11-08-2016 | | By Paul Whytock
There are many industry experts that believe Lidar sensing will help to improve the safe operation of autonomous cars but whether it will win over the vast majority of dubious drivers remains to be seen.
Currently most drivers say they don’t want to be driven by a robotic car. Nearly 80% of mature motorists say no thanks to autonomous cars whilst the younger generation are more open to the idea with about 50% saying they would try them.
Despite these negative reactions there are companies that are pouring substantial amounts of corporate cash into creating safe autonomous vehicles. Notable among these are Google and Tesla. But both have recently been in the headlines regarding vehicle crashes.
The Google incidence was when its car thought a bus would give way to it and then didn’t and caused a collision. This is indicative of no matter how good the technology is its never really going to be able to calculate the endless vagaries of human nature, particularly when there is a human behind the wheel of an oncoming vehicle.
The Tesla crash was serious and ended in the death of the driver when the car’s sensors system failed to distinguish a large white truck and trailer crossing the highway. The car attempted to drive full speed under the trailer with catastrophic results.
Tesla’s vehicle autopilot system uses cameras and radar to detect and avoid obstacles. According to Tesla the cameras struggled with the glare from the sun and the radar decided that what looked like an overhead road sign meant there was no need to implement braking.
So here’s the question.
When you have multiple sensors providing conflicting information how does the vehicle decide which is the correct one to respond to? There is no doubt that indecision regarding the information from an autonomous cars sensing and imaging systems is a fundamental operating flaw.
So despite all the large sums of investment, the huge amounts of R&D work and all the real-world road testing of them it’s understandable that the average driver remains unconvinced.
Laser or infrared-based Lidar technology will help vehicle sensors pick up shapes and colours of obstacles which normal cameras are unable to detect, particularly at night. The principle behind Lidar is straightforward. Project a light at a surface and measure the time it takes to bounce back. Light travels at about 186,000 miles/sec and the technology that measures and interprets these needs to be equally fast.
Lidar basically creates 3D images of the surroundings and it works in light and dark conditions, clouds or sun, shadows or sunny conditions but it doesn’t like snow and rain. But then again nor do camera-based sensing systems.
But it is currently very expensive. Google, for instance used €80k worth of Lidar with its early autonomous car designs.
Part of the cost problem relates to there being no economy of scale production advantages to drive price down. High resolution Lidars are made in small quantities and can cost more than a car.
And Lidar, although very good, still needs to improve technically. Resolution and range are two such areas. Resolution-wise they achieve around 65 pixels and can see well for 75 metres.
There is also the operational factor that Lidar scanning happens as a vehicle is moving which means that images can be distorted because one section of the scene being scanned is captured at a different time to another part of the scene. And Lidar sensors capture a better image if they are positioned outside of the vehicle rather than having to operate through glass.
But despite these current technological and cost shortcomings here is no doubt that Lidar is going to be a must-have technology if autonomous cars are to become a reality on our roads.
But costs need to be driven down and there has to be some extremely sophisticated development work on how a car’s systems will decide which of its sensing system to believe in an emergency situation, especially given that making such a decision and then reacting has to be done in microseconds to be of any value.
Having said that there are many industry pundits who think that the concept of a totally autonomous driverless car is just to far-fetched.
In their view many of the safety improvements to vehicles that could reduce road casualties should come from improving conventional cars with examples being things like accurate lane control, collision avoidance systems and satellite navigation that will see hidden road accidents around bends and over hills and warn the driver. And it doesn’t have to be very sophisticated systems. Helping drivers to park safely and keeping them patient in stop go rush traffic would all contribute to road safety.
But even if driverless car technology does succeed there is still another major obstacle in its way and that’s the legal questions.
What happens when an autonomous car makes a mistake and kills a pedestrian? Who will be liable? The owner of the vehicle even if they were not driving, the manufacturer of the vehicle, the company that wrote the software or made the sensing hardware?
The legal issues themselves present a formidable red light for driverless vehicles.