Transportation
❯
Autonomous driving
❯Autonomous trains (Unattended train operation (UTO))
Autonomous apron truck
For:
Transporter
Scope:
Automated transportation of luggage (carts) to requested destinations on an airport apron while following local traffic rules and resolving unplanned
conflicts.
Goal:
Improve Operation Efficiency
Government
Defense & Military - Autonomous Vehicle
For:
Military Operatives.
Scope:
Using computer vision to implement perimeter safety for Autonomous Vehicles.
Goal:
Improved Product Development / R&D
AI components for vehicle platooning on public roads
Scope:
Trains of vehicles that drive very close to each other at nearly equal speed
(platoons) on public roads, in particular platooning trucks on motorways.
Goal:
Improve Operation Efficiency
Autonomous trains (Unattended train operation (UTO))
Scope:
Freight and passenger trains operate autonomously, excluding any crew
presence on board, but with remote operator attention involved.
Goal:
Improve Operation Efficiency
Self-driving aircraft towing vehicle
Scope:
Self-Driving towing vehicle for aircrafts, operating on an airfield autonomously.
Goal:
Improve Operation Efficiency
Autonomous trains (Unattended train operation (UTO))
Goal:
Improve Operation EfficiencyProblem addressed
The critical objective of automation in trains is to provide extra reliability and
safety and to prevent accidents on railways, which tend to be caused by human
error. Moreover, the provided innovation leads to energy consumption
optimization, transport capacity increases, and, eventually, possible reduction of
personnel costs due to the autonomous operation.
Scope of use case
Freight and passenger trains operate autonomously, excluding any crew
presence on board, but with remote operator attention involved.
Description
There is a lot of information about self-driving automobiles.
Developing computer vision technology, reliable navigation,
and radio communication makes creating self-driving trains
technologically feasible.
Compared to cars, trains have a long braking distance. This
means that autonomous trains are expected to have a unique
obstacle detection system, which can spot obstacles up to 1
000 m away and more.
Both conventional and autonomous railway systems consist
of fleet and infrastructure. Current interaction between
locomotive and dispatcher is realized by voice
communication. For autonomous train use, digital
communication with formal commands for train control is
necessary.
Key AI development realized into the obstacle detection
module can be achieved with both computer vision methods
by processing data received from sensors (LIDARs, RADARs,
infrared and electro-optical cameras) and by positioning and
localization based on prior electronic map information and
obtained data from GPS information. This system can work
under differences in light, weather, and timing conditions.
The data collected from sensors with a varied range of
actions and purposes is processed by classical image analysis
and deep learning approaches; it is then fused. Methods such
as semantic segmentation, object detection, LIDAR points
clustering, tracking, localization, and mapping are used. All
in all, this leads to clear scene perception and safety system
responses. The machine can trigger the alarm, halt, apply the
brakes, or accelerate based on information about the
environment.
However, a remote driver is still needed to resolve
complicated cases, which the on-board system is not able to
process correctly. Considering that the system's priority is
safety, such examples most commonly include false-positive
object occurrences. It is important to stress that one remote
driver operator can control the performance of several
automated trains at the same time.
Three autonomous shunting locomotives are already in
operation at Luzhskaya Marshalling Yard in Moscow, Russia;
and parallel deployment for passenger trains is currently
under testing on the Moscow Central Ring.
Live Video
AI: Perceive