FIELD // KHUSHI SHAH

02 / Field Discovery · Mission-Critical Systems

Friend / Foe

Finding the real workflow inside
a high-stakes environment.

Through 30+ interviews, simulator observation, and live-exercise shadowing, we found that friendly-force information already existed — but lived outside the commander's sightline at the moment it mattered most.

The information already existed. The problem was that it was not available where the decision happened.

ATR / IFFM1 ABRAMSCOMPUTER VISIONH4D / DISF-C

H4D capstone with JMRC — Part of the Defense Innovation Summer Fellowship (DISF-C) Summer 2025.

Follow the workflow ↓

The Situation

Situational awareness inside tanks
fails more often than we admit.

Tank commanders on M1 Abrams may have ≤5 seconds to identify a target before engaging. Friendly-force data already existed through BFT, MFoCS, and JBC-P — but commanders still had to manually cross-check grid references while smoke, dust, weather, and motion overwhelmed their optics.

Every failed identification puts lives, tanks, and missions at risk. The problem was fratricide — friendly fire caused by hesitation, second-guessing, or misidentification under extreme time pressure.

The Combat Environment

≤5 seconds to ID a target — one mistake can cost lives

Smoke, dust, weather, and motion overwhelm optics

Manual grid checks split focus when precision is critical

Hesitation means the enemy gets the upper hand

01

At first, this looked like a software problem.
We later learned the information was already available — in the wrong place.

The initial assumption was that a cleaner interface would improve friendly-force awareness. Operators already had access to BFT, MFoCS, and JBC-P — the gap was not missing data, but the manual cross-checks required to use it under fire.

Initial belief

Better interface → faster identification.

If operators had a better screen, they would make faster identification decisions.

What fieldwork revealed

Place the right signal inside the existing operational display.

Commanders needed faster, clearer friend-or-foe identification embedded where they were already looking — not another system to cross-check.

What the decision workflow looked like

01Thermal / CITV view
02JBC-P cross-check
03BFT / MFoCS lookup
04Manual grid correlation
05Firing decision

The critical bottleneck was not information access. It was decision latency.

The Research Insight

Every additional screen increased cognitive load, instead of eliminating second-guessing.

Tank crews validated the need for identification that arrived at the point of decision — not information that required leaving the sight picture to find.

02

What field research surfaced

01

Commanders often ignored JBC-P during active engagement — there was not enough time to leave the thermal view.

02

Friendly-force data from BFT and MFoCS existed, but required manual grid checks that split focus under fire.

03

Operators relied on thermal imagery and visual confirmation; optics were routinely degraded by smoke, dust, and motion.

04

Crews described second-guessing identifications even when data was available — hesitation gave the enemy the advantage.

05

30+ interviews across JMRC and DCTC Poland validated demand for faster, clearer friend-or-foe at shooter level.

03

We went looking for the real workflow.

Through 30+ interviews with Staff Sergeants, Master Gunners, and Tank Commanders across JMRC Germany and DCTC Poland — plus simulator observation and live exercise shadowing — the team learned how identification decisions were actually made under time pressure.

Questions we asked

What commanders saw

What systems asked them to do

What people ignored under pressure

What changed our assumptions

01Problem brief
0230+ stakeholder interviews
03JMRC · DCTC Poland
04Simulator observation
05Hypothesis shift
04

Embed identification in the Commander's Display Unit —
not the operator's head.

Current Workflow

Thermal / CITV view

JBC-P cross-check

context switch

BFT / MFoCS grid lookup

context switch

Manual grid correlation

context switch

Firing decision

Proposed Workflow

Thermal / CITV view

ATR overlay — real-time friend-or-foe ID

integrated

ADP-01 friendly symbology

integrated

Firing decision

integrated

The proposed direction was not to replace BFT, MFoCS, or JBC-P. It was to use computer vision to surface friend-or-foe identification directly inside the Commander's Display Unit — eliminating manual cross-checks at the moment of decision.

05

Autonomous Target Recognition
embedded in the Commander's Display Unit.

Team Tank proposed an ATR computer-vision layer for Identification Friend-or-Foe (IFF) on ground maneuver elements — starting with M1 Abrams tanks. The model would track and classify armored vehicles in real time, embedded directly in the CDU with seamless connection to BFT, MFoCS, and JBC-P.

What This Layer Makes Possible

01

Real-time target ID — AI classifies armored vehicles instantly

02

Commander-first CDU integration — no manual cross-checks

03

Built to fit the fight — connects with BFT, MFoCS, and JBC-P

06

The concept evolved through research,
not feature accumulation.

STAGE 1

Real-time target ID

AI model tracks and classifies armored vehicles instantly from the thermal feed.

STAGE 2

CDU integration

Embedded directly in the Commander's Display Unit — no manual cross-checks required.

STAGE 3

ADP-01 symbology

Military-standard friendly-force symbols crews already recognize under pressure.

STAGE 4

BFT / MFoCS / JBC-P

Seamless connection with existing battlefield systems — built to fit the fight.

07

Where I contributed

01

Research leadership

Led research strategy and conducted 30+ stakeholder interviews with SSGs, Master Gunners, and Tank Commanders across JMRC Germany and DCTC Poland — synthesizing findings into product direction.

02

Field discovery

Planned site visits to JMRC and DCTC Poland, coordinated stakeholder meetings, identified operational roles to interview, and captured workflow observations during simulator demonstrations.

03

Product strategy

Defined the ATR/IFF scope for M1 Abrams, prioritized MVP assumptions, reframed beneficiary segments around shooter-level decision dominance, and developed value propositions.

04

Product development

Guided MVP iterations toward CDU-embedded computer vision, helped prioritize interface decisions including ADP-01 symbology, and connected technical feasibility to product recommendations.

05

Commercialization (DISF-C)

Continued as Team Tank through the Defense Innovation Summer Fellowship Commercialization Program — pursuing annotated datasets, simulation testing, field integration partners, and SBIR pathways.

What deployment would require next

The most important questions were still ahead.

These are not failures. They are the work required before the concept could become a real operational product.

01

Technical

Access to annotated CV training datasets

Real-time model latency on CDU hardware

Integration with BFT, MFoCS, and JBC-P

Secure battlefield data access

Certification and deployment architecture

02

Operational

Simulation environments for overlay + model testing

Vehicle-platform variation across armor fleets

Field testing with active tank crews

Training integration and maintenance ownership

03

Commercial

Acquisition pathway and pilot deployment

Procurement model for ground maneuver elements

SBIR and dual-use commercialization

Partners for system integration

04

Research

Validation across larger troop exercises

Cognitive-load and decision-time measurement

Decision-confidence improvement metrics

Insights from acquisition teams

A good first direction makes the next questions clearer.

The Takeaway

Eliminating fratricide one troop at a time — by turning hesitation into decision dominance at shooter level.

Seconds matter.

Placement matters.

Confidence eliminates second-guessing.

08

What I would validate next

I would validate whether the ATR overlay reduced identification time below the ≤5-second threshold, eliminated second-guessing during simulator and rotational training, and improved decision confidence for tank commanders under realistic conditions.

Testing Loop

AnnotateddatasetsSimulationtestingFieldexerciseModeliterate