Whenever I present Eloy’s work at conferences, including at last month’s Transport Technology Forum, I begin by introducing the Eloy Drive app. Available on Apple iOS and Google Android, it’s our most visible piece of software.
But truth be told, the app is really only the mechanism we needed to build to deliver many of the longer-term software solutions that Eloy is developing. Our main goal is to understand C-ITS: Coordinating Intelligent Transport Systems, where we aim to coordinate road vehicles with software. To be able to do that we need a method to communicate with drivers – hence the creation of the Eloy Drive app.
Coordination is a game changer
Connected, and eventually autonomous, vehicles have the great advantage of working together to optimise our road transport system. The major difference between human-driven and machine-driven is the higher ability to absorb and process information, and the willingness to comply with centralised rules. Autonomous coordination looks to optimise the entire system instead of having every vehicle trying to optimise for itself, which often leads to many of the problems we have on the roads.
Human-based coordination can still be emphatic because we can provide directions through a SatNav, where the SatNav has a back-door link with central road control. We can provide some message types to drivers via in-vehicle systems that can be acted upon. This alone can not only create a lot of benefit but also provide real-world learning for implementation in autonomous vehicles.
Connected coordination could be win the ultimate boss battle
There is also an overlap with human-in-the-loop, where we have autonomous and human-driven vehicles on the roads at the same time, possibly for decades.
This is what excites Eloy, and where the bigger pieces of software can be built with the Eloy Drive app as our main mechanism. We can iterate and test ideas, run pilots of different sizes, and learn what works in the world today for application in the world of the future.
What software is Eloy building?
The main rationale for Road Traffic as a Software is to help drivers (and in the future autonomous vehicles) make safer and more efficient decisions with data gathered from other vehicles.
Autonomous vehicle software will need to have high safety levels, and the additional connected data will be required to increase vehicle movement while maintaining these safety levels, in effect increasing the speed at which they can travel (within legal limits).
Greater efficiency will include the reduction in traffic, and faster and optimised journeys.
Here are the current big ideas we are working on:
Virtual traffic lights (stop and go protocols)
We have been a bit casual with our description as technically we should refer to these as Virtual Traffic Signals (VTLs). But for most drivers, Traffic Lights is universally understood.
The aim here is to think about how we can stop traffic anywhere on the road network to achieve a better traffic management outcome. This could be because as accident has occurred and we need to stop traffic from proceeding towards it, or if we want a priority system for vehicles such as on single track country lanes.
In the preliminary stage where we want VTLs to simply mirror how physical traffic signals behave there are 2 main parts to building the software.
We have some logic that determines the flow pattern of how we want to trigger the VTL, such as the timing loop between Stop and Go. We also have some geospatial information, including vehicle locations and intended journey route, that feeds into the VTL logic to determine optimal sequencing.
Temporary traffic signals could be replaced with virtual traffic lights
VTLs have various potential applications but one obvious autonomous opportunity is at roundabouts to improve flow. Human drivers often exhibit ‘polite’ or ‘aggressive’ behaviours at roundabouts that help keep traffic flowing or allow ‘stuck’ traffic to start flowing again. For autonomous vehicles arriving at a roundabout together, VTLs can signal which cars should go first, preventing a situation where they are all too ‘polite’ to move.
GLIDE (optimal speed)
Our road system is basically a binary system. Cars are either stationary or they should be travelling as close to the speed limit as is safely possible. The obvious question is, should vehicles travel at a different speed if it is safer and more efficient to be substantially below the speed limit.
Examples to consider are smoothing out stop/start congestion patterns, slowing earlier before needing to brake, slowing to allow other vehicles into lanes on motorways, or slowing on country lanes where visibility is poor, or the road surface quality is poor.
Autonomous vehicles can also benefit from GLIDE, such as coordinating adaptive cruise control, essentially creating platoons of vehicles with better air resistance patterns. Humans will fail to maintain these sorts of patterns due to human reaction times to slight changes in speeds, but autonomous vehicles can be more effective at picking up minor speed changes in other vehicles with V2V communications.
Coordinated and structured parking
Anyone who has travelled to a large venue by car, such as for a concert, a football match, or Christmas shopping, or even for the rush hour commute will know that it isn’t’ just roads that cause problems.
Getting in and out of car parks can be a nightmare and sometimes take longer than the rest of the journey. A significant proportion of traffic congestion is caused by drivers looking for parking spaces and 30% of all car accidents actually happen in car parks.
Improve car park packing to lower congestion on the wider road network
Coordination across parking will be an important area for autonomous vehicles because they will be able to park differently from human-driven cars. Passengers will be able to disembark at drop-off areas and the autonomous vehicles will enter a car park driverless. Cars will be able to pack more tightly and then allow specific cars out when their owners return to the collection area.
We can start to understand how this might work today with human-driven cars. The coordination in and out of car parks and which car park is most suitable is a coordination challenge that can save significant amounts of time and set the learning path for how road authorities will coordinate autonomous vehicles around venues. It is far easier to learn with semi-compliant humans than requesting access to take directional control of an autonomous vehicle from a private company.
Motorway lane risk patterns
When driving on motorways or multi-lane highways, we can sometimes get uncomfortable with the vehicles around us. Are they traveling too fast relative to us? Are they changing lanes aggressively? Do we make worse driving decisions in this situation, for example changing lanes that forces other vehicles to brake?
Connected vehicles can provide significantly better information on traffic flow and road accidents. Data collected on either side of incidents can be analysed and conclusions drawn. We already see this with variable motorway speed limits adjusting when traffic starts to move too fast, or the vehicle density increases to riskier levels.
We can start this work by looking at patterns of vehicles. Which lanes are they in and far away are they? Do drivers like to switch lanes frequently?
If this presents a higher risk threshold, we can either lower the variable speed limits or provide specific in-vehicle guidance to some drivers ie please move to the slower lane for a few miles, or please refrain from changing lanes. Small actions such as this might break up the risk patterns with far less intervention.
Optimal service station arrival and departure patterns
On longer journeys we often have several options about where to refuel, recharge, or take a rest stop. Many of us might have a favourite destination or certain places we would avoid at all costs. However, this might vary based upon how busy each service station is. Less busy stops might be more desirable than a very busy preferred location, particularly if electric vehicle wait times are high or there are long queues for food and drinks.
We don’t just need to provide information to drivers; we can coordinate them too. If we know driver and passenger preferences, we could adjust the SatNav to show the most suitable locations to stop so that service stations don’t’ get too crowded. If there is road congestion ahead, we might want more vehicles to take their rest to lower traffic density.
Again, this is an activity we can do now and will be highly valuable for autonomous vehicles.
Electric vehicle charging locations and timings
We briefly mentioned this in the service station section, but in the future the vast majority of cars will be connected and electric. Some but possibly not all, will be autonomous.
Whilst many people will charge their EV at home, many households won’t be able to, with up to 50% unable to have a home charger installed. As a result, if we want drivers to move to lower emission EVs, we will have a huge coordination challenge.
This will require an interesting mix of telling drivers where EV charging is along with algorithms to choose the most suitable locations that may want with information about the length of stay to optimise charging across all drivers.
