MIT's Cheetah 3 Robot Climbs Stairs With No Visual Sensors Or Cameras To Help

MIT has a new robot that researchers have been working on that could prove very useful in a disaster zone or environment that is inaccessible to humans. These same environments could likely have very low visibility meaning that cameras and visual sensors on the bot would be of little or no use. MIT's Cheetah 3 robot can climb stairs without the aid of any visual sensors or cameras.

Advertisement

With the addition of stair climbing ability, Cheetah 3 can now leap and run across rough terrain, climb stairs littered with debris, and recover balance quickly if shoved or pulled over. The bot weighs in at about 90 pounds putting it on par with a full-grown Labrador. The bot replaces its eyes with the ability to "feel" its way around its surroundings using "blind locomotion" in a manner similar to how we find our way through dark rooms.

The researchers say that visual systems might not always be accurate or available so they are relying on tactile information to allow the bot to be faster. The fancy vision-free capabilities of Cheetah 3 will be presented in October at a conference on intelligent robots. Along with the blind locomotion capability, the researchers will also show off the bots improved hardware.

Advertisement

That improved hardware gives the bot an expanded range of motion compared to previous generations. This allows it to stretch backwards and forwards and twist from side to side. Part of the reason the bot is adept at moving without seeing comes in new algorithms the team has developed.

One of the algorithms is for contact detection and the other is a model-predictive control algorithm. Contact detection allows the bot to determine when the best time is to change from having a leg in the air to having it on the ground. That allows it to react differently when stepping on a large rock or stepping on a small twig. The other algorithm can calculate multiplicative potions of the robot's body and legs a half second in the future.

SOURCE: MIT

Recommended

Advertisement