MIT Made a Blind Cheetah Robot That Can Leap Into Dangerous Situations

Why the team ditched the vision sensors. Also, Uber files for controversial patents, and CanguRo can be your R2-D2.

Finally, Your Own R2-D2

Researchers have developed a morphing autonomous tricycle personal assistant that they call the CanguRo robot. It got the name because the team says that it resembles a kangaroo when in personal assistant mode.

According to Digital Trends, it gets its name from the Italian word for kangaroo, even though it certainly sounds like it's English counterpart as well.

Developed by a team from the University of Tokyo and the Future Robotics Technology Center (fuRo) at Chiba Institute of Technology, CanguRo works a lot like R2-D2 in Star Wars. It will follow its master, using AI, it will even talk to you using a series of familiar beeps and chirps, and it can be used as a communication device — no word on whether or not it is a hologram. Using an app, you can even send it to your location.

So, if you want to walk home, the robot can follow you and carry your bags, or you can transform into a tricycle and have it carry you to your destination at 6.2 mph.

The CanguRo uses image recognition combined with mapping and positioning data for autonomous control, although you can still take control if you wish ... or if your careening into a pothole, you just lean left or right if you want to turn.

According to The Japan Times, CanguRo is one of a series of machines that is called RidRoid. The robot is still under development.

Next, the researchers are trying to ditch the beeps and chirps and make it talk.

Uber Files for Hot-Button Patents

Uber is stirring up controversy with a new patent that others have referred to as common sense.

According to reports from Motor1 and Motor Authority, the ride-sharing company filed a patent for a new concept that would keep riders out of unsafe areas and let you know if your approaching driver is going to pull up in a beater.

The patent, which was filed at the beginning of the year and updated at the end of June, would offer a "safe routing" option. The app would pull information from social and news media as well as publicly available data on violent crime and let the rider choose to take a safer, but possibly longer, route.

The passenger would also be able to check out weather, current traffic conditions, and even info on the driver. That info would include the safety of the vehicle, which would be derived from the make, model, year, and health of the car.

Some say that this is a way to discriminate against poorer neighborhoods, but you could likely still go through or to the neighborhood, and the trip might even be cheaper since it is shorter. If it's just the option to go around the one street in town with statistically higher rate of your Uber getting plugged by indirect fire, that doesn't sound discriminatory to me. That sounds like common sense.

As for the beater meter, you’ll just have to wait for a classier car if you can’t stand anything made before 2017, but it might also incentivize drivers to upgrade their rides when they start missing out on fares.

MIT Made a Blind Cheetah Robot

Last week, MIT celebrated freedom on the Fourth of July by showing off the Cheetah 3.

Cheetah 3 is a new robot from the Biomimetic Robotics Lab that can run 6.7 mph, climb stairs, and even make a nearly 3-foot leap. What's crazy is that it does all of this and more while essentially blind.

According to MIT, the 90-lb robot ditched the cameras and instead uses "blind locomotion," like when you're stumbling towards the fridge in the middle of the night for some water.

According to the researchers, vision can be loud, inaccurate, and/or unavailable, which can slow down the robot or render it useless. Which would be a problem if, for example, the robot was the WALK-MAN 2.0 and it was trying to save you from a burning building.

So, how does the Cheetah 3 blindly walk up staircases and navigate unknown terrain? It has a pair of new algorithms: a contact detection algorithm and a model-predictive control algorithm. The contact detection algorithm tells the Labrador-sized robot if and when it needs to shift it's weight to stay balanced.

If one leg steps into a pothole, it tells the other legs how the need to compensate to keep the Cheetah moving forward. The model-predictive control algorithm helps the Cheetah maintain balance when someone kicks it, or as in the tests, jabs at it with a 2x4. It predicts how the robot should compensate to for the force.

The robot also has a number of improvements from the Cheetah 2, notably the ability to stretch backwards and forwards, and twist from side to side. this will help prepare it for future roles as a first responder in otherwise inaccessible areas, like one of the many explosions at industrial facilities this week alone. 

The researchers will present their findings in October at the International Conference on Intelligent Robots in Madrid.

This is Engineering By Design.

More in Product Development