Wastewater networks are hitting a tipping point. Aging assets, tighter budgets, climate-driven extremes, and higher public expectations are exposing the limits of reactive maintenance. The old cycle, wait for failure, respond under pressure, repair at premium cost, isn’t just expensive. It’s disruptive, difficult to prioritise, and often avoidable.
A better direction is emerging, treat inspection as high-fidelity measurement, not just documentation. Capture repeatable data, build usable models, integrate them into asset systems, and move from “what happened?” to “what’s likely to happen next?”
That’s the quiet revolution underway in sewer inspection, a shift towards subsurface intelligence powered by robotics, 3D sensing, and, increasingly, AI.
What “digital twins” actually mean underground
“Digital twin” can be a buzzword, so it’s worth being specific. In subsurface infrastructure, the value isn’t a pretty 3D model, it’s a measurable, searchable representation of the asset that supports engineering decisions.
A useful underground digital twin enables things like:
- Quantitative geometry, deformation, ovality, and cross-section change
- Accurate location of defects and features for targeted rehab
- Baseline comparisons over time, including change detection and progression tracking
- Capacity-relevant information, such as sediment volumes, debris build-up, and partial blockages
- Evidence-based prioritisation, risk plus consequence plus condition, rather than intuition alone
If inspection outputs can’t be measured, compared, or integrated, they often remain “inspection artefacts”, videos and reports that are read once, filed away, and rediscovered when something goes wrong.
Choosing the right platform: there isn’t one “best” tool
The biggest misconception in modern inspection is that one platform replaces the rest. In reality, the best programmes match the platform to the pipe environment and the decision you need to make.
Robotic crawlers: the workhorse, and still essential
Crawlers have been the backbone of sewer inspection for decades because they’re stable, repeatable, and suited to many low-flow conditions. With the right payload, they can produce consistent imagery and 3D data, especially where pipe conditions are relatively accessible and obstructions are limited.
Their trade-offs are familiar, tether management, access constraints, and challenges in higher-flow or heavily obstructed sections.
Aerial drones: fast reconnaissance where ground systems struggle
Drones can be extremely effective for rapid visual scouting, especially where you need quick situational awareness, tight-space navigation, or you’re operating above the flow.
But they have natural limits, mission duration and coverage per flight, payload constraints, and, crucially, water is a hard boundary for optical and laser sensing. Drones can map the air gap quickly, but they can’t see through water, which matters in many real-world sewer conditions.
Floating platforms: multi-sensor specialists for large assets and mixed conditions
Floating platforms come into their own in larger-diameter interceptors, culverts, and assets where you need both above-water geometry and below-water profiling.
This is where multi-sensor approaches matter. A combination of LiDAR, for dry geometry and structural detail above the waterline, and sonar profiling, for submerged debris, silt, and channel condition, can give a more complete picture of the asset’s true state, especially when sediment, debris, or high flow would compromise other methods.
The real breakthrough: complementarity, not competition
Aerial drones and floating platforms aren’t rivals, they’re complementary.
Drones are excellent at quickly mapping what’s visible in the dry zone. Floating systems can capture the dry zone and what lies below the waterline, plus can collect data over much longer timeframes and distances.
When you combine modalities, you close the gaps that would otherwise show up later as surprises, submerged build-ups, sediment accumulation that reduces hydraulic capacity, or debris fields that don’t appear clearly on video.
The practical takeaway is simple, use the fastest tool for reconnaissance, and the most complete tool for quantification. In many networks, that means multi-platform workflows where each tool is used where it is strongest.
Why safety and productivity are becoming inseparable
A key goal in modern inspection is reducing, or eliminating, the need for hazardous confined space entry. This is not just a safety aspiration; it’s a productivity shift.
Robots, drones, and floating platforms can operate in environments that include:
- Low visibility
- Difficult access
- Hazardous gases, for example hydrogen sulphide and methane
- Unstable locations with unknown obstructions
Keeping teams at ground level while still collecting engineering-grade data is one of the most meaningful improvements the sector is making, because it reduces risk and increases coverage.
Turning measurement into decisions: the missing middle
Collecting data is getting easier. Turning it into decisions is still the bottleneck.
Many organisations can now capture video, point clouds, LiDAR scans, and sonar profiles. But the value often stalls when the output isn’t aligned to GIS, standardised, comparable over time, or easy to interrogate by stakeholders beyond the survey team.
The next phase of inspection maturity looks like this:
- Capture repeatable, high-fidelity measurements
- Process into engineering-ready outputs that go through QA, registration, and referencing
- Integrate into cloud platforms plus GIS asset management tools
- Prioritise based on condition, consequence, and progression
- Act with targeted maintenance and rehab planning
When those steps join up, inspection becomes a continuous feedback loop rather than a periodic event.
Practical progress in remote 3D mapping: Telesto example
One practical example of this shift is how inspection payloads are evolving. At Headlight AI, we’ve developed Telesto a system designed for remote 3D mapping in sewers, tunnels, and culverts, with an emphasis on collecting usable measurement, rather than just “capturing footage.”
What’s interesting, and broadly relevant beyond any single product, is the retrofit mindset. Instead of requiring a single bespoke vehicle, the sensing package can be deployed across different platforms depending on the job, supporting consistent data outputs in varied environments.
More recently, combining LiDAR-based mapping with sonar profiling has been a useful step toward a genuinely holistic view, geometry above the waterline, plus quantification of submerged conditions like sediment depth and location (see example below).
Example section of a 3D point cloud showing above the waterline LiDAR data (from Telesto) fused with below the waterline sonar scans for accurate silt deposit locations, clearly showing the position of a major silt deposit in the sewer.
The broader lesson is the workflow it represents:
- Modular sensing
- Multi-platform deployment
- Outputs that engineers can measure, compare, and act on
AI in sewer inspection: promising, imperfect, yet inevitable
AI-assisted defect detection and reporting is advancing quickly. In some contexts, it can dramatically reduce review time and improve consistency, especially for high-volume programmes.
But the industry’s caution is understandable. Accuracy varies by asset type, capture quality, and training data. Many teams still need to validate outputs before trusting them for high-stakes decisions. The workflow can lose a lot of its benefit if AI outputs require extensive re-checking.
The near-term benefit isn’t “replace manual inspection.” It’s:
- Standardise coding
- Reduce reviewer workload
- Speed up first-pass assessments
- Focus human expertise where judgement matters most
As confidence grows, the value will compound, especially when AI is paired with high-quality, structured data, and not noisy, inconsistent capture.
Beyond inspection: building subsurface intelligence
The long-term trajectory is bigger than better inspections. It’s moving towards systems that can inspect, map, and ultimately intervene with increasing autonomy, especially in environments where access, safety, and cost make traditional approaches difficult.
That’s why research directions like the PIPEON Project are so interesting. Pipeon is exploring how robotic systems might not only support inspection and mapping, but also enable elements of repair and maintenance, bringing the possibility of faster response cycles and reduced operational costs.
From an asset-owner perspective, that matters because the value isn’t just knowing more, it’s closing the loop, detect issues earlier, quantify them reliably, prioritise with confidence, and execute interventions efficiently. If inspection is the “eyes,” projects like Pipeon are exploring what “hands” could look like in the same environment.
The practical question utilities should be asking now
Not “Do we have the latest gadget?” but:
Are we capturing data that’s measurable, repeatable, and integrated, fast enough to influence decisions before failures occur?
Because the shift from preventive to predictive maintenance doesn’t happen through intention alone. It happens when data quality is high enough to trust, workflows are integrated enough to act, and outputs are usable enough to move budgets and schedules.
A question for the industry
Where do you see the biggest bottleneck today, access and deployment, high-flow conditions and waterline limitations, turning raw scans into engineering-ready outputs, integration into GIS and asset systems, or confidence in AI-assisted reporting?
If you’re working on programmes, or research, that are pushing the field forward, especially around multi-platform sensing, digital twins, or autonomous intervention, I’d be genuinely interested to hear what’s working and what’s still painful.
