We have been creeping towards autonomous vehicles for years and now we appear to be on the verge of actually having them in consumer's hands. The momentum is now building at an alarming rate. The starting gun was really fired when the U.S. state of Nevada legally authorised the operation of Autonomous vehicles on state road back in 2011. Since then California, Florida, Michigan, North Dakota, Tennessee and Washington D.C. have also followed suit with legislation. This year, 2016, in Canada the province of Ontario has begun to allow testing of autonomous on Ontario roads. This is the standard US to Canadian exchange rate for tech adoption. Even the Institute of Electrical and Electronic Engineers (IEEE) have weighed in stating that by 2040 up to 75 per cent of all vehicles could have autonomous capabilities to some extent. There are some variations but most require a driver in the car currently and this, my friends, is what really cheeses me off! You see the driver as "the driver" must adhere to the intoxication laws and thus you still need a designated driver if you are on a bender with the lads. All my hopes for driving to the pub having a few pints and letting the car drive me safely home are smashed! So why bother you are asking as am I. Well actually there is a continuum put forward by the
National Highway Traffic Safety Administration (NHTSA), in the U.S.A that has 5 distinct levels. They are as follows:
- Level 0: Human driver is in complete control of all functions.
- Level 1: A single function is automated.
- Level 2: Multiple functions are automated concurrently but driver must remain attentive.
- Level 3: Driving functions are automated such that driver can safely engage in other activities.
- Level 4: The car can drive itself without a human driver.
Currently we are at level 2 legally speaking
although the technology is already well into level 3 and even level 4. So there is hope! why is it the lawyers slow everything down? So the day will come that our cars will be able to get us safely home form the pub all by themselves! So when pub trip utopia comes how will these predecessors of SkyNet get us safely home? That is, of course, if they believe we are worthy.
To do this let's have a look at the original Google autonomous car in a little more detail. The picture to the left shows the various sensors and of course what car is complete without an
Awesome Spoiler. The GPS give the car a general understanding of where it is. GPS is only accurate to a about 3m and can drop out if blocked by buildings or bridges or tunnels. Also it is a well know statistic that 500 people per year are killed by GPS and that 90% of statistics are made up. That is why the first and foremost element of the system is the laser range finder mounted on the car's roof.
(Velodyne LiDAR HDL-64E) This sub-system creates a detailed 3D model of the environment around the car. This is accurate to less than 2cm (See Picture 1) This model is then layered over high-resolution maps for the area the car is currently in. (See Picture 2) In Picture 2 you can easily see light blue laser data overlaying the high resolution map. This gives the car textured view of the world around it. There are actually 4 models created by this process. (See Picture 3) Moving left to right there is one that helps the car understand where it
could drive, one that provides an elevation or obstacle model, one that identifies the objects expected to be fixed and use that to identify objects that could move such as traffic, pedestrians, bicycles etc., and finally a model that tells the car where it
should or is suppose to drive. This helps the car react to changing conditions and how to make safe evasive manoeuvres when some
dolt walks out in front of it while texting. In addition, there is a forward facing camera mounted by the rear view mirror that does recognition on traffic controls, like traffic lights, to ensure the car
obeys them appropriately. The wheel encoder and inertial motion sensors (not shown) help the car very accurately locate and track its position on the detailed maps.
 |
| Picture 1 |
 |
| Picture 3 |
 |
| Picture 2 |
This is all good for the environment immediately around the car as it passes at low speed but how does it deal with high speed objects hurtling at it such as the condition on the highways. Since the laser array has an effective radius of 200 ft that gives the car 2.1 seconds to react to an in coming object travelling at 65mph if the car is standing still. If both objects are moving at 65mph it would be half that time. Even though the computer can easily calculate how to react it does not have the time to actually react in the real world. Obviously more time is required to react. Radar is added to give the car the ability to sense object in the distance beyond the reach of the laser. There are three Radar sensors covering the front and front/sides on the front bumper and another on the rear bumper to cover the rear. This give the car the ability to sense and perceive the world around it and is the first step to giving the machines the a picture of just how daft humans are.
So the car now can perceive the environment, can make the calculations and required adjustments to navigate through any road or street according to the rules of the road. Problem...most human drivers do not follow those rules explicitly. We go when it is not our turn, we yield when we shouldn't, drive too fast or worse yet too
slow in the fast lane with the blinker on! So the car needs to be ready to adjust based on interacting with other vehicles. So in the case when the car and another car arrive at an intersection at the same time the autonomous car will based on the rule of the road yield accordingly and then start to signal intent by advancing slightly. In essence the car becomes more "aggressive" till it can clear the intersection. No mention of a robotic arm option that could signal "disappointment" in others driving abilities.
The
Google Self-Driving Car Project has advanced well beyond the prototype out lined above and clearly well well into Level 4 of the continuum. So have no fear! Someday, in the not too distant future, all cars will come with "Autonomous Mode" or as I like to call it "Take me drunk I am home" mode.
To do this let's have a look at the original Google autonomous car in a little more detail. The picture to the left shows the various sensors and of course what car is complete without an Awesome Spoiler. The GPS give the car a general understanding of where it is. GPS is only accurate to a about 3m and can drop out if blocked by buildings or bridges or tunnels. Also it is a well know statistic that 500 people per year are killed by GPS and that 90% of statistics are made up. That is why the first and foremost element of the system is the laser range finder mounted on the car's roof. (Velodyne LiDAR HDL-64E) This sub-system creates a detailed 3D model of the environment around the car. This is accurate to less than 2cm (See Picture 1) This model is then layered over high-resolution maps for the area the car is currently in. (See Picture 2) In Picture 2 you can easily see light blue laser data overlaying the high resolution map. This gives the car textured view of the world around it. There are actually 4 models created by this process. (See Picture 3) Moving left to right there is one that helps the car understand where it could drive, one that provides an elevation or obstacle model, one that identifies the objects expected to be fixed and use that to identify objects that could move such as traffic, pedestrians, bicycles etc., and finally a model that tells the car where it should or is suppose to drive. This helps the car react to changing conditions and how to make safe evasive manoeuvres when some dolt walks out in front of it while texting. In addition, there is a forward facing camera mounted by the rear view mirror that does recognition on traffic controls, like traffic lights, to ensure the car obeys them appropriately. The wheel encoder and inertial motion sensors (not shown) help the car very accurately locate and track its position on the detailed maps. 
Currently we are at level 2 legally speaking 
