A doorway fills the screen. Beyond it, the interior is dark. Broken masonry covers the floor, daylight cuts through damaged walls, and another soldier moves across the camera’s view. There is plenty to look at. Understanding it takes longer.
For military VR training, that gap deserves attention.
The accompanying footage is attributed to Ukraine’s Omega unit during operations in a settlement near Kostiantynivka. It moves through damaged buildings and overgrown outdoor spaces, with views interrupted by walls, vegetation, other people and clouds of dust or smoke. The camera gives us fragments. It never gives us a complete picture.
Watching from a screen, we can pause, replay and study those fragments. The people in the footage had to act as events unfolded.
An edited recording cannot tell us everything they knew, heard or had been told. It cannot establish why each decision was made. But it gives training designers something useful to examine: how much work is involved in making sense of the surroundings.
A virtual building can have detailed brickwork, convincing shadows and carefully modelled weapons, yet still make that work too easy. If every room is legible, every relevant person is visible and every event arrives with an obvious cue, the trainee receives information that the setting should require them to interpret.
That matters because an exercise can reward a fast response without testing how the trainee reached the decision to act.
In this footage, damaged structures break up the view. Openings reveal only part of the space beyond them. Vegetation obscures the outdoor areas. The camera turns, and something that was visible a moment earlier disappears. A viewer may know that another person is nearby without being able to follow their position continuously.
Those are useful considerations for scenario design. Visibility, lighting, clutter and the movement of other people should affect what information is available to the trainee. Their inclusion needs to serve a defined training objective. Adding smoke to every room or filling every surface with debris would create visual difficulty, but difficulty alone does not make a good exercise.
There is also a physical limit to what VR can reproduce. A virtual pile of rubble can obstruct a view. It does not reproduce the footing, resistance or effort of crossing broken masonry. Training design needs to recognise that distinction and retain practical exercises for the demands that simulation cannot represent.
The same care belongs in the debrief.
After an exercise, an instructor can examine a replay with knowledge the trainee did not have. Positions are known. Events can be revisited. The outcome is already clear. It becomes easy to judge a decision using information that only became available afterwards.
A useful review should establish what the trainee could see at the time, what other information they received and what they understood from it. First-person playback can support that discussion, but the recorded view alone does not prove what someone noticed. The instructor still needs to ask.
This is directly relevant to our work at Agincourt. BattleVR supports immersive team training, repeatable exercises and after-action review. Those capabilities give instructors a way to examine decisions across a session, then return to the exercise with specific learning points.
Our design priority is to make the information available within an exercise part of the training itself. That means treating visibility, uncertainty and the behaviour of other people as training requirements, and assessing how well each scenario represents them.
The footage is a useful reference for that work. It shows how little a single view can explain, even when the image is detailed and the viewer has time to study it.
When reviewing a simulator, watch what information it gives the trainee. Then ask what they must work out for themselves.
Video attribution: Footage circulated as showing Ukraine’s Omega unit near Kostiantynivka. The visible watermark reads “Group Two”. Location and filming date have not been independently verified. Public post carrying the attribution.
