Do I need real-time routing?

When considering optimising your company’s routes, the first question you should ask is ‘do I need real-time routing?’ Real-time routing arises when your set of planned routes has to change in response to new information coming in - for example, new or cancelled jobs, delays or a broken-down vehicle. As a result, unlike non-real-time (a.k.a. ‘static’) routing, you cannot plan a single unchanging set of routes for a day’s work and have your fleet follow them exactly through the day. Instead your routes need to evolve throughout the day, adapting hundreds or thousands of times as circumstances change.

Technically every route planning problem is real-time, as the world is always changing and a driver’s routes will never run exactly to schedule. Practically speaking though, some applications and industries are far more ‘real-time’ than others. The top ones we’ve come across so far are:

• On-demand deliveries - takeaway food, alcoholic drinks, laundry collection etc.

• Taxi services, both for general public and specialised services for the elderly and people with special needs.

• Field force optimisation, repair technicians, surveyors, engineers or any other mobile workers.

• Time slot generation for home delivery networks, e.g. offering efficient delivery time-slots to customers browsing an online supermarket, based on planned routes and already booked orders.

• Freight collections, in traditional freight networks (e.g. palletised goods), the same vehicles delivering items from a central depot will also make collections at short notice.

Developing an effective optimisation system for real-time problems is hard, requiring various technical and research hurdles to be overcome. Firstly, the planning engine must be ‘100% incremental’, meaning that when input data changes happen (e.g. a cancelled job), the engine can adapt the routes in a second or less, because it reuses and updates its existing calculations. For example, the engine incrementally adds to its matrix of travel times between locations, and re-uses its current planned routes as a starting point for calculating new routes.

Real-time vehicle routing problems also come with a host of extra features which are either not applicable or not relevant for static problems. The classic example is a ‘soft end time window’, which allows you to minimise how late a delivery is made or a location serviced. In a static setting, a single plan is made at the start of the day which assumes all stops will be served on-time (i.e. by their end time window). However, in a real-time setting minor delays can and will occur - for example due to unexpected traffic - and being late for a delivery often becomes unavoidable. As a result, a real-time optimiser needs to minimise lateness whereas as a static optimiser can just work with a single, inflexible hard end time window. Other examples of real-time only problem features are (a) ensuring deliveries already loaded on-board a vehicle are prioritised and (b) handling a human user’s manual overrides to a plan, adapting and optimising around them.

Sufficiently accurate travel time estimation is key to route optimisation - both for real-time and static problems. Accounting for rush hours can be tricky and integrating real-time traffic updates is harder still.

Another key issue with real-time route optimisation is how to properly model historic routes, so you can run ‘what if’ scenarios and estimate improvements to KPIs such as mileage from adopting a real-time optimiser. This can only be done with a discrete event simulation tool, which properly simulates the evolution of the planning process over a delivery period, re-planning whenever new jobs are added without assuming any future knowledge (i.e. no ‘crystal ball’ assumption).

Lastly any real-time system must be robust to hardware failure, making cloud-based hosting the only viable solution. However, a real-time optimiser is by its nature always running - making significant demands on server CPU usage, data transfer, memory etc. Scaling this up to run many models can only be done by making an intelligent and automated allocation of computing resources across different routing models.

ODL Live is a real-time optimisation engine developed by Open Door Logistics, which solves the many difficulties arising from real-time problems. ODL Live is designed to integrate via its webservice API with your existing drivers apps and backend logistics management system. Visit the ODL Live website for more information.

An integration is also planned with Intelligent Routing’s fully-featured dashboard and drivers app suite. If you are interested in using ODL Live for real-time routing with Intelligent Routing’s management dashboard and apps then please contact us using the Intercom button in the bottom right-hand corner of your screen or email us at [email protected]