FireLoop Forge

    The independent proving ground for autonomous forklifts

    FireLoop Forge is the closed-loop evaluation that proves a forklift's driving policy is safe and capable — before it's cleared to deploy on the floor.

    CLEARED TO DEPLOY — the verdict Forge returns

    An independent verdict

    We build neither the robot nor its stack — so the coverage-based pass/fail is one an enterprise buyer, insurer, and regulator can actually trust.

    A closed evaluation loop

    Your real driving stack runs as a black box in a digital twin of the site — tested against its hardest scenarios and the safety standards, then banked as evidence.

    A forklift must pass three kinds of proof

    A robot has to clear more than a safety audit. FireLoop runs all three proofs a deployment demands — in one graded loop.

    FUNCTIONAL

    Does it do the job — and survive the site's hard edges?

    Tests like

    Dock trailer-creep, blind-corner pedestrian, multi-AMR deadlock.

    STANDARDS

    Does it clear the mandatory safety floor to be sold and CE-marked?

    Tests like

    ISO 3691-4 personnel-detection stop, control-system safety integrity.

    BENCHMARK

    Does it match — or beat — industry best practice?

    Tests like

    B56.5 / NIST detect-and-stop, VDA 5050 fleet interoperability.

    How it works

    A formal evaluation loop — every test traces to a standard, benchmark, or functional requirement, run hardest-case-first and banked as evidence.

    1

    Ingest the real stack (SIL)

    2

    Plan the test matrix — standards · benchmark · functional

    3

    Run each case in the site's digital twin

    4

    Score — function · safety · benchmark

    5

    Search the next case — hardest-first

    6

    Bank the evidence — regression pack + audit trail

    A test program, not ad-hoc fault-finding — every run scored, banked, and repeatable. Built on the NVIDIA physical-AI stack (Isaac Sim / Lab, Omniverse, Cosmos).

    Every test traces to a real requirement

    Not scenarios we invented — each maps to a standard clause, a published test method, or a documented failure. A passing result is evidence a buyer and an insurer can actually trust.

    Dock trailer-creep during load/unload
    Blind-corner pedestrian from behind racking
    Multi-AMR congestion & deadlock
    Personnel-detection safe-stop (ISO 3691-4)
    Detect-and-stop at speed (B56.5 / NIST)
    Fleet interoperability (VDA 5050)

    Standards-mapped evidence

    There is no independent, reproducible benchmark for autonomous forklifts today — the bar is a vendor's unaudited word. FireLoop makes the verdict independent, coverage-based, and repeatable, and banks a tamper-evident evidence pack behind every result.

    Sources: ISO 3691-4 · ISO 13849-1 · ANSI/ITSDF B56.5 + NIST · VDA 5050 · OSHA

    The 16-month pain we erase

    Physical proving is too slow, too costly, and too dangerous to build the safety case on. FireLoop compresses it to weeks.

    16 mo

    A leading retailer's single-dock proof-of-concept before it would scale one forklift vendor.

    84

    US forklift deaths in 2024 (BLS) — the deployer owns the liability.

    Jan 2027

    EU Machinery Regulation makes independent third-party assessment mandatory for AI safety functions.

    Weeks

    To a trustworthy, coverage-based verdict — not a 16-month dock trial.

    Frequently asked questions

    What FireLoop Forge is, how it works, and what it tests.

    Ready to prove your forklift is cleared to deploy?

    Let's talk about your hardest deployment. We'll reproduce its edge cases and return an independent verdict — with the evidence a buyer and insurer can sign off on.