The Ultimate Guide to Assessing On-Demand Transit Platforms


What do you think of when you hear “on-demand transit”? On-demand transit is a system where transit riders book pick-up and drop-off destinations and times, often using smartphones, and transit agency vehicles are directed, by on-demand software platforms, to transport these riders without following fixed routes. Rideshare services like Uber or Lyft may come to mind. But they are only a tiny representation of on-demand transit solutions.

This guide will answer the following questions:

  • What are the biggest problems experienced with on-demand transit
  • What are routing algorithms, and how do they affect the deployment of on-demand transit services?
  • What are the different routing algorithms and their applications?
  • What kind of routing algorithm is the most operationally efficient and convenient for the rider?
  • How can I tell which type of algorithm is in a transit system?
  • How are leading transit agencies using the most cutting-edge on-demand transit software to create the best routing solutions?
  • What questions should I ask potential on-demand transit software vendors?

Transit Transformation Using On-Demand Software

Often, public transit service planning results in a transit system design characterized by under-used fixed routes, typically found on evenings, weekends, and off-peak hours.

Unfortunately, it also produces “transit deserts,” where transportation services are limited, inefficient, and more costly to operate when compared to dense, well-traveled urban areas. It’s not feasible to provide a fixed route service in transit deserts like many rural areas and suburbs. Therefore, many people are outside the system, defeating the purpose of public transit, which is to provide equitable and accessible transit services to its citizens.

Today, on-demand software is an innovative tool for agencies to deploy to help provide efficient and more cost-effective transit solutions to transit deserts.

On-demand software manages all features of an on-demand transit system, including:

  • Routing Algorithms & Parameters
  • Service Configuration
  • Service Administration
  • Rider Management
  • Ride Management
  • Booking Methods
  • Dispatch
  • Driver Management
  • Vehicle Configuration
  • Scheduling
  • And many other functions

On-Demand Transit And The Traveling Salesperson Problem

There’s one problem that people talk about in the routing world but may not know the whole story – The Traveling Salesperson. The Traveling Salesperson has many places to visit during their day, including offices and coffee shops for meetings, conference centers for networking events, and back home after work. They may even travel to multiple cities in a day.

Nevertheless, the Traveling Salesperson doesn’t like wasting time – They want to travel from destination to destination in the shortest, most convenient, and cost-effective way possible. After all, they are busy!

On-demand transit software tries to solve this problem. But this problem becomes even more complex when accommodating multiple “salespeople” (or transit riders) who all have to reach essential destinations, including places of employment, education, community centers, retail, friends, and home. Providing an efficient solution to accommodate these trips is entirely dependent on the software powering the platform, which is computationally complex, requiring a lot of computing power, and must deliver a great customer experience for riders by eliminating transfers and reducing ride and wait times.

Simply put, the Traveling Salesman problem is challenging to solve, particularly for multiple salespeople (or riders). The ability to create a transit service with the most time- and cost-effective solution for traveling between different locations — Routing Optimization — is only possible with the most advanced technology and computing infrastructure.

Concerns With Common On-Demand Deployment Models

Generally, on-demand transit services roll out in phases. First, the agency selects a small area for a transit pilot and provides a small-scale service using one vehicle. Then, after the pilot, the agency analyzes the pilot’s performance and makes an informed decision on whether or not to expand to a larger area with more vehicles.

The idea is to get a feel for how an on-demand transit system works and how well it can handle the rider demand. Also, it is possible to fix any issues before the service expands further or cancel it altogether if the performance is poor.

However, there are problems with this approach!

For example, what if the service area grows too big for one vehicle to handle?

Most on-demand platforms use a different routing algorithm to handle multiple vehicles. As a result, one-vehicle service areas and multi-vehicle service areas have other characteristics, and as a result, generate different services and outcomes.

Also, it is difficult to use a small transit model like this typical one-vehicle phase to plan a long-term service.

Routing Algorithms Define the Rider Experience

Routing algorithms are the “secret sauce” in the software that determines how well an on-demand service performs. Routing algorithms assign requested trips to vehicles that cause no or very few violations, including lateness, long ride times, and transfers.

The routing algorithm affects:

  • Booking policy
  • Cancellation policy
  • Size of the service area
  • Number of riders serviced per hour
  • Number of transfers required
  • Number of vehicles on the road
  • Coordination with other services

Different routing algorithms create vastly different outcomes.

For example, some can handle one vehicle at a time, while others can manage many vehicles. Other, newer solutions are flexible and can increase or reduce the active fleet size in real-time — like Pantonium.

Some routing algorithms work best when scheduling trips in advance, and others are capable of ASAP requests and ad hoc boarding.

Three Major Routing Algorithm Properties

  1. Vehicles: How many vehicles can the routing algorithm handle in one area?
  2. Rides: How many rides can the algorithm handle at one time?
  3. Real-time: How well can the algorithm handle sudden changes at any time, including right now?

A Great Routing Algorithm Is Convenient And Efficient

The goal for the transit agency is to provide an optimal rider experience using existing or fewer assets. Cutting-edge routing algorithms create routes that accommodate many riders per hour with quick and reliable service.

The best algorithm:

  • Produces good routes continuously
  • Provides short wait times and trip times for the riders
  • Limits the number of transfers for riders (including when coordinating with existing fixed routes or rapid transit systems)
  • Uses a reasonable number of vehicles (to serve more than one rider at a time and improve operating efficiencies and route productivity)

Different Routing Algorithms And Their Applications

The following descriptions outline the various standard algorithms used in software platforms and the associated characteristics that can help you identify what performance goals line up with your on-demand project objectives and desired outcomes. Use this information to inform your software selection better and ask more informed questions.

One-Vehicle Routing Algorithms

Systems using one-vehicle algorithms are suitable if the service area is small. One vehicle is sufficient to meet the demand for service if the demand is under four or five trips per hour.

1. Do Nothing Algorithm, also known as Brute Force

A “Do Nothing” or “Brute Force” algorithm randomly assigns trips to the route or handles them on a first-come-first-served basis by adding new ride requests to the end of the queue.

Performance considerations include:

  • Trip processing (typically up to 5 trips per hour max)
  • Cancellations and no-shows are hard to control and worsen route quality and customer experience
  • Only accommodates ASAP trips

2. Best Position Algorithm

The “Best Position” algorithm receives each trip request and simulates all the positions possible to inject it into the route. It then chooses the best place to insert the pick-up and drop-off points.

Performance considerations include:

  • Handles the first few rides well, but the route quality and customer experience deteriorates quickly
  • Cancellations and no-shows are hard to manage and degrade route quality
  • Works best for short service or very few rides

3. Greedy Algorithm

A “Greedy Algorithm” chooses the closest stop to the current stop as the next stop. Although it works better than the Brute Force and Best Position options, it cannot handle requested times well. All trips are ASAP, so the rider cannot plan a journey, for instance, commuting home from work.

Concerns include:

  • Handles a few trips per hour (up to 5 trips per hour)
  • Hard to handle a specific request time
  • Forces all trips to be ASAP
  • Cancellations and no-shows are hard to handle

4. Traveling Salesperson Problem-Solver (TSP) Algorithm

A fast-running TSP is the best solution for routing with one vehicle. TSPs create the best quality routes because they constantly seek the “least worst” routing options. In addition, great TSPs are fast enough to handle real-time events, including ad hoc boarding.

Characteristics include:

  • An effective TSP solver can handle more than ten rides per hour
  • Needs to be real-time implementation, or it can’t handle cancellations well
  • Needs to run fast to accommodate more than a couple of rides

Multi-Vehicle Routing Algorithms

A multi-vehicle routing algorithm is necessary when:

  • The service area is large;
  • many vehicles are needed to meet the demand for service; and
  • the demand for service is over five trips per hour

1. Insert to Best Vehicle Algorithm

The “Insert to Best Vehicle” algorithm is a scaled version of the one-vehicle “Best Position” algorithm applied to multiple routes. The “Best Position” algorithm runs on every route. It then chooses the route that provides the best overall transit solution. Next, the software adds the requested trip to the vehicle in the best position to accommodate that particular trip and at the trip booking time.

Concerns include:

  • Route quality and customer experience deteriorates very quickly
  • Handles cancellations and no-shows poorly
  • Can’t move trips from vehicle to vehicle

This algorithm has the same limitation as the Best Position algorithm — adding rides degrades the route quality quickly. Therefore, the number of rides per hour that this algorithm can handle is not consistent.

2. Real-time Global Optimization Algorithm

The “Right Way” To Optimize A Route

The “Real-time Global Optimization” algorithm continuously optimizes the entire fleet and all the trips, dynamically and in real-time. It results in the most effective on-demand service convenient for the rider and efficient for the transit agency.

Benefits include:

  • Can move trips from vehicle to vehicle
  • Route quality and customer experience is always good
  • Handles cancellations and no-show very well
  • Service areas can be any size (as long as there are enough vehicles to service the area)
  • Vehicle capacity can be any size

A Real-time Global Routing Optimization algorithm can handle small or large service areas with many vehicles, riders, and trips. Riders can request trips and make changes, in advance and in real-time, including accommodating ad hoc boarding by riders hailing a vehicle.

The algorithm quickly adjusts to accommodate any changes and is minimally impacted by cancellations, no-shows, and detours. In addition, riders don’t need transfers because the algorithm allows vehicles to travel freely and optimally within a large service area, generating a great customer experience by eliminating frustration and time lost for riders.

Furthermore, a Real-time Global Routing Optimization algorithm is flexible in coordinating with fixed routes and rapid transit services. It works well in combination with scheduled stops. The optimal urban route configuration mixes fixed routes servicing consistent and dense demand, with all surrounding areas using real-time, on-demand services to address changing and uncertain needs.

A Real-time Global Routing Optimization algorithm allows vehicles to go to where the demand is in real-time.

If a Real-time Global Routing Optimization Algorithm is the best solution, why isn’t everyone using one?

In short, it is challenging to create an algorithm that runs fast enough to be both valuable and able to adjust to real-time changes quickly. This is a very computing resource-intensive software program to create!

Where other platforms fall short, Pantonium leads the way in powerful, cutting-edge Real-time Global Routing Optimization software that will help you create an efficient and convenient transit service, which consistently delivers a superior level of customer service.

How Leading Transit Agencies Are Using Pantonium To Create Efficient and Convenient Routing Solutions


Learn more about how Pantonium transforms public transport with on-demand transit.

Ask Vendors The Right Questions

Looking at the summary in the Table below, you can develop questions to better understand how the software platform is designed.

Here are a few questions that you can ask to ensure that you are getting a solution that will fit your needs.

  • Are there limitations on service area size?
  • What is the impact of allowing ad hoc boarding, where the driver creates the trip for the rider?
  • How many vehicles can be efficiently routed within a service area?
  • Can vehicles be shared between service areas?
  • How does your platform handle last-minute cancellations?
  • Can the rider book a pick-up time or a drop-off time? How is the drop-off time incorporated into the routing?
  • How far in advance can a rider book? Will this trip be slotted into a time window, or will the platform optimize it with other asap trips?
  • How does the algorithm coordinate with fixed route schedules? Does it plan the rider drop-off to align with fixed route schedules?


Get Started With Pantonium

Pantonium created and implemented a Real-time Global Routing Optimization algorithm for on-demand transit.

Pantonium is Global

  • Uses all the routes and trips to compute the solution in one service area
  • Handles many vehicles at the same time
  • Requires no zones or transfers

Pantonium is Real-time

  • Adjusts continuously to changes happening in real-time
  • Allows ad-hoc boarding
  • Automatically adjusts to cancellations, including no-shows

Pantonium is Flexible

  • Allows scheduled stops
  • Coordinates with fixed routes and rapid transit
  • Does not cannibalize productive fixed routes
  • Handles very long trips
  • Automatically handles co-mingling

Get in Touch

Additional Questions, With Answers From Pantonium


How do real-time changes affect the algorithm?

Consider a bus stuck in traffic and will be 15 minutes late to its next stop — Pantonium recognizes this and runs this new information through the global routing algorithm. The algorithm finds a unique solution — to send a different, nearby bus to the bus stop because there is the capacity for riders and will prevent them from waiting very long for service. This is a massive convenience for riders, who may be commuting to work, or standing in the cold or rain at the bus stop, or any other reason.

What are the conventional solutions to rolling out an on-demand transit service?

One conventional solution to rolling out an On-Demand Transit Service is to hire staff who manually handle requests, cancellations, and changes in advance (not in real-time).

Another solution is to “divide and conquer” by chopping the service area into smaller zones, each served by one vehicle. However, this is not convenient for riders and not efficient for the provider agency. Small, unconnected zones can significantly increase the need for transfers. Inconveniently, riders may wait a long time between transfers.

What are the limitations of fixing service areas with conventional on-demand transit solutions?

  • Vehicles can’t cross zone boundaries to move to where the current demand is
  • Not possible to adjust zones to different demand patterns across the day (like rush hour for commuters)
  • Cannot add more vehicles to a busy zone or remove vehicles from a quiet zone

Read More: Re-Designing Public Transit to Meet Today’s Needs