AlgoPlanner Math
06/08/2017

Dynamic Route Planning: The BIG Buzzword!

What do people mean when they say “Dynamic Route Planning”?

It is my experience that this expression has a large variety of meanings depending on who you ask. In fact is has so many meanings that I would argue that it has become meaningless in many cases. Here are some of the examples we have come across in discussions, tenders and other encounters with our customers and prospects.

A: “A pre-planning phase where routes are automatically re-planned by a computer as new orders are added/removed/changed”

B: “A planned route which is automatically re-planned by a computer as changes occur in the course of a working day – hence, including during the route execution.”

C: “A comprehensive planning of several Distribution Routes in a defined area by the application of more or less advanced heuristic, iterative algorithms”

D: “A route planning that takes road conditions and/or live traffic information into account, when calculating the route(s).”

E: “A vague idea/fantasy/wet dream – that somehow a computer is able to solve the problems with over- or under dimensioned capacity on a given fleet of vehicles and drivers. Sometimes even including vehicles of different sizes and scope, drivers with varying work hours etc… the wish list seems endless.

F: “Any mix of the above alternatives”

G: and…. “Probably many more interpretations prevail…”

As anyone with practical experience with distribution will have learned, the planning of distribution routes may often prove one of the most complicated and distressful jobs anyone can think of.

In my opinion, there are two main different ways of approaching the challenge of planning distribution routes. Different pieces of information enter and alter the process at different stages.

1. Entity View: You can see each order as an initial set of information – an entity – being progressively enriched with more and more information as the time passes; until completion.

The order is received, evaluated (where from/to, how much, special conditions), a time and a date is set for the execution, a driver and vehicle is added, a route is assigned, and then later on it is executed. Hereafter again, unforeseen events may require that further information be added; customer not at home, item damaged or missing etc. etc. The original information, however, is not forcibly altered in the process by pre-defined rules and therefore it is only sporadically changed.

OR

2. Process View: You can see the order as going through a process of checkpoints, each stage “leading the order in the right direction” – so to say. With the process view the order is still evaluated upon receipt, but this is where similarities end. The order should not be passed on to the next stage in the process until the data quality lives up to standard. Correctly spelled address, no duplicate barcodes, workable time window, etc. The next steps concerning assertion of physical location, condition of the goods followed by the placement on the best possible route etc. again are seen as checks on a “road with stoplights” where the order is not passed on to the next stage until the light is “green” from the preceding checkpoint… you get the idea.

Most distribution companies we have met tend to accept option 1; often without really thinking about it. Why should they, it works!

Yes it does, of course it works, otherwise procedures would have been changed. The reason why it works, however, is more often than not due to good staff and not good data or comprehensive procedures. The driver in particular is an often under-appreciated hero, and he often solves many problems that a computer is not likely to be able to. Anything from badly spelled addresses to volumes differing in reality from the information on the order, quick assessment of what is the more important/urgent choices when priorities have to be made, what to leave behind and what not etc. etc.

To sum up:

In order for any mathematics and computerization to be of any use:

  1. The data quality has to have a certain, and in fact quite high standard, and
  2. The prerequisites for the execution have to be predictable, and then
  3. You must determine when and where in the process it makes sense.

In theory, the Dynamic Route Planning is an excellent idea. It is our experience, however, that the challenges with data quality and predictability of the prerequisite conditions more often than not dictate a different or at least strongly adjusted use of mathematics.

It should be obvious that any ambitions to ask a computer to calculate anything for you require a serious scrutiny of routines and tools to standardize the underlying data. So, if you do have any ambitions to install Dynamic Route Planning in any form, we suggest a pragmatic approach. – We suggest that you view your business as a three stage process. Of course there are many more stages, but for simplicity I here categorize them into 3 simple or main stages, if you like.

Stage 1 Orders arrive (Entry):

Challenges: varying formats, varying reaction time and deadlines, varying data quality when it comes to addresses, dimensions, dates etc. etc.

Stage 2 Dispatcher’s desk (Massage):

At some point in time, the dispatcher enters the stage and tries to make the best of the situation. The situation may of course well be more or less frantic, and the dispatcher(s) may well be more or less experienced.

Challenges: Not all data has had the quality improved to “perfection”, there may be varying weather and road conditions, some drivers are late/sick, some vehicles are out of order, and replacement is not comparable in size and scope, some customers impose last minute changes to address, time window, etc. again…. The list is endless.

Stage 3 Pick Up/Delivery (Execution):

The Driver loads his vehicle. Also here – a number of things may go wrong. Goods missing, space in vehicle and size of goods do not match, mis-sorts, driver showing up late, bad weather etc. etc. this list is also almost endless. The driver leaves the hub and takes his route. Again…. Packages missing or delivered in the wrong place, address does not “exist”, goods damaged, changes in route sequence, recipient not home, refusal to accept delivery, no place to park the vehicle etc. etc. again an almost endless list.

Therfore… the best way to utilize mathematics and computing power for most distribution businesses is by analyzing your data over time – for instance 6 months. Based on the data from that period, you can use algorithms and computing power to suggest the best route areas for the forthcoming period of, say 1 or 2 months. Hereafter you can repeat the process after scrapping the oldest 1 to 2 months, and adding the newest. In this way you utilize the information from the most recent past to create a coarse automatic route optimization between the “Entry” and the “Massage”.

You may then use a different set of algorithms to optimize the “Execution” process, to get the best possible support during the day the goods are actually picked up and/or delivered.

We are here to assist you, and we have the knowledge and the software to help you optimize the process in your distribution enterprise.