Arclane: Self-Hosted Fleet Routing and Dispatch Optimization

Arclane reads a JSON file of your stops and vehicles and prints back a plan: which vehicle takes which stops, in what order. Copy and paste as-is.

Set the image once:

IMG=709825985650.dkr.ecr.us-east-1.amazonaws.com/zece-holdings/arclane:1.0.0

1. CHECK IT RUNS

docker run --rm "$IMG" --version

Expect: arclane-cli 1.0.0

2. CREATE plan.json

One depot, two vans, four stops:

{ "depots": [{ "id": 0, "x": 0, "y": 0 }], "vehicles": [ { "id": 1, "depot": 0, "capacity": 100, "shift_start": 480, "shift_end": 1020 }, { "id": 2, "depot": 0, "capacity": 100, "shift_start": 480, "shift_end": 1020 } ], "stops": [ { "id": 101, "x": 10, "y": 4, "demand": 12, "tw_start": 540, "tw_end": 720, "service": 10, "label": "Baker St" }, { "id": 102, "x": 12, "y": 9, "demand": 8, "tw_start": 540, "tw_end": 720, "service": 10, "label": "Oak Ave" }, { "id": 103, "x": -6, "y": 11, "demand": 20, "tw_start": 600, "tw_end": 840, "service": 15, "label": "Mill Rd" }, { "id": 104, "x": -9, "y": 3, "demand": 15, "tw_start": 600, "tw_end": 840, "service": 10, "label": "King St" } ] }

x, y where it is, on any consistent grid tw_start, tw_end earliest and latest arrival, minutes after midnight (540 = 09:00, 720 = 12:00) service minutes spent at the stop demand, capacity space a stop takes and what a van holds; any unit, so long as it matches shift_start, shift_end when the driver clocks on and must be back

3. SOLVE IT

From the folder holding plan.json:

docker run --rm -v "$PWD:/data:ro" "$IMG" solve /data/plan.json

Expect: {"distance":52.3,"vehicles_used":1,"unassigned":0,"wall_ms":1}

One van covers it, 52.3 of travel, nothing left out.

4. GET THE ROUTES

docker run --rm -v "$PWD:/data:ro" "$IMG" snapshot /data/plan.json > solved.json

solved.json holds a "runs" list: one entry per vehicle, "stops" in driving order:

vehicle 1 -> 104, 103, 102, 101

Van 1 does King St, Mill Rd, Oak Ave, Baker St. That is the plan.

5. WHEN A STOP CANNOT BE SERVED

Arclane never drops a stop silently. Anything it cannot place appears under "unassigned" with the reason:

CAPACITY_EXCEEDED vehicles are full; add a vehicle or capacity WINDOW_INFEASIBLE nobody can arrive before tw_end; widen it SHIFT_OVERRUN it would push a driver past shift_end; extend it SKILL_MISSING no vehicle on shift has the required skill

Each code names the thing to change. Change it, run again.

To see it: keep only Van 1, set capacity to 30, solve again. The four stops need 55, so two come back as CAPACITY_EXCEEDED.

That is the whole product.

OPTIONS

Deadline add --time-limit-ms 5000 to solve; it returns in 5 seconds with the best plan found, whatever the size Skills vehicle "skills": ["reefer"], stop "skill": "reefer". The names are yours, they only have to match Depots add more to "depots", point each vehicle at one via "depot"

ALL COMMANDS

--version print the version --help print the command list solve <file> solve, print a one-line summary --time-limit-ms N stop after N milliseconds --patience N how hard to try (default 120) snapshot <file> solve, print the full plan as JSON gen --stops N --vehicles K make a test instance of any size bench benchmark on your own hardware

Exit codes: 0 fine, 1 bad input, 2 wrong command. Errors go to standard error in plain words, so a failed batch job tells you what to fix.

No server, no network calls. It starts, solves, prints and exits, which suits a scheduled ECS task or an EKS CronJob. One CPU per plan.

support@zecehold.com