From Basic to Advanced: adding real world complexity
We’ve discussed the basic routing optimizer and the load optimizer which are the most used optimization algorithms currently used in practice. When the size of the company or the complexity of the routing restrictions become more intense we go from basic to advanced algorithms. Examples are axle weight restrictions with compartment routing, Departure smoothing to balance the workload in the warehouse and a very intense one where the liquid collection routing takes axle and trailer weight into account when doing collections based and weight restrictions on roads and in different states.
Since the invention of the computer according to Moore’s law, which became reality, computers have increased in power eight thousand times. With the increase in research in Operations Research, the power of algorithms in the same time has increased by a factor 80.000. Advanced techniques such as in recent history column building, simulated annealing, where lots of solutions are tried and discarded, have become available.
And in very recent history, even random destruction methods which always improve the solution given time, these basically randomly select some collections and/or deliveries after the solution is finished and take them out of the solution. This is the random destruction part. Then the improvement comes when running the construction part of the algorithm again.
As mentioned before, basic algorithms already include: collection, trailer-drop, multi-trip, multi-depot and transfer scenarios. All kinds of real-life aspects like time windows, traffic congestion and compartments are also taken into account. What distinguishes basic from advanced is complexity and thus even more real-life restrictions. Some of these restrictions are very industry or truck specific. Others are country specific and some are just very difficult to put into an algorithm.
Currently with the latest advances in the science of Operations Research and the increasing complexity of real-life restrictions, especially due to the rise of mega-cities and increased qualification of the workforce, the following main business applications have become viable and/or necessary:
- Axle weight restrictions while loading pallets/cases, bulk and liquids
- Departure smoothing to reduce number of teams needed to load at any given time
- Door-to-door and middle of road returns for e-commerce with lots of close stops
- Dynamic Sourcing, where based on sourcing and distribution costs I dynamically during routing choose my source location
- Re-allocation of lower priority orders into earlier trucks if it reduces costs and has no impact on priority
- Routing of liquid collection taking road and state weight restrictions into account.
- Dynamic inventory where the routes are based on VMI orders.
- Real-time rerouting, meaning new orders/priority orders show up during the day due to customers’ urgency or trouble with already existing routes. Rerouting the already existing plan in real time to deal with these new orders.
Currently the big elephant in the room is the time dependent restrictions where many city centers are off limits for certain or all delivery vehicles during rush hours or during the day. Everybody is working hard to get a feasible solution for this problem (meaning a solution that actually gives a solution within 30 minutes for 50 trucks). I’ll have to post a specific update just on this topic when it comes out because it will change the way routing happens.
In the near future we’ll get the time dependent routing in. What the future will bring is uncertain. Already we’re seeing an advance time-slot allocation for supermarkets where customers can pick a timeslot without having completed their order. This will increase in complexity where we forecast potential orders based on which the whole supply chain is already optimized months in advance to guarantee lowest costs in prearranged contracts based on even longer forecasts of customer behavior and the algorithms will be attuned to the whole chain, not just to local optimum.
Operations research has come a long way. Many more things are possible now in terms of optimization and restrictions. Many more businesses are now being served and attended at their level. Advanced algorithms are a huge plus for the specific businesses where they are relevant at the moment. In the future, they will become the bread and butter of the entire company making it an indispensable competitive advantage.
The post From Basic to Advanced: adding real world complexity appeared first on This Complex World.