Behind the wheel, your brain does a lot of processing you take for granted, such as calculating the paths and speeds of nearby vehicles so you can successfully make a lane merge. Ford is turning to Stanford and MIT researchers to come up with computer algorithms to mimic that processing of environmental data.
Last month, Ford showed off the Fusion Hybrid autonomous car research vehicle, fitted with four lidar sensors, it was using to develop future driverless systems. The company will implement new algorithms developed at the universities to test these driving behaviors in the cars.
MIT gets to take on the task of predictive path analysis based on current sensor data. The researchers can look at the path and speed of a car, and use physics to determine where that car is likely to be at a set time in the future, such as 10 seconds, 30 seconds, and possibly a full minute. Likewise, the algorithm could determine where that other vehicle could not possibly end up in the same amount of time.
A predictive algorithm such as this could tell an autonomous car where it is safe to make a lane change, for example.
Ford notes that this predictive analysis will also apply to pedestrians, letting the car know where people in its vicinity are likely to be a few seconds into the future. That aspect of the research could make a driverless car much safer in urban areas with a lot of foot traffic.
Stanford's part will be to work out for autonomous cars something drivers do every day, maneuver to get better forward visibility. In what Ford calls a "peek ahead" move, drivers frequently shift within a lane, either steering or adjusting speed, to see around a bigger vehicle just ahead. Stanford researchers will try to emulate that behavior in a car.
Of course, rather than the limitation of a driver's two forward-facing eyes, the Fusion Hybrid is equipped with sensors providing a broader range of perception. The Stanford research could implement behaviors such as moving the car to the right side of a lane to get a better view around a left-lane-hugging vehicle.
The Stanford research won't be quite as vital if vehicle-to-vehicle (V2V) systems are implemented in the near future. With these systems, all cars broadcast their position and speed. V2V would let a car know what traffic ahead of a larger vehicle is doing, before a driver or even sensors could perceive it.
Ford demonstrated its V2V research during the recent CES.
As for fully autonomous cars, Ford sets a date beyond 2025 for these to hit the road.