Driver less cars is all the talk in tech circles now, not because of any excitement of replacing the need for a driver – we all love driving don’t we, but because of the challenge and potential benefits it poses not only for the environment and reduction of required infrastructure investment but its exciting from a how do we do it fit for purpose clean sheet of skyward thinking. Its the type of imaginative solutioneering that is behind most Sci-Fi programs, I have an idea but no clue how to get there YET so lets get started.
That is how I will approach this too…… Lets GO…..
What do we start with, lets say a standard car of today which is fitted with the usual scary stuff like electronic controlled controls, i.e. brakes, throttle, Engine Management, Steering, 4G data connectivity (we will only use this for updates not control), Lights etc. I say this is Scary as I am a mechanical traditionalist and believe the driver should have a mechanical link to the road, not ask a computer “is that ok with you” and then be overridden because of a faulty sensor.
On Aircraft this is normally ok as you have replicated systems, heavily tested and manufactured to the highest levels with highly skilled maintenance crews with very high moral standards on rigid maintenance plans. When you buy your car do you think of quality and component life or how much Gizmo you get for your money or when you take your car for maintenance, do you think of the bill or how experienced and trained the mechanic is, so this is a big problem when we talk safety of authoritative control systems in cars, buses and transport systems.
Then we have the security, we all know that even the most expensive cars can be hacked by as certain skilled hackers have done it, so we have the problem today of someone applying the brakes or turning off the engine to steal the vehicle but this results in the future of road transport being prey to terrorism attacks while being hacked over the soon 5G wireless network.
Actually the worst disaster part of that can be cleverly avoided by a ROM state checking computer that prevents certain patterns of signal behavior but that’s another topic for another day and obviously incurs a cost.
So what is the biggest risk and hurdle to full automation transportation on the roads, well we just touched on it, a computer can be controlled by another computer and talk to it thousands of times a second but people are independent and random in comparison so its the transition phase. We are already transitioning in technology that’s built into our cars (and I will use the term cars now rather than road transport) and we know we can make robots move and clean your living room, if you have the money, even using LiDaR (Light Distance and Range).
LiDaR is currently incorporated in many driver less cars to monitor the surroundings but processing the received data can be costly on computing resource if it has to build a picture of the environment to determine what is moving, what is fixed, what is in the distance when most cheap LiDaR is accurate from 10m up to approx 200m. Yes there is LiDaR that’s higher range but its the cost and how much data that can be computed in on board devices. We can use Radio Wave detection techniques like RADAR or Ultrasonics for close up work of about 2m but high power radio waves are detrimental to all life and there are strict limits on electromagnetic radiation (electrically generated radio waves) allowed by devices (see ICNIRP website) not to mention the interference it would create with other communication devices so its not really an option at this time.
So now we have the negative out of the way, and we have a basic of the platform we are starting from, Lets Get Creative….
So for those that are following BIM (Built Information Modelling), you will be aware of how we can capture data from Laserscans to create virtual models. We can then process these models with software to pick out assets like valves, pipework, buildings, lights etc. We can do this by going one step further and Scan all our road network so we have a true virtual world model of not only the road network but all the infrastructure on and surrounding it. This will be a cloud hosted model and would take time to build so you may have one area able to support driverless cars and new areas added daily.
This Virtual World Model (from now VWM) would be GeoLocated to within cm accuracy which can be done, today, with the SPECTRA SP20 and scanners from the FARO range. Its not a big ask on capability but the cloud hosted data space would be huge.
This creates 3 problems,
- Who will pay for the scanning
- If the scanned Data is Huge then how do you access it in a rapid time for the travelling speed of the vehicle
- How could our wireless infrastructure support such data transfers.
Who will pay – Infrastructure partners
Huge Data – Store areas locally and reduce data from full scan data to GeoLocated 3D CAD files or models in opensource format on the vehicle. For those not familiar with 3D modelling yet, look at Blender, file size is small, coded in python and can be adapted to your needs. It also works with Armory3D which will also demonstrate how you can apply physics into 3D space. I can feel a tutorial coming on for that.
Does this seem far fetched – If you play any video game, it shouldn’t as they all now, in the main, effectively work in a virtual space to replicate real world physics. One of the things that made the Grand Theft Auto games so realistic and addictive.
Ok so we have this VWM but how does that create control in the real world – The key is GeoLocation. This feature is so much more powerful than having thousands of people walking around chasing little fluffy virtual creatures although it was great that developers got people physically active in their games rather than stuck in front of a computer screen. If only there was such a thing for Bloggers 🙂
So Geolocation of scan data works by picking a few points, called Ground control points (GCP), and measures your location from those points and equates it to a GeoLocation in the real world. This is not a processor intensive operation so is widely used in the BIM software world. It can do this as the GeoLocation in the real world is anchored as it is in the VWM.
But hold on, how can a distance from a point give you a Geolocation. In the early days of Mapping and still used today, the world was divided into Degrees Minutes and Seconds where 1 minute (on the Latitude) equated to 1.852km or 1 nautical mile, so if we know a distance from a known point in meters in the VWM then we can find it in the Real World. for this post, I will leave it there as that’s all we really need to know.
So in the VWM, we have a dumb model so far but now link in asset databases of what is now geolocated assets. Traffic control Lights – yes somewhere there will be a database of which red light is on, if its working etc even if its local to the asset, it can be connected via IOT. once you geolocate the asset you have geolocated the database and the control methodology of that intersection or road. You can also add a simple NO GO signal from the light locally much like the railway networks Auto-Brake(Dead Man) switch. Now apply that to smart road speed controls and speed sensors then we can create a fully autonomous road control network preventing traffic hotspots.
The vehicle will obviously need a GNSS aerial/receiver to accuracy of about 0.5m to 1m so high precission without RTK/VRS. A standard mobile phone type with uncorrected GPS/GLONASS will not be accurate enough at approx 4m at best. Why so accurate, you have to manage lane control. So we could use advanced lane departure systems which would have the computer equivalent of different coloured reflectors (cats eyes) so this would be great and answer which lane you are in and which side of the road could be coded in too but this is an infrastructure cost to maintain and living in England, looking at the Pot Holes and general embarrassing state of the roads, I am thinking reducing infrastructure maintenance requirements is the only way.
What about Pelican Crossings……. Yes they need to be replaced. As I said before, random actions are not good to computer systems and the data generated by push button operation of a traffic light controlled crossing will help locate crossings at better locations improving road design and safety overall so I imagine would be cost neutral in the BIG picture.
Ahh so what about someone walking in front of the car or falling in the road…..
We incorporate LiDaR on each vehicle that is autonomous capable which can compare not only what the road should be in the VWM and stop if there is a conflict but what is moving or calculate a way around the obstacle. The important thing is that it knows what Safe looks like in any location and environment and if it is not Safe then it stops.
When I spoke to someone about this, they said what if a dustbin blew into the road, you would stop a motorway, I replied, I would not have driven into a dustbin either….and what are passengers for.
Others have said, this technology is nothing new, but that is the point, we can do this today if we are just smarter about using what we have available now. We do not need a super AI to take over the world, we just need passion and imagination.
Oh yes I nearly forgot, the transition phase……
Ok so we have lots of vehicles that are semi autonomous but have someone behind the wheel. These can be upgraded with a new control module and sensor pack retrofitted option/package and those vehicles that cant be upgraded or too costly to do so would just not be allowed in. Once upgraded, they can pick up on a signal entering the zone and go into autonomous mode with a confirmatory reply signal sent to the checkpoint.
There is no problem with this approach, in fact its well suited. It would require an autonomous zone permit which would require the vehicle to be fitted with autonomous technology much the same way as London vehicle emmision zone rules work. That is simple. enforcing in real time would require numberplate recognition cameras as already fitted today. The only extra cost would be that of the police resource and their police overridden autonomous vehicle which would control the other traffic automatically, via the VWM with a wireless No Go Signal as back up in case of Data Network Failure, to safely create space and work area to pull over the vehicle.
We just hit on another benefit, imagine how safe our roads would be from a civilian casualty perspective oh and no get away cars in the first place. this could also be applied to Road workers and breakdown assistance personnel.
One of the other benefits is the pollution reduction and fuel savings that would be generated due to road network management. We could go on all day about the benefits but for me, my favorite part about driving is the freedom to go where i want when I want (cost of fuel permitting) so scenic areas of non-autonomous driving must be kept but the motorway cruise there or city driving could definitely give me more sleeping time so I can enjoy my time when I get there.
Another side thought is if the traffic is computer controlled effectively, then why not have conveyor road networks. Bear with…….
Lets say you have planned a route from London to Birmingham, would it not be cool to join a slot in a 140mph conveyor. This could be done by road design and electromagnetic coupling. Once your route planner on board your car has decided you are going to Birmingham, It can contact the M6 controller and gain a slot in the superfast lane. as directed by the M6 controller you could be directed down a sliproad, may need adaptions to your vehicle, and directly coupled to a conveyor or rolling chassis powered by electric (like a train car) thus only using fuel for shorter journeys. The chassis would be controlled by the M6 controller and moved as required. A conveyor system may have short sections of conveyor to accelerate your car to freewheel down train tracks and then re-accelerated again at specific sections. Anyone that remembers the Super Nintendo game F-ZERO will instantly associate with this.
So after now showing my age, I invite you to please leave any comments.