As engineers, one of our biggest challenges is to lower costs. Being able to prove concepts with a nice, large budget and having access to sophisticated sensors and mechanisms is great, but ultimately, if humanity, or at least the consumer, is to benefit from our work, we will have to find a way to make it cheap, simple, and with the same reliability as the "fancy" version.
Google, using its substantial surpluses of money, can afford to attempt fully automated vehicles with 360-degree rotating scanners and state-of-the-art LIDAR totaling around $70,000. Researchers from the Oxford Mobile Robotics Group are building a cheaper version that can use sensors in smart ways to minimize cost and ultimately increase availability. As opposed to the autonomous cars of Google and Audi, the Oxford group has found a method for using cheap sensors with an adequate computer network and architecture to do the many jobs an autocar can do.
The sensory and autonomic systems that make up Oxford's autocar, named RobotCar, seem remarkably simple. The Oxford team has created a car with two laser scanners under the front bumper with an 85-degree f.o.v. and a pair of stereo cameras, which continuously analyze and memorize environments. After a few weeks of study, it learns to navigate your morning commute, slowly learning more and more over time.
The scanning system sees obstacles 164 feet in front of the car, and it will not proceed until the roadway is clear. It also anticipates movements, like pedestrians crossing the road. The system is not dependent on GPS, because it's less than ideal for an autonomous car. RobotCar needs accuracy within centimeters in all weather conditions, and in indoor or subterranean parking garages, etc. In addition, the team is not counting on big infrastructure additions to accommodate RobotCar. Instead, the car can "see" road and lane markings, traffic lights and signs, and even accesses aerial views.
The car thinks, using three computer systems. The Main Vehicle Computer (MVC) can control all the car's systems including sensory and automotive, of course, and oversees communication between the Lower Level Control (LLC) and an iPad mounted on the dash. When the car's MVC recognizes a route, it asks the driver to take over with a prompt on the dashboard iPad. If there is any discrepancy between any of the three computers, the car will not ask to drive, and if the drive does not respond to the prompt at all, the car will slow to a halt. At any time while the car is in motion, the driver can tap the breaks, to instantly regain control.
The 22-man Oxford team envisions a future where cars earn more driving responsibilities little by little, doing more detailed and sophisticated work by themselves. Combining this philosophy with a cost as low as $150, RobotCar could potentially revolutionize the auto industry. The group wants to develop cars that offer autonomous features for an additional cost, low enough for middle class incomes to afford -- a situation similar to the Ford Model T's but taking the automobile to a whole new level in the 21st century.
It will be interesting to see how autonomy alone influences car sales. It may not be the flying car of the future, but a fully automated fleet might let us fly through commonly congested corridors. This dream is still far off, and there will be a time when humans and RobotCars share the roads. What will become of road rage?
The research group is already testing RobotCar on private roads near the Begbroke Science Park near Oxfordshire and is in talks with the British department of transportations for testing on public streets.
To watch more videos, go to YouTube and search "RobotCar UK."