Get best of the week

Joint work of unmanned vehicles can accelerate movement by 35%


Researchers have shown that a fleet of unmanned vehicles, combined together to provide smooth motion, can optimize traffic flow by at least 35 percent.
“For the safe use of autonomous cars on real roads, we need to know how they will interact with each other.”
- Amanda the Prophet
Researchers at the University of Cambridge programmed a small fleet of miniature robotic cars to ride on a multi-lane highway and watched how the traffic flow changes when one of the cars stops.

When the cars drove separately, the cars behind the stopped car had to stop or slow down and wait for the clearance in the stream, as is usually the case on a real road. A queue quickly formed behind a stopped car, and the general flow of traffic slowed down.

However, when the cars communicated with each other and drove together, as soon as one car stopped in the inner lane, he sent a signal to all other cars. Cars in the outer lane, in the immediate vicinity of a stopped car, slowed down a bit, so cars in the inner lane could quickly drive past a stopped car without having to stop or slow down significantly.

In addition, when a human driver was driving with unmanned vehicles, who moved along the track in an aggressive manner, other cars could give way to avoid an aggressive driver, thereby increasing the level of safety.

The results, which will be presented today at the International Conference on Robotics and Automation (ICRA) in Montreal, will be useful for exploring how autonomous cars can communicate with each other in the future, as well as with cars driven by a human driver, on real roads.

"Autonomous cars could solve many different problems associated with driving in cities, but there must be a way to work together," says Michael He, co-author and undergraduate student at St. Johns College, who developed the algorithms for this experiment.

“If various car manufacturers are developing their own unmanned vehicles with proprietary software, then all of these cars should interact effectively with each other,” said Nicholas Hildmar, also a co-author and undergraduate student at Downing College, who designed most of the hardware for the experiment.

These two students completed their undergraduate research project in the summer of 2018 in the laboratory of Dr. Amanda the Prophet of the Cambridge Department of Computer Science and Technology.

Many existing tests for several autonomous cars without a driver are performed in digital format or with large-scale models that are either too large or too expensive to conduct experiments in a vehicle fleet.

Starting with low-cost, large-scale models of commercially available cars with realistic steering systems, Cambridge researchers have adapted cars with motion capture sensors and the Raspberry Pi so that cars can communicate via Wi-Fi.

Then they adapted the lane change algorithm in unmanned vehicles to work with the fleet. The original algorithm decided when the car should change the lane, based on whether it is safe to do and whether changing the lane of the car will help you move faster on the road. The adapted algorithm allows cars to be denser when changing lanes and adds a safety limit to prevent accidents at low speeds. The second algorithm allowed cars to detect a car in front and free up space.

Then they tested the fleet in “egocentric” and “cooperative” driving modes using both normal and aggressive driving, and watched how the fleet responded to a stopped car. In normal mode, cooperative driving improved traffic flow by 35% compared with egocentric driving, while in aggressive mode - by 45%. Researchers then tested how a fleet reacts to a single person-driven car using a joystick.

“Our design allows us to conduct a wide range of practical, inexpensive experiments on autonomous cars,” said the Prophet. “For the safe use of autonomous cars on real roads, we need to know how they will interact with each other to improve safety and optimize traffic.”

In further work, the researchers plan to use the fleet to test multi-vehicle systems in more complex scenarios, including roads with a large number of lanes, intersections, and also plan to work with a wider range of types of vehicles.


image

About ITELMA
- automotive . 2500 , 650 .

, , . ( 30, ), -, -, - (DSP-) .

, . , , , . , automotive. , , .

Read more useful articles:

More articles:


All Articles