Agriculture Training Simulator

Training operators on the real machine was slow and costly, burning fuel and time. The venture moved that training into a simulator course for dealerships, farms, Agrishow and other farm events. My studio built the software and a partner company sold the physical simulators, from a simple tier with market controllers to a premium cabin made by hand. That split defined the design problem. New Holland and Case had tried other simulators and were unhappy with the look and the design, so the software had to teach the real job and also justify the hardware it ran on. The partner had the final word on the business, and the product direction was set by me and my tech partner, recognized by Unity for his vehicle physics. I led everything outside the code, from the experience and the art to the pitches, the roadmap and the nine people in business, marketing and assembly.

The fantasy at the heart of the simulators is simple. An operator who drives one of these machines every day should sit in the cabin and feel at home, with the same controls, sounds and view, on a farm they recognize. Familiarity comes before any instruction, because an operator who recognizes the place trusts the simulator and listens to what it teaches. So the world is the headquarters of a medium to large farm in the Brazilian countryside, with a dirt football pitch and the sheds and light poles operators see every day, under a warm light with seasons and day and night. Inside the cabin a radio plays music with ads where the brand talks about its machine, which turns a long session into a working day and gives the manufacturer a voice inside the experience. The farm also stays open for free drives, so a visitor can take the machine anywhere.

A real machine carries more systems than a new operator can absorb at once, and a fair adds needs of its own, a crowd watching and only a few minutes for each visitor. I introduced each system as a step of the course, at the moment the procedure needs it, so the operator first does a task by hand and then feels what the system takes over, as a real operator learns it, one system at a time. Cruise control holds the speed between manual driving and the GPS autopilot, cameras move between the cabin, the outside and a stand-by view, and the cabin display mirrors the machine's own monitor. For fairs, a vehicle swap cuts the walk between machines, and mirror screens fed by cameras on the machine and an outside screen for the public route each image to whoever needs it, the operator inside the cabin or the line waiting outside.

Learning a combine harvester means fifty-two steps in a strict order while driving a heavy machine at real speed. A checklist does not work in a moving cabin, so the course had to make the next action obvious and keep both hands on the controls. I set the order of every step with the engineers and operators who taught us the industry's standard procedure, and used five-hundred-page manuals, hours of training videos and my own time driving the machines to understand each part. Then I translated that procedure into quests on waypoints and highlight areas. Only the next area glows on the field, an arrow points to it from the machine, and the step completes on its own when the machine reaches it or the action is done. I wrote every quest text, tooltip and narrator line, and the same quests run on the keyboard, on both panels and in VR, with the same order and the same steps.

I approached the map from the procedure outward, laying out the farm around the places each training needs before placing a single house or field. The first rule was physics. A combine harvester cannot turn tight, so on fields of real, uneven sizes I placed every waypoint at the distance that lets each machine draw its turn on its own, which makes guided maneuvers feel instinctive. Then I shaped the space to guide the driver. Props and paths funnel the tractor around the crop and away from areas the course does not use, so the farm itself shows the way without a single sign. The built area is about a fifth of the planned map, which also had a town, a fuel station and a muddy area for freeing stuck carts. The result is a map that teaches without words, where every guided maneuver happens where the physics of the machine makes it feel natural, from the first row to the last turn.

The simulators were sold as hardware, and an operator judges a simulator first by how the seat, the wheel and the joystick feel in the hands, long before the training starts. So I designed the cabin around the physical experience of the real machine, defining the seat and its distance to the screen and the wheel, the height of the panels and controls, and the icons of the physical and digital panels. The real parts could not stay, so we read the machine's panel, wheel and joystick through reverse engineering and rebuilt them as our own, which let the partner produce a premium and a standard series without the original parts, with every button mapped to the software. The partner's team assembled the simulators and I coordinated that work, so every one came out with the right design, sizes and measures, and I set them up at the fairs, from the cabin to the VR headsets.

The real limit of the look was the hardware. The same farm had to look good on PC, in VR and on mobile, so the level of realism was set by what performance allowed. Pure realism also looked plain, so I defined a stylized realism that reads as real while the colors, the sky and the light are tuned to look more beautiful, light enough to run, rich enough to impress and composed to be pleasant to whoever drives. We moved between Unity's render pipelines more than once to get there. I directed the whole look and built most of it, from the lighting, cameras and post-processing to the materials, plants and crops, and rebuilt the machines from the manufacturers' engineering files. The harvest itself, with stubble, flattened plants and straw, was one of the most iterated systems of the whole venture, and every visible 3D model is mine except the machines themselves.