How is reducing congestion...

...with the help of Dynamic Resource Management

back to our blogs

How is reducing congestion...

...with the help of Dynamic Resource Management

Solving the problem of traffic jams in our cities - once and for all

The problem

In any complex, modern city some things seem as if they will never change. During rush hour, you can never find a seat on the underground – and when it rains, the cabs simply disappear.  We have all read about smart cities – those urban hotspots-of the-future where all infrastructures work and a shiny, happy populace goes about its business at super-slick levels of efficiency. But right now, our cities seem as unsmart as humanly possible. Especially when you are waiting in the rain on a Thursday afternoon for that ever-elusive cab.  

Step forward Dynamic Resource Management; an AI-based system for the future which will liberate commuters from the misery of getting wet while waiting for a cab that is unlikely to turn up.  And in its wider applications, it will achieve much more: not least a massive effect on commercial revenues for those companies who choose to adopt it.

According to the Economist (Aug 12, 2017) “Electric propulsion, along with ride-hailing and self-driving technology, could mean that ownership is largely replaced by “transport as a service”, in which fleets of cars offer rides on demand… That could shrink the industry by as much as 90%. Lots of shared, self-driving electric cars would let cities replace car parks (up to 24% of the area in some places) with new housing, and let people commute from far away as they sleep—suburbanisation in reverse.”

So what is it, and how does it work?  I have been thinking about Dynamic Resource Management since I worked on navigation systems with Nokia in Berlin. The problem I was faced with then and which still plagues all current navigation systems - among them, Waze, Google Maps and so on - is that there was no effective way to dynamically allocate 100s of thousands of vehicles in a city to the road networks, so as to not cause the traffic bottlenecks. All current navigation systems send all drivers down the same path to avoid a traffic jam - thereby creating new traffic jams.

Dra 2 V1 4

The map of traffic in the city (far left) shows the areas of congestion, next to screen grabs of three different map apps. There is some traffic on route A, so all the apps are re-routing traffic via route B – result total chaos, since the jams are now even worse.

The cutting edge research teams at have harnessed their mathematical expertise and their knowledge of probabilistic modelling to produce a model for city taxi firms that will prevent cab drivers driving round empty and marry them up with the people who need a ride.  It is a complex problem – for drivers, cab companies and passengers - and right now is on the brink of providing a solution.

Demand for cabs is often cyclical – or as we put it dynamic. People tend to want them all at the same time, but it’s not always easy to predict when that time might be. The parameters of demand are always changing.

Potential cab users tend to be distributed all over a city. Somehow the taxi driver – and the controller back in their office - has to figure where the demand is likely to fall. Yet the demand for cabs between the hours of say 6am and 6pm is changing all the time – and is subject to a whole range of external parameters.

It is impossible for a driver to predict where they might be needed and at what time.  As a result, they are engaged in what mathematicians refer to as a “random walk” - in essence, they are driving around empty, looking for passengers.  They are burning off petrol and no one is making any money. The technical term might be random walk – but what we are really looking at is dead time. Time where the driver and the cab company are actually losing money. And the potential passengers, shivering while waiting for a cab in the rain aren’t happy either.

Dra 3 V1 2

In a cab, a driver can only see what is in the immediate vicinity, the traffic itself blocks the view - and the apps currently on the market can't always help. This image shows the limited area of perception around each taxi in a fleet whilst on a "random walk". Even though there are pickups to be made, the drivers are unaware of them – and revenue is disappearing.

The cab drivers’ main problem is that they have only a very limited vision of the demand for a cab. They can only look out of his windscreen and see a few yards in front.  He or she has no idea of the big picture – how many people want cabs in a different part of the city, how many people have just come out of a meeting, how many people are en route to a meeting.

They are hoping that past experience may give them an indication of where they might drive to but it is impossible for the human brain to compute all the possible eventualities.  If you ask a cab driver where on August 15 in the city it was raining and where in the city was the greatest unfulfilled demand for cabs, they wouldn’t have a clue.

Dra 4

This shows the chaotic, disconnected system that taxi firms are forced to operate around. Even the controllers are unaware of what’s going on – they are as much in the dark as the drivers.’s research and our platform takes into account the extraordinary range of variables that might influence the demand for a cab service in a city.  A lot of human activity is cyclical. If there is a regular meeting every Tuesday, in a faculty of engineering say, close to 20 people are going to come out. There might therefore be a demand for 20 cabs. But at the moment there is no system to tell the cabs to be there – no system to co-ordinate them at the level of a city.

You need a system that will tell you in advance of the meeting that 20 people are going to need cabs – you need to be pro-active in that… anticipate the demand before it is there. Taxi drivers don’t know, shifts are changing all the time and different drivers are out on the road.

You have different pockets of demand appearing and disappearing – should the controller in the office send 20 empty taxis, or do they have a drop-off model that shows that every week ten people get dropped off for another meeting.

The problem of allocating taxis is not simple. Different size of taxis, different sized groups of people – all of these are parameters that have a different cause and effect.

The solution’s research and work on Dynamic Resource Management offers a potential solution to the problem that cab drivers – and cab companies – face in the modern city. Trials in simulators have shown that 75 percent of taxis were able to arrive within half the time compared current state of the art benchmarks.

Dra 5 V1 2

Our unique model allows us to take account of large amounts of changing factors and put them all together. These variable parameters allow us to take decisions based on hundreds of different and dynamically changing factors. When we order a cab, we want it to arrive as soon as possible.

At, we are able to do much better forecasting and models than any other ride-hailing or taxi company, reducing the waiting time which is ultimately what the passengers actually want and need. These parameters might include the estimated time of arrival, the number of passengers, and the size of the vehicle required, whether it’s going to rain, if there are roadworks. Other variables might be associated with the cab driver – they might be in need of a rest break, a meal or a simple cup of coffee. All of these elements will be factored into the “model”.

The model

Dra 6 V1 2

This graphic shows what happens when’s agents are added to the system – they are able to provide each user - drivers, controllers - with the information to make the best decision.

The approach to Dynamic Resource Management will, once it is widely adopted, change the face of cab services in cities.  It will improve the utilisation rate of taxis, offering the cab companies a software platform effectively designed for their individual needs. The overall aim is to optimise the taxi and the driver – reduce the waiting time for passengers and thus make money. The cab companies will be buying into a subscription-based model. In a matter of months, we will build the front-end for a client.  When the system is in place, we will do the hand-holding and the system will take over, building a model that will keep taxis as well as passengers on the move.

The taxi model is likely to have a huge effect on transport in our cities. It is the first phase in a range of AI applications based on Dynamic Resource Management which will be rolling out within the next few months: other areas in which it will be applied include fleet management, financial services and last mile delivery. We can all look forward, it seems, to a well-ordered highly-dynamic future. And maybe one day we can leave our umbrellas at home.

Join us to make AI that will change the world

join our team