HomeQube

Problem essay · field evidence

Why construction robots still fail on real sites

Labs can show a robot working. Construction sites usually can’t keep it working. Here is what the field literature shows — and why we start with Machine Identity, not another isolated robot stack.

What the literature shows

Seventy-five deployments, five failure dimensions

A 2026 systematic review of 75 peer-reviewed field deployments (2016–2025) finds robots underperform on actual sites not because the idea is wrong, but because deployment fails across several dimensions at once ([Kim et al., 2026].

Three dimensions dominate:

  1. Environment (~60%)

    Weather, dust, lighting, clutter, GPS loss, uneven ground. Lab assumptions don’t survive the site.

  2. System integration (~59%)

    Hardware/software mismatch, BIM and digital-twin drift, vendor lock, fragmented workflows.

  3. Hardware (~52%)

    Battery, actuators, payload, fragile sensors, limited mobility in cluttered terrain.

Less common, but real: cost and scale (~20%), and human–robot interaction (~19%) — hard interfaces, constant supervision, safety near people. Failures co-occur. The most frequent pairs are environment + integration, environment + hardware, and integration + hardware. The authors argue mitigation must work at two levels: the robot itself, and the site — infrastructure, workflows, and workforce readiness.

Our reading

A join-the-network problem — not “buy a better robot”

That pattern is not solved by another proprietary stack. A machine that cannot hold a permanent address, a trusted heartbeat, and a place in shared coordination will keep failing the same way the literature describes — isolated, brittle, and hard to supervise.

HomeQube starts one layer earlier: Machine Identity → stamp → heartbeat → Alive on one coordination socket. Watching is free. Outcomes come later. Field evidence says adaptability, interoperability, and robustness are the gap; we treat those as reasons to give every machine a network identity before demanding full autonomy.

How Machine Identity works — Add, Stamp, Heartbeat →

What does Alive mean? — signal on the socket →

Put one machine on the wire

Start on the porch, or mint if you already know the path.

Reference

Citation

Kim, H., Hong, J., Song, E., Hong, T., & Choi, J.-K. (2026). A multi-dimensional failure factor affecting the on-site adoption of construction robots. Developments in the Built Environment, 25, 100841. https://doi.org/10.1016/j.dibe.2025.100841. Open access (CC BY 4.0).

This page is HomeQube’s problem framing and opinion. It does not reproduce the paper’s full dataset or claim endorsement by the authors.