Gearing Up For A Driverless Future
There’s been a lot of talk lately about autonomous cars. Countless companies are claiming they’ll be the first to release revolutionary self driving cars. Competition is heating up between companies like Volvo, Google, Uber and Tesla to name a few. Uber piloted self-driving cars in Pittsburgh last month, overseen by a Driver and Engineer. Singapore-based software developer nuTonomy has also launched trials into self-driving cars with hopes to launch them in 2018 for public use.
Each year, a mammoth 1.2 million people are killed on the world’s roads. 90% of these tragedies being down to ‘human error’ – drunk driving, tiredness, road rage, visual impairment. The technology used in creating the driverless car makes it possible to reduce the number of fatalities.
There are numerous systems working together to power driving the car. Lidar Sensors, highly accurate sensing systems that work by rotating and bouncing pulsating laser light off of the surroundings. This produces ‘time-of-flight’ measurements of range, accurate to the centimetre, producing millions of data points per second. The car is able to maintain a precise position thanks to ultrasonic sensors in the wheels. Altimeters, gyroscopes, and tachymeters are also at work to make sure the car keeps the correct and safe distance from the curb, in lanes and from other vehicles.
Deep Learning algorithms work to process this sensory data, with Convolutional Neural Networks processing the visual data points collected via the 360° camera. These algorithms are fed large sets of data on road signs, lanes, traffic light systems, cyclist and pedestrian signals, including being taught human-like driving. This allows the algorithms to help navigate the vehicle by making informed decisions based on this data. This is all processed and analysed by the central computer system which uses the information to assist in the steering, accelerating and braking.
Without a doubt, we will see a Big Data-driven future of connectivity, with car data fed into a centralised system. Not only will this help decrease accidents but also make that morning commute a bit easier by eliminating traffic congestion through pre-planned routes and the removal of human navigational errors and road rage. Tracking owner data, as well as vehicle data, does naturally raise issues with cybersecurity and data protection. Authorities will have to review insurance policies, licenses and liable to suit this new driver-less world. On a larger – and costlier scale – there will also be significant changes to infrastructure, land speeds, road markings and signage. Not to mention the problems surrounding job loss as the need for public transport, haulage and chauffeuring services look to be all but eliminated.
But just think, the next generation of drivers may never have to take a driving test – or at least the kind we have now. Car accidents and road fatalities could be a thing of the past. Disabled and vision impaired people will have the freedom to ‘drive’ themselves and not rely on laborious public transport journeys to get about.
In the US, drivers get into fatal accidents 1 in every 100 million miles driven. Self-driving cars will need to prove that they are at least this safe. Google’s self-driving cars only managed 1.3 million miles between 2009 to 2015 so it could take years for companies to really prove the driverless cars’ safety.
Although legislation and regulations will need to be ironed out and public trust won, with large scale investment and countless companies hard at work on this technology – the finishing line is definitely within sight!
Other blog posts
Deep Learning, Guest Blogs
Python Frameworks for Your Upcoming Deep Learning and Data Science ProjectsRead More