Autonomous mobility scooter runs new algorithms that work both indoors and outdoors.

Move over, autonomous
cars; there are now self-driving scooters. Massachusetts Institute of
Technology’s (MIT) Computer Science and Artificial Intelligence Laboratory
(CSAIL) team has unveiled a driverless scooter
that uses algorithms to not only allow users to sit back and ride but also
can be applied across a spectrum of
different vehicles. It allows those who are mobility-impaired to switch from
scooters that move inside buildings to golf carts in car parks to autonomous vehicles
on the road.

mit-scooter-1

The device demonstrates a new comprehensive
mobility system designed by CSAIL in partnership with the National University of Singapore and the
Singapore-MIT Alliance for Research and Technology by using sensors and
software from experiments with autonomous cars and golf carts. The researchers
note that the reason they reused this
configuration was to concentrate on new uniform navigation and control
algorithms.

The algorithms are based on three layers of software. The first layer is made of low-level algorithms that respond
quickly to changes in the vehicle’s environment in
order to avoid accidents and stay on course. The second layer is
composed of localization algorithms used to construct maps and navigate. The
third layer is a scheduling algorithm that allows the vehicle to operate and be shared with several users.

These new algorithms work both indoors and
outdoors and could even navigate through MIT’s famous Infinite Corridor, which
is a straight hallway, 825 feet long, and is
deemed difficult for robots to orient themselves. The algorithms in the
scooter could handle it without getting lost.

As for the ambitions behind the scooter, MIT’s
engineers see the development as a potential standard system for all types of autonomous
vehicles. By setting this standard, users could switch from one type of ride to
another, such as from a scooter to car to golf cart depending on which is most valued for the journey.

According to the team, the new algorithms are
more practical for performing reliability analysis of the system’s performance.
The design makes it easy to verify if the layers are properly working, is less
complex, and requires no need to reinvent systems for different vehicles.
Additionally, data can be transferred
from one vehicle to another, and because the algorithms interact with the environment
in various devices, they learn to improve their performance.

mit-scooter-2

The scooter debuted in April 2016 at an event
where over 100 people took part in testing the software. Before doing so, the
participants were asked to rate the safety on a scale of one to five. According
to the team, the scores were between an average of 3.5 and 4.6.

The team’s complete
findings can be found here.

Source: News Atlas