Is a Computer Enough When Generating Routes?

As a smart transport technology partner, our business involves pooling passengers in shared vehicles and generating the optimal routes to pick them up and drop them off. The route may be optimal from a computer’s point of view, but is it optimal from a human standpoint? As it turns out, there is a strong relationship between human psychology and the perceived quality of the ride, both from the passenger and driver’s point of view.

In theory, one would expect the quality of the ride or the trip to be judged based on the absolute trip distance or trip time, or the difference between requested and offered service times. However,  passengers in general have negative perceptions towards trips that are not smooth, trips with backtracking or loopy route patterns, even if they reach their destinations within the scheduled time windows. Here, we will show some examples of routes that are seen as unideal even though they fulfil all trip conditions, and how our interventions have “fixed” them.

‍

Backtracking routes

Below is an example of a backtracking route generated. To fix this, we would drop passenger 4 off first, then 3 and 2 in that order.

‍

Loopy routes

Here is an example of a loopy route that both passengers and drivers are not happy about. How do we fix this?

When we apply some manual considerations to change a few stops, the route is significantly smoother.

‍

Entering and exiting freeways multiple times

This route takes the passengers from the East end of Singapore all the way to the North of Singapore, with only one passenger living in the North (Woodlands). This trip is not ideal for the last passenger, being the only one living in that area and dropped off significantly later than the rest.

Also, this route requires the driver to exit and then re-enter the expressway several times, which makes for a very unpleasant driving experience for the driver.

One way to fix this is to divide areas into zones (below). By adding the zoning constraint, the previous route would not have been generated.

‍

Passengers prefer to be dropped off/picked up around the same area

Another insight is that people prefer the pick-ups of all the passengers to be around the beginning of the trip, and similarly, the passenger drop-offs to be all be towards the end of the trip, as opposed to having pickups and drop-offs throughout the entire trip. Even though the journey time may be the same, they perceive the former as more efficient.

This is another case of a route generated that would make either passenger A or B very unhappy if they both were put in the same bus service.

‍

While technology is computationally efficient in route generation, we still have to constantly input these manual considerations, to make the algorithm smart enough to offer routes that would maximise the passenger experience.

‍