Route optimisation software plans the route of each delivery rider or van: which stops it makes, in what order and at what time. It also assigns new orders to riders as they arrive during the day, and tracks every delivery until it is done. It is also called route planning or dispatch software.
The problem is harder than it looks. A rider with 30 stops could visit them in more possible orders than there are grains of sand on all the world's beaches. No computer can try them all. So the software uses methods that find a very good order quickly, without checking every possibility. It must also respect limits. Each van carries only a certain weight. Some customers want delivery between 6 and 8 p.m. A shop accepts deliveries only before 11 a.m. And each rider's shift ends at a set time.
The software needs good maps. It reads roads, one-way streets and traffic from digital maps and location data, and converts each address into a point on the map through location APIs. In India, addresses are often written loosely, such as "near the temple, behind the school". So the software also learns, from past deliveries, where each address really is.
Here is a typical case. A courier's delivery centre in Indore has 1,400 parcels to deliver one morning, and 40 riders. At 7 a.m., the software divides the city into 40 areas with about 35 parcels each, so that riders do not cross each other's paths. It orders each rider's stops to cut the distance travelled, and puts the parcels with an evening time slot at the end. Planning by hand used to take the centre's supervisor two hours. The software takes a few minutes. At noon, a seller books an urgent pickup in one area. The software adds it to the route of the rider who will pass closest, and reorders his remaining stops. Each rider's phone shows the next stop, and each customer receives a message with the rider's live location.
The same software runs in very different businesses. A quick commerce app uses it to send riders out of a dark store within minutes. A food delivery app uses it to give one rider two orders from nearby restaurants. A diagnostic lab uses it to send its technicians to collect blood samples from homes. The planned routes also feed delivery time prediction, which tells the customer when the order will arrive.
Newer versions make more of these decisions without a person, and change them as the roads and orders change, as described on the page on physical AI.