, but it broke a dish once per week? What if your child brought home a “#1 DAD!” mug she painted at school, but the robot discarded it into the trash because it didn’t recognize it as an actual mug?

Unlike the controlled and programmable environments of the factory where robots are typically developed, Tedrake says the home is a “wild west” full of objects and circumstances that are impossible to completely predict. It is therefore necessary to build a robot that can adapt to unpredictability, but despite advances in artificial intelligence and machine learning, there’s still a long way to go.

The Toyota Research Institute has been developing a dishwasher-loading robot that uses stereo cameras mounted around the sink and deep learning algorithms to identify objects in the sink before placing them in the dishwasher. There are thousands of variables involved: the shape, size, and orientation of the objects, whether the object even belongs in the dishwasher, the process of gripping and releasing the objects, and so on.


Toyota Safety Sense: A semi-autonomous safety net for your car

In the video below, you can see that TRI’s robot is actually pretty good at its job, which Tedrake says is thanks to the company’s use of simulation. “An exciting achievement is that we have made great strides in making simulation robust enough to handle the visual and mechanic complexity of this dishwasher loading task and on closing the ‘sim to real’ gap,” he says. “We are not able to design and test in simulation and have confidence that the results will transfer to the real robot.”

The robust correlation between simulation and reality allows TRI to do all of its development in simulation, which it uses to find and test scenarios in which the robot might not behave as desired. Because of the law of diminishing returns, this is actually one of the biggest hurdles TRI faces. “The search space is so huge, and the performance of the system so nuanced, that finding the corner cases efficiently becomes our core research challenge,” Tedrake explains.

The dishwashing robot is just a start. TRI isn’t just hard at work on building home-assistance robots, but AI technology that will drive our cars as well — and the margin for mistakes on the road is arguably even smaller than in the home. No wonder it’s taking so long.


Family Fun: Places to go on an adventure this summer