The challenge
A food processing company set out to modernize and digitise its production and logistics processes. Daily delivery route planning was done manually by a small group of dispatchers who held irreplaceable know-how and were approaching retirement. The goal was to capture that knowledge in an algorithm before it left the company, speed up planning and optimize cost.
At a glance
ProblemsTSP + bin packing, solved jointly
TechnologyCustom C/C++, processor-optimized
Operating modelHuman-in-the-loop approval
What we did
SWICON solved two NP-hard problems jointly: the traveling salesman problem for routing and the bin-packing problem for truck loading. Off-the-shelf solvers were evaluated and failed on the client's specific side constraints, most notably the physical layout of the hanging hooks inside the trucks, which standard bin-packing implementations cannot express. The final algorithm was written in C/C++ and directly processor-optimized. The solution is semi-automated by design: the algorithm plans and optimizes, the dispatcher reviews and approves.
The result
AI-supported route planning resulted in 39% fewer trips compared to previous operations. Planning runtime dropped from around 6 hours to 30 minutes per day, and the dispatchers' knowledge is now encoded in the system.
