UHarvest Route Planner Delhi NCR · B2B grocery dispatch
No warehouse —
Cluster Analysis

Plan a dispatch

3 steps
Warehouse & date
Upload files 0 of 2 required
Choose or drop orders CSV / XLSX
order_id, type, weight, lat, long, delivery slot
Choose or drop trucks CSV / XLSX
type, capacity_kg, fixed_cost, units_available, max_drops_per_trip, max_drops_per_day, is_nep
Drop limits may match; use 0 for unlimited.
Choose or drop history CSV
customer_id, lat, long, vehicle_type
Algorithm & constraints
Distance: OSRM (road network) ·
With second trips on, a trip that waits this long for a slot is cut in two so the truck can be reused.
1.0 = none, 1.1 = Delhi legs take 10% longer than the time-of-day profile.
Advanced settings
Fill & distance targets
Prefer keeping most routes at or below this (e.g. 64 = 80% of an 80 km cap).
Best-effort fill by weight or drop count. Under-filled trucks are topped up or downsized.
Sparse routes below this are merged or consolidated.
Slots, dwell & docking
Stop dwell times
A truck may reach a HoReCa outlet up to this many minutes before its slot opens. 0 = not before the slot.
Higher = solver works harder to keep GT on time (may use more trucks / km).
Warehouse docking
Service times
Added to every leg for parking and reaching the shop, on top of drive time.
Fleet search & zones
Cluster mode partitions today's orders by saved polygons (NEP-aware fleet filter).
Each iteration runs the full solver (~3–5 min). Try 180–300 s for 400+ stops.
Zone clustering & driver preference
5% cost focused40% sweet spot100% zone focused
City is divided into this many pie-slices from the depot. 8 = 45° slices.
0 = off. A route with a stop farther than this only carries stops near that stop.
Load balancing
PyVRP genetic search
One search decides trucks, routes and stop order. Travel times are static in the model; ROE's simulator re-times every plan, so the margins below are safety buffers.
× midday leg time (OSRM × traffic factor). 1.0 = as simulated.
Model closes HoReCa slots this early.
0 = GT must arrive before close. Capped by the GT lateness cap at validation.
Model shift = max shift − first load − this.
One truck = monthly cost × this, in metres (2.0: a 70k truck ≈ 140 km). Higher = fewer trucks before km.
With HoReCa-first on: extra cost of a truck that carries HoReCa, so HoReCa concentrates on few trucks.
3 ≈ 10.8 km per hour of route time; discourages waiting at shops.
Driver ease: km added per full sector crossing × driver-ease weight (20 × 0.6 = 12 km, about 9% of a truck, as in ROE). Raise it for tighter areas. It also counts against the km cap in the model, so the cap is conservative.
Km added to a leg between stops of different zones, so a route stays in one zone unless crossing pays. 0 = off.
Orders are clustered into zones of about this many stops. Stored cluster polygons are used instead when they exist.
A truck may carry up to this much over its kg capacity in exceptional cases, at the premium below. 0 = off.
Extra cost of an overloaded truck, so it is used only when it saves a truck.
Every stop of a route must lie within this many km of the route's anchor order. 0 = off.
Dispatch & execution rules
Upload orders and trucks CSVs to continue

Runs history

No plan yet

Build today's dispatch in three steps, or open a previous run from Runs history.

  1. Pick the warehouse and delivery date
  2. Upload the orders and trucks CSVs
  3. Choose an algorithm and press Optimize routes
Unassigned orders (0)
Google Maps mode is on — add your API key in Settings if you have not already.