If you live in the town of Frisco, Texas, and see a bright orange minivan on the road, chances are very good that it is a self-driving car from Drive.ai that has officially launched its on-demand service in order to take people to and from places they want to go.
Drive.ai has been testing its routes in Frisco since April, collecting data for the company’s deep learning artificial intelligence systems, documenting the scenarios and creating custom simulations to improve how the vehicles perform on the roads. During these four months, Drive.ai has geo-fenced routes in Frisco, which includes crossing six lanes of traffic, parking lots, pulling over and dropping passengers off on-demand.
The vehicles are orange because Drive.ai wants the cars to be easily seen by other cars or by pedestrians. There are also front facing, rear and side LED screens that give directions to pedestrians in order to help with the safety aspect of when crossing the streets or identifying where the car may be turning.
While the vehicles are designed to be driven autonomously and software makes the vehicles heed all traffic rules, there will be safety drivers on board. Drive.ai uses simulations to create scenarios that include unlikely human behavior — a person jaywalking, random objects on the road, heavily populated streets, navigating busy parking lots or bicyclists.
These scenarios were tested in order to teach the vehicles how to approach variable driving conditions so when it happens in a real-world setting it will have the experience to know what action to take or not take.
From driving the streets of Frisco, Drive.ai was able to convert the logs of these trips to simulations in order to see how the technology responds to different scenarios. Original simulations were also introduced to craft unusual situations for the vehicles such as piloting around double-parked cars or steering through tight turns. It is a stress test for the vehicle technology to ensure safety for the now launched on-demand program.