Researchers have developed algorithms to free the Cheetah robot from visual devices, suggesting potential applications in rough terrain access
The Cheetah 3 robot at MIT is based on blind locomotion and is designed to handle unexpected terrain variations without external cameras or environmental sensors. This model proposes a solution to the inaccuracies or limitations of coverage that visual devices possess.
Overcoming dependence on cameras and sensors
The Cheetah 3 robot weighs approximately 41kg (90 pounds) and is said to be of a size comparable to that of a full-grown Labrador. It is programmed in a way to avoid dependence on vision as it can be noisy and unavailable sometimes.
Researchers said that the robot relies on tactile information and can also stretch backwards and forwards and twist from side to side. It has been designed to carry out tasks that can be dangerous or inaccessible for humans to execute.
The team has developed contact detection and model-predictive control algorithms that form the foundation of the robot’s ability to ‘blindly’ move in unstructured terrain while maintaining balance.
The aim of the contact detection algorithm is to tell the robot the time and kind of forces to be applied in order to move the body in the right way. The model-predictive control algorithm intends to predict the multiplicative positions of the robot’s body and legs half second in advance.
Exploring dangerous and inaccessible zones
The researchers tested the robot in experiments on laboratory treadmills and staircases. Random objects such as wooden blocks and rolls of tape were also spread on the surfaces.
The robot’s movements are developed from the idea of a human’s movement with closed eyes. It is a step towards preparing a preconceived mental model for where the ground could be.
This can eventually free machines from their dependence on sensors and cameras, making it possible to ‘blindly’ function in areas that are too remote or dangerous for humans to access.