An AGV follows a path you install. An AMR decides its own path around whatever is in the aisle. That single difference explains most of the price gap, most of the integration work, and most of the disappointment when the wrong one is bought.
Both move material. Both can be leased or bought. In an Indian plant the choice usually comes down to how stable the route is, how much the aisle changes, and whether anyone on site can own the fleet software after the vendor leaves.
The difference that actually matters
An automated guided vehicle follows a defined route using magnetic tape, reflective markers, QR codes, wire guidance or a fixed laser target map. An autonomous mobile robot builds a map with lidar and cameras, localises itself against it, and plans around obstacles it did not expect.
| Question | AGV | AMR |
|---|---|---|
| How does it know where it is? | Fixed tape, markers or reflectors | Onboard map and sensor fusion |
| What happens when a pallet blocks the aisle? | Stops and waits for a person | Replans around it, or waits if there is no gap |
| What does a route change cost? | New tape, new markers, downtime | Drive the new route once, save the map |
| People and forklifts in the aisle | Needs segregated lanes or safe zones | Can share space with configured safety fields |
| Payload ceiling | Often higher, simpler frames | Lower to moderate on standard models |
| Price for the same payload | Usually lower upfront | Higher upfront, lower change cost |
The table is a starting point, not a verdict. A plant with a fixed 80-metre run between two conveyors is an AGV job. A warehouse where pallets are parked in different places every shift is an AMR job.
What an AGV is good at
AGVs win on predictable, high-volume movement with a stable layout. Tape-guided tugs, assembly line sequencers and fixed transfer cars have run for decades because the route never changes and the payload is heavy.
- Heavy payloads, including towed carts that would be awkward for a wheeled AMR.
- Very stable routes where the installation cost is paid once and amortised over years.
- Simple fleet logic: one route, one direction, fixed stops.
- Lower upfront price at the same payload, which matters when the budget is fixed.
- Predictable behaviour that a conservative safety review can accept quickly.
The hidden cost is change. Every new stop, aisle or product flow means new tape or markers, and that work usually happens during production hours. If your layout changes twice a year, the AGV’s cheap purchase becomes an expensive habit.
What an AMR is good at
AMRs win where the environment is shared and the flow changes. They also win where installing tape is impractical: cold stores, cleanrooms, leased buildings where you cannot alter the floor, and multi-tenant warehouses.
- Mixed traffic with people, forklifts and pallet trucks in the same aisle.
- Changing routes, new pick faces and seasonal layouts.
- Fast deployment: map the site, mark the zones, run the fleet.
- Scaling by adding robots rather than rebuilding the route.
- Better fit for existing buildings where floor work is disruptive or forbidden.
The hidden cost is discipline. AMRs need clear floor markings at pick and drop points, a map that is kept current, and someone who understands why a robot refuses to move when a mirror or a shrink-wrapped pallet confuses the lidar.
Cost comparison in rupees
Prices move with payload, brand, battery and fleet software. Treat the figures below as indicative bands for planning, not quotations, and ask for a landed cost that includes installation, mapping, training and the first year of support.
| Item | Typical AGV | Typical AMR |
|---|---|---|
| Vehicle, 500 kg class | ₹6–12 lakh | ₹10–20 lakh |
| Route or map setup | ₹40,000–1.5 lakh (tape and labour) | ₹50,000–1.5 lakh (mapping and tuning) |
| Charging station | ₹40,000–1 lakh | ₹60,000–1.5 lakh |
| Fleet software | Often basic or included | Licence, often annual |
| Layout change later | High, plus downtime | Low, often in-house |
Do not compare the vehicle price alone. The right number is cost per moved pallet over five years, including the labour hours your team spends keeping the system running. A cheaper AGV that needs a technician every quarter is not cheaper.
The infrastructure question
AGVs need the floor. Tape wears, markers get hit by pallet trucks, and reflective targets drift when racks are moved. Budget for a maintenance routine, not just an installation.
AMRs need the network. A fleet that loses Wi-Fi mid-task stops and waits. Before the pilot, walk the route with a laptop and measure signal strength at every pick and drop point, including the corners behind racking. If the site cannot hold a stable signal, fix that first; no robot vendor can code around a dead spot.
Fleet software and integration
One robot is a purchase. A fleet is a software project. Before signing, ask what happens when two robots want the same lift, how tasks are prioritised, and how the system talks to your WMS or ERP.
- Task assignment: first available, nearest, or fixed by zone?
- Traffic management: does it reserve intersections or rely on sensor avoidance?
- Integration: REST API, MQTT, or a database view? Who writes the connector?
- Reporting: can you export trips, wait time and utilisation without asking the vendor?
- Exit: can you export the map and the task history if you change vendors?
A pilot that proves the robot moves is not enough. Prove the software hands tasks correctly and that your own team can add a new pickup point without a support ticket.
A decision sequence that avoids rework
- Count the moves: how many trips per shift, at what payload, over what distance.
- Map the variability: does the pickup point move, and how often does the aisle change?
- Check the floor: condition, slope, gaps, door widths and lift access.
- Check the network: signal at every stop, and who owns the access points.
- Write the acceptance test: trips per hour, success rate, intervention rate and charging time.
- Only then compare quotes on landed cost and support response.
If variability is low and the route is long, price an AGV first. If variability is high or the building cannot be altered, price an AMR first. Most failed projects skipped this step and bought on demo-day impressions.
Pilot design and acceptance test
Run the pilot on the real route, with real loads, during real shifts. A weekend demo in an empty aisle tells you almost nothing about a Monday morning with forklifts, visitors and a wet floor near the loading dock.
- Two weeks minimum, including one week of production-level volume.
- Measure completed trips, failed trips and manual interventions per shift.
- Record charging time and whether the robot blocks a working aisle while charging.
- Test a deliberate obstruction: pallet, person, spill, and a dropped carton.
- Test network loss and power loss, then confirm the restart procedure.
- Ask two operators, not the vendor, to add a stop and change a route.
Define the pass mark before the pilot starts. A common one is 95 percent completed trips with fewer than one intervention per 50 trips, but the right numbers depend on your flow. Write them down and hold the vendor to them.
Common mistakes
- Buying an AMR for a route that never changes, then paying for autonomy you do not use.
- Buying an AGV for a warehouse that relayouts every season.
- Forgetting the charging location in the aisle design, so the robot parks in the busiest lane.
- Assuming the vendor will integrate with the WMS for free.
- Ignoring the safety review until after the purchase order.
- Running the pilot without the people who will operate the fleet.
Whichever technology you choose, the project succeeds on the boring parts: a stable network, a clean floor, a named owner and a written acceptance test. The robot is the easy part.
Can an AMR work in a facility with no Wi-Fi?
Only in a degraded mode. Most fleets need a stable network for task assignment and traffic management. Some robots continue a current task after a drop, but they will not accept new work reliably. Fix coverage before the pilot.
Do AGVs still make sense in 2026?
Yes. Fixed, heavy, high-volume routes remain cheaper with AGV technology, and the simpler safety case is easier to document. The mistake is using an AGV where the layout changes often.
How long does an AMR deployment take?
A single-robot pilot can be moving loads in one to three weeks. A multi-robot fleet with WMS integration usually takes two to four months, mostly software and change management rather than hardware.
What payload can a typical AMR carry?
Common models cover 100 kg to 1,500 kg, with heavier custom platforms above that. Payload, gradeability and battery life trade against each other, so check the continuous rating rather than the peak.
Do I need to change my floor for an AMR?
Usually not, but the floor must be reasonably flat and clean. Large gaps, steep ramps, oil films and loose gravel cause localisation faults. Mark pick and drop zones clearly and keep them clear.
The fleet decision usually follows the flow and the software layer, which the five-layer cost model explains.
Mobile robots move goods while ASRS stores them, and most large sites end up using both.
Once you choose AMRs, the next question is how they are coordinated; see fleet management.
Sources and further reading
- International Federation of Robotics — mobile robot statistics and definitions.
- A3 Robotics — AGV and AMR guidance from the industry association.
- ISO 3691-4 — safety requirements for driverless industrial trucks.
- Nav2 documentation — the open-source navigation stack used by many AMRs.
