Cars have been around for years now, with one of the earliest cars ever made was back in 1769. Over the course of three centuries, the motor car has evolved into more than just a machine made for convenience. The car as we know it today has transformed into a tool for comfort and competition. While the internal mechanics and general purpose of cars has evolved through the decades, one of the main components of the driving experience i.e. the driver behind the wheel always remained constant. However, the rapid advancements in various technological fields is all set to change that.
The birth of transistors and subsequent developments in the fields of digital computing has led to advanced AI (artificial intelligence) projects from various corporations such as Google, Apple, Boston Dynamics and Honda. Despite the sci-fi legacy of artificial intelligence, we are indeed implementing AI in almost every possible field already. While most real world applications of AI is quite dull, some establishments produce immensely interesting and questionable products such as the very first non-human world chess champion (IBM’s Deep Blue) to a rat brain (The SyNAPSE Project).
The automotive industry too has begun the march towards AI. During the late 20th and early 21st century, contemporary advancements in semiconductor technology disallowed the implementation of AI in cars. These days, premium car makers such as Mercedes Benz, Audi and BMW now fit various AI cores in numerous safety systems. Car manufacturers are now taking the next step into the future: autonomous vehicles.
Autonomous driving in its most basic sense is a vehicle that drives itself with the belief that a inorganic machine behind the wheel should theoretically reduce the number of accidents that occur on the road. While humans fall prey to distractions, rage, carelessness or general ineptitude, a machine specifically designed to manage traffic and navigate itself to a destination does not drop the ball. The CPU or computer of the car manages the steering, brake, throttle and gear position and constantly monitors its surroundings for dangers that lie ahead. Various manufacturers implement driverless cars in various different ways. Currently, Tesla, BMW, Audi and Google have been making headlines for their achievements in the field of autonomous driving.
For an autonomous system to navigate itself, it must first be able “see”. For this, common camera systems are not enough as raw video/image data is near impossible to accurately decipher by our silicon counterparts (this is also the basic idea behind captcha or Completely Automated Public Turing test to tell Computers and Humans Apart). For better accuracy and resolution, most autonomous cars implement a combination of 360 degree RADAR & LIDAR systems, ultrasonic sensors and cameras. RADAR and LIDAR systems implement radio waves and laser beams respectively.
These radio waves and lasers are shot in every possible direction and the time taken for them to return is measured. This provides the distance between the point of origin (car) and the obstacle (wall, other cars, pedestrian etc). All these systems help generate a 3D map of the environment and endows the car with “sight”. For a better understanding of what a car perceives as vision, ScanLAB Projects conducted an experiment for the New York Times providing insight into the ocular capabilities of a driverless car.
The on-board computer’s memory may also store map data and high resolution images of streets to compare with its immediate surroundings and determine where it is and where it should be going. The on board computer also contains algorithms on basic road etiquette, ovetaking maneuvers, laws that govern the road and more importantly collision avoidance tactics. To avoid any mishaps on the road, most drivers must have rapid response times. While computer systems these days are immensely powerful, speed is of the essence. As such, developers such as Google make use remote computer farms for computation capability to ensure the on-board systems are under less stress and focus more on immediate responses. The computer then adjusts the steering, throttle and gears to pilot itself to the specified destination.
Driving is a skill developed over time. Over the course of many years, a driver learns where the potholes maybe, how reliably his car handles in various weather conditions, estimating possible collision situations and molding (this very human characteristic may prove to be a boon or bane depending on circumstances) his fellow drivers’ mental state, driving standards and etiquette to adjust his driving style. With an autonomous car however, none of these things are possible as these scenarios are decidedly anthropomorphic in nature. The perils of autonomous driving are multiple.
However, the pros far exceed the cons and the rise of general AI will usher in a new era for humans and cars alike.