Robotic Vehicles Work Best Off Public Roads: Mines, Ports, Farms
The driverless taxi gets the headlines, but most robotic vehicles doing daily work run in mines, container ports, warehouses and fields. Why closed sites came first, and what they teach about the open road.

Ask people to picture a robotic vehicle and most will think of a car without a driver gliding through a city. Such services exist in a small number of cities, and they attract most of the attention. Yet the robotic vehicles that do the most routine work today look very different. They are haul trucks in open-pit mines, container carriers in ports, tow tractors in warehouses and tractors in fields.
These machines share one feature. They operate where the owner controls the ground. That control turns a very hard problem into a manageable one. The vehicle does not need to understand every situation on earth, only the ones that can occur on a site that people designed for it.
Looking at where autonomy already works explains a lot about where it still struggles. It also gives operators a practical guide for deciding whether a robotic vehicle makes sense on their own site.
Why controlled sites came first
A self-driving system has to answer three questions all the time. Where am I? What is around me? What should I do next? On a public road, each question has countless variations. Construction zones appear overnight, cyclists swerve and police officers wave traffic through red lights.
On a private site, the operator can narrow those variations. Routes are mapped in detail and change only when the operator changes them. Speed limits are set to suit the vehicles. People who enter the area are trained, carry tracking tags or are kept out entirely. When something unusual happens, the vehicle can simply stop, and a supervisor nearby deals with it.
Stopping is the key point. On a highway, an abrupt stop can cause a crash. In a mine or a yard, a cautious stop is usually the safest default. That makes it far easier to build a safety case that regulators, insurers and workers accept.
Mines, ports and warehouses
Open-pit mines were among the first places to run large driverless trucks in daily operation. The trucks follow fixed haul roads between loading points and dumps, guided by satellite positioning, radar and lidar. A control room monitors the fleet. Because the trucks can run around the clock at steady speeds, they also tend to reduce tire wear and fuel use compared with uneven manual driving.
Container terminals followed a similar path. Automated guided vehicles and straddle carriers move boxes between ship cranes and stacking areas inside fenced zones. In warehouses and factories, mobile robots and tow tractors carry pallets and parts along aisles, often sharing space with forklifts and staff. Here the main sensors are cameras and laser scanners, with safety zones that slow the vehicle when a person comes close.
The business case in these settings is rarely just about cutting driver costs. Many sites struggle to find staff for night shifts, remote locations or dull repetitive routes. Others want fewer accidents in areas where heavy vehicles and people mix. Predictable cycle times also make planning easier for the rest of the operation.
Fields and orchards
Agriculture has used satellite guidance for many years. Plenty of tractors already steer themselves along rows while the farmer handles the implement. Fully driverless operation is a further step. Some machines now work fields with no one in the cab, watched by a farmer on a tablet.
Farms are private, but they are not tidy. Ground conditions change with the weather, animals cross fields and the boundary to a public road may be nothing more than a ditch. Driverless field machines therefore rely on geofences, obstacle detection and conservative behavior. They tend to start with simple, repetitive tasks such as tillage or mowing, where an occasional pause costs little.
Sidewalks: the halfway step
Small delivery robots that roll along sidewalks sit between the private site and the open road. They travel at walking pace, weigh little compared with a car and can stop almost instantly. Those traits limit the damage they can do. Several cities and university campuses have allowed them, usually under local rules.
They still meet the public, and that brings new problems. Pedestrians, wheelchairs, strollers and dogs share the same narrow space. Curbs, snow and parked scooters block routes. Many services keep remote operators who can guide a stuck robot. Public acceptance depends on small things, such as whether the robot gets out of the way on a crowded path.
Autonomy does not arrive everywhere at once. It spreads outward from places where someone controls the ground.
The hard step to public roads
Open public roads remove most of the advantages that private sites offer. Nobody controls who uses the road. Rare events become common at scale: a mattress on a highway, a child chasing a ball, a traffic light out of order. The vehicle must handle them at speed, and often it cannot simply stop.
Driverless taxi services and highway truck trials have shown that progress is possible within defined areas and conditions. Engineers call that envelope the operational design domain. It may limit a service to certain streets, hours or weather. Outside it, the vehicle must not drive itself. Even inside it, most operators keep remote staff who can give instructions when the vehicle is unsure.
Liability and rules remain unsettled in many places. Who is responsible after a crash, how incidents must be reported and how a vehicle proves it is safe are handled differently by country, and sometimes by city. These questions slow deployment as much as the technology itself.
What an operator should check first
For a company considering robotic vehicles on its own site, the experience of mines, ports and warehouses suggests a sober checklist:
- Can the routes be fixed and mapped, or do they change every week?
- How often do people and vehicles share the same space, and can that be reduced?
- Is there reliable wireless coverage across the site for monitoring and remote stops?
- Who will maintain the sensors, update the maps and respond when a vehicle stops?
- What happens during the transition, when driverless and manual vehicles work side by side?
- Do insurers, worker representatives and the local safety authority accept the safety concept?
Sites that answer these questions well often find that the vehicle is the simpler part. The larger work lies in changing the routines around it. That lesson from private ground will matter on public roads too, once the technology is ready to leave the fence behind.
Questions and answers
What is the difference between a robotic vehicle and a self-driving car?
A self-driving car is one type of robotic vehicle. The broader term also covers driverless trucks in mines, warehouse robots, farm machines and small delivery robots. What they share is that sensors and software handle steering, speed and braking without a person at the controls.
What do the SAE levels of driving automation mean?
The SAE scale runs from level 0, no automation, to level 5, full automation in all conditions. Most driver assistance in today's cars sits at levels 1 and 2, where the driver must keep watching. Level 4 means the vehicle drives itself without supervision, but only within a defined area or set of conditions.
Which rules apply to driverless vehicles on private sites?
On private ground, the main rules usually come from workplace safety and machinery law rather than traffic law. Operators typically need a documented risk assessment and safety concept. Details vary by country, so a site should involve its safety authority and insurer early.
Do robotic vehicles replace drivers completely?
On many sites, drivers move into new roles such as fleet supervision, maintenance or handling exceptions. Some jobs do disappear, especially on long repetitive routes. The transition tends to go better when staff are involved early and trained for the new tasks.
Why do delivery robots sometimes need human help?
Sidewalks are full of obstacles a robot may not know how to handle, such as construction barriers, crowds or snow. When a robot gets stuck, a remote operator can look through its cameras and guide it. This backup keeps the service running while the software improves.


