Skip to content

Industry News

A £15,000 Robot Is Now Walking UK Construction Sites Alone

Tilbury Douglas, one of the UK's top-tier construction contractors, has deployed a £15,000 Unitree G1 humanoid robot on a live build site, becoming the first major UK contractor to do so. The robot, named Douglas, autonomously walks the site capturing 360-degree imagery, LiDAR point clouds, and health and safety data, saving an estimated 40 hours of skilled worker time per month.

By Cara Voss · April 25, 2026

A £15,000 Robot Is Now Walking UK Construction Sites Alone

A £15,000 Unitree G1 humanoid robot named Douglas is now walking the floors of a live UK construction site, autonomously navigating stairs, using lifts, opening doors, and logging 360-degree imagery and LiDAR point clouds — all without a human escort. The deployment by Tilbury Douglas, one of Britain's top-tier contractors, marks the first time a major UK construction firm has put a commercially available humanoid robot to work on an active build.

The robot, whose name is a nod to the company, isn't swinging a hammer. Its job is data collection: capturing daily progress photos from identical positions, scanning for construction defects, monitoring health and safety compliance, and feeding results into three separate project management software systems. Before Douglas arrived, that routine took site staff roughly two hours per day on foot.

Modern UK construction site interior with laser scanning technology and digital data overlays AI-generated image

A live UK construction site now hosts a humanoid robot performing autonomous data collection. Source: Biped.News

Key Stats

£15K

Purchase Price

40 hrs

Saved Per Month

30 kg

Robot Weight

10 wks

Trial Duration

How Douglas Works on Site

The trial has been running for 10 weeks across different build phases — from open shell environments to near-complete interiors — giving Tilbury Douglas a broad picture of where the robot adds value and where its limits are. Mark Buckle, Technical Director at Tilbury Douglas, describes the setup as the next step in a digital strategy the firm has been developing since 2020.

Every day, Douglas walks the site and photographs each area from precisely the same position. That consistency matters: AI software compares the sequential images to automatically flag construction progress and deviations, a task that's almost impossible to replicate with ad-hoc human walkthroughs. The robot also carries laser scanning hardware that generates a LiDAR point cloud — a three-dimensional map of the space — which can detect installed components that deviate from design specifications.

Three software systems receive the data outputs: a programme management tool, a defect tracking platform, and a health and safety reporting system. The integration means site managers get automatic reports without manual data entry.

📷 360-Degree Imagery

Camera mounted on top of the unit captures identical-position photos each day. AI software compares shots to auto-detect progress and flag deviations from the build plan.

📡 LiDAR Point Cloud

Built-in laser scanner produces dense 3D maps of site geometry. Software identifies incorrectly installed components that cameras alone might miss.

🦺 Health and Safety Monitoring

Douglas conducts automated H&S walkarounds, checking site conditions and flagging issues directly into safety management software.

🚪 Autonomous Navigation

Climbs stairs, uses lifts, opens doors without assistance. High-vis markings and a safety beacon allow it to operate around plant machinery. Excluded from areas with fall risk.

For more complex environments, an optional computer backpack can be attached to give Douglas additional onboard processing power, increasing its autonomy in areas with complex layouts or heavy plant activity.

Under the Hood: The Unitree G1

Douglas is built on the Unitree G1, a commercially available humanoid robot from the Hangzhou-based manufacturer that has quietly become the most widely shipped humanoid platform in the world. The G1 stands 132 cm tall and weighs approximately 35 kg — compact enough to navigate standard UK building staircases and fit in most commercial lifts.

Advanced LiDAR sensor array and 360-degree camera technology used in construction site monitoring AI-generated image

LiDAR and 360-degree camera sensors enable precise site mapping and automated defect detection. Source: Biped.News

The base G1 has 23 degrees of freedom across its legs, arms, and waist. The EDU research variant extends this to up to 43 DOF with optional dexterous hands and additional wrist joints. Onboard compute runs an 8-core CPU, with the option to add an NVIDIA Jetson Orin module for heavier AI inference tasks. Battery capacity is 9,000 mAh with a stated runtime of approximately two hours, though Tilbury Douglas has not published specific runtime figures for construction site conditions.

The robot retails at approximately $13,500 USD through Unitree's commercial channels — well below the six-figure price tags of competing platforms from Figure AI or Boston Dynamics. Buckle's preference for calling it "less than a small car" is a deliberate attempt to reframe the ROI conversation for cost-sensitive construction firms used to tight margins.

Spec Unitree G1 Boston Dynamics Atlas Figure 03
Height 1.32 m ~1.5 m 1.70 m
Weight ~35 kg ~80 kg ~70 kg
DOF 23–43 (EDU) 28 65+
Battery Runtime ~2 hrs ~1–2 hrs ~4 hrs (work)
Base Price ~$13,500 ~$420,000 (est.) $50K–$100K+ (est.)
Commercial Status Available now Industrial pilots BMW / select partners
2026 Shipment Target 10,000–20,000 units Limited Production ramp

Technical Context

Sensors: Depth camera, 3D LiDAR, 4-microphone array, WiFi 6, Bluetooth 5.2 — all standard on the G1.

Compute: 8-core CPU base; optional NVIDIA Jetson Orin for heavier AI workloads.

Connectivity: WiFi 6 enables real-time data streaming to site management software.

Navigation: Stair climbing, lift use, door opening — all autonomous on current hardware.

Who's Behind the Deployment

Tilbury Douglas is a UK Tier One contractor with revenues exceeding £1 billion annually, operating across building, infrastructure, and fit-out sectors. The company's digital transformation strategy dates to 2020, and the Douglas deployment sits within a broader program of automation, digital twin adoption, and AI-assisted site management. Buckle frames the initiative as workforce amplification rather than replacement — an important distinction for a sector where the skills shortage is structural.

Unitree Robotics, founded in 2016 and headquartered in Hangzhou, China, has carved out its market position by making humanoid robots commercially available at price points competitors can't match. The company shipped 5,500 units in 2025 and is targeting 10,000 to 20,000 in 2026 — a volume trajectory that no other humanoid manufacturer is close to matching. Unitree has also used its own robots in its own factories, with G1 and H1 units assembling robot components as part of a self-production capability.

Unitree G1: approximately $13,500 retail, available now, 200-plus research institutions deployed globally.

Boston Dynamics Atlas: Electric platform, targeting industrial pilots with Hyundai, estimated at roughly $420,000.

Figure 03: BMW factory deployment, 67-hour autonomous run confirmed in Q1 2026, $50K–$100K+ estimated.

Tesla Optimus Gen 3: Internal Tesla deployment at Fremont, production ramp announced April 2026.

Agility Digit: Seven units in commercial operation at Toyota Manufacturing Canada as of March 2026.

What This Means for Construction

Construction site workers reviewing 3D digital schematics and AI-powered progress monitoring dashboards AI-generated image

Digital monitoring and AI-assisted reporting are reshaping how construction sites track progress and safety compliance. Source: Biped.News

Construction has been a laggard in robotics adoption compared to manufacturing and logistics. Warehouses and auto plants have structured layouts, predictable environments, and high task repetition — exactly what current robot AI handles well. Construction sites are almost the opposite: constantly changing geometry, unpredictable surfaces, weather exposure, and dense human activity. Most humanoid deployments in 2025 and early 2026 deliberately avoided the sector.

The Tilbury Douglas deployment sidesteps these challenges by keeping Douglas in a data collection role rather than a physical task role. It doesn't pick up materials, operate machinery, or work near the building envelope. That's a sensible scope limit that makes the deployment tractable with current hardware — and it's not a compromise. The data collection workflow is genuinely valuable and difficult to automate any other way.

The skills shortage context matters here. The UK construction sector faces a persistent gap: the Construction Industry Training Board has estimated the sector needs hundreds of thousands of additional workers by 2028 to meet infrastructure demand. Automating routine site monitoring frees experienced workers for tasks that actually require their expertise.

Key Insight

The £15,000 price point is the strategic story here. At that cost, construction firms don't need to model a multi-year ROI to justify the purchase. Buckle said the investment decision wasn't driven by a straightforward ROI calculation — it was about freeing skilled workers from 40 hours of monthly admin. That framing will resonate across an industry that has historically resisted automation.

2 hrs

Daily site walk replaced by robot

3

Software systems fed automatically

$8.78B

Humanoid market est. by 2035

The 12-Month Outlook

Tilbury Douglas plans to deploy Douglas from the earliest stages of future projects — including foundations and groundworks — so that progress records exist from day one. That's a significant expansion of scope from the current fit-out and shell-and-core deployments. The firm also intends to take the robot to schools and colleges as a recruitment tool, using it to signal that construction is becoming a technology-led career path.

Buckle said the company is exploring fleet deployments and partnerships to accelerate adoption. If 40 hours of monthly savings comes from one robot on one site, a contractor running 50 active sites simultaneously could be looking at meaningful aggregate productivity gains — without a single new hire.

Unitree, for its part, is targeting 10,000 to 20,000 shipments in 2026, up from 5,500 in 2025. A planned IPO on China's STAR Market would give the company capital to expand its international support infrastructure — a prerequisite for wider European construction deployments where on-site technical support matters. At around $13,500 a unit, the economics work at scale in a way that six-figure competitors simply cannot match for data collection tasks.

Frequently Asked Questions

Which Unitree robot model is "Douglas"?

Douglas is built on the Unitree G1 platform, a 132 cm, 35 kg bipedal humanoid available commercially at approximately $13,500 USD (£15,000 in the UK). The G1 is sold through Unitree's commercial channels globally and is distinct from the larger H1 (roughly $90,000) and the budget R1 (roughly $5,900).

Is this the first humanoid robot on a construction site?

It's the first deployment by a Tier One UK contractor on a live active build — which is the meaningful threshold here. Smaller robotics trials have happened on construction sites before, including spot-checking robots and drone surveys, but a commercially available humanoid conducting daily autonomous walkarounds on a live site with data feeding directly into project management software is a new development for the sector.

Can the robot do physical construction tasks?

Not in this deployment. Douglas is strictly a data collection platform — cameras, LiDAR, and reporting. It doesn't carry materials, operate tools, or interact with the build physically. Tilbury Douglas has specifically ruled out operating it in areas with fall risk. Physical construction tasks remain well beyond current humanoid capabilities in real-world unstructured environments.

What's the actual cost justification at £15,000?

The robot saves an estimated 40 hours of skilled worker time per month. At UK construction day rates, that's significant labour cost recovery. Tilbury Douglas frames it less as an ROI calculation and more as a productivity multiplier for an understaffed workforce. A fully loaded site worker at £300–£400 per day covering 40 saved hours per month equates to roughly £24,000–£32,000 in annual labour value, so payback on the hardware comes in under a year by that measure.

Will Unitree expand into more construction deployments?

Unitree is targeting 10,000 to 20,000 G1 shipments in 2026, and a planned IPO on China's STAR Market would provide capital for international expansion. The construction data collection use case can be replicated with low integration effort at any firm using standard project management software. Multiple UK contractors are expected to evaluate the Tilbury Douglas results before deciding whether to follow.

The Construction Door Is Open

The Tilbury Douglas deployment is not a proof of concept. It's a 10-week production trial on a live site, with quantified time savings, three integrated software outputs, and a clear expansion plan. The £15,000 price point means other Tier One contractors don't need a lengthy business case to replicate it.

Watch for two near-term signals: whether competing UK contractors announce similar programs in Q2 and Q3 2026, and whether Unitree's STAR Market IPO proceeds on schedule. If it does, Unitree will have the capital to build UK and European support infrastructure — currently the limiting factor for broader enterprise adoption outside China.

The Bottom Line: A £15,000 off-the-shelf robot just walked onto a British construction site and started doing the job of a full-time data analyst — and the only real barrier to every major contractor doing the same is whether they've seen this story yet.

The humanoid robot market has spent years demonstrating capability in controlled labs. Tilbury Douglas just demonstrated it on a muddy, active construction site. That's the gap the industry has been waiting to see closed.