1. Problem Statement
The United States operates approximately 2.3 million miles of buried drinking water transmission and distribution mains. A 2025 comprehensive study surveying over 800 water utilities — representing 17% of total US and Canadian pipe inventory — found that 33% of all water mains (approximately 770,000 miles) exceed 50 years of age, and nearly 20% (452,000 miles) have surpassed their engineered service life without replacement (Barfuss SL, Fugal M. “Water Main Break Rates in the United States and Canada.” Journal AWWA, 2025; 117(2):22–33. DOI: 10.1002/awwa.2401).
The consequences are measurable. US and Canadian water utilities report 260,000 water main breaks annually, costing $2.6 billion in direct emergency repair expenditures. The American Water Works Association (AWWA) State of the Water Industry Report has identified renewal and replacement of aging infrastructure as the top concern of water utility managers for five consecutive years. The Infrastructure Investment and Jobs Act of 2021 allocated $55 billion for water infrastructure, but the estimated replacement deficit stands at $452 billion — an order-of-magnitude gap between available funding and accumulated need.
Water utilities currently lose approximately 11% of treated water to distribution system leakage, a figure that rises to 30–50% in older municipal systems. At the national scale, non-revenue water losses represent billions of gallons of treated drinking water — and the energy, chemical treatment, and labor embedded in producing it — escaping through undetected pipe failures every day.
The fundamental obstacle to proactive infrastructure management is the absence of condition data. Water utilities cannot see inside their buried pipes without excavation or service interruption. External acoustic sensors detect active leaks but cannot assess pipe wall thickness, internal corrosion, joint condition, or incipient failure modes before a break occurs. Internal inspection using tethered closed-circuit television (CCTV) crawlers requires dewatering the pipe segment, isolating it from the distribution network, and dispatching a crew — a process costing $10–15 per linear foot that disrupts water service. These constraints limit internal inspection to fewer than 1% of distribution mains in a typical utility’s annual assessment program.
An autonomous, tetherless robot that enters live pressurized water mains through existing access points (fire hydrants, valve chambers), navigates pipe networks without service interruption, and returns multi-modal condition data (visual, acoustic, ultrasonic wall thickness) would transform water infrastructure management from reactive emergency response to predictive, condition-based maintenance. The technology exists at prototype scale. The commercial product does not.
2. State of the Art
Four independent research programs have demonstrated autonomous or semi-autonomous in-pipe robots for water distribution networks, each advancing different aspects of the core technical challenge.
Pipebots (University of Sheffield, 2019–2024)
The largest coordinated research program in this space, funded by the UK Engineering and Physical Sciences Research Council (EPSRC) at £7 million with £2 million in university co-investment, involving researchers from the universities of Sheffield, Leeds, Bristol, and Birmingham. Pipebots developed miniature robots (40 mm width) equipped with acoustic sensors, cameras, accelerometers, gyroscopes, magnetic field sensors, and ultrasonic transducers. The robots demonstrated autonomous navigation through pipe networks of 75–900 mm diameter using computer vision and inertial navigation without tethering. In 2024, the consortium was awarded an additional £9 million through Ofwat’s Water Breakthrough Challenge for deployment with UK water utilities, with Phase 2 completion scheduled for June 2026.
MIT Mechatronics Research Laboratory
Professor Kamal Youcef-Toumi’s group at MIT has developed a series of tetherless in-pipe robots for leak detection in pressurized water mains. Their ellipsoidal micro-AUV navigates 4-inch (100 mm) diameter pipes at 0.4 m/s with a turning radius of 1.5 cm. The leak detection principle exploits the pressure gradient near a leak point, translated into force measurements via instrumented silicone fins that contact the pipe wall. A critical contribution is the demonstration that autonomous navigation in live pressurized mains is feasible without dewatering — the robot uses the water flow itself as a propulsion assist while rim-driven propellers provide active steering.
SubMerge (Dutch Water Utilities Consortium, 2017–present)
A collaboration among Dutch drinking water companies Vitens, Evides, and Brabant Water, with technical development by Demcon and research support from KWR Water Research Institute. The SubMerge robot is a 2-meter-long, 19-module articulated platform with 15 individually driven wheels, operating in pipes of 90–300 mm diameter. The sensor suite includes cameras, ultrasonic wall-thickness measurement, hydrophones for leak detection (detecting leaks from 250 L/min), and a positioning algorithm for mapping inspected segments to utility GIS databases. The robot achieves speeds up to 360 m/h and travels up to 6 km on a single charge, covering approximately 66% of main distribution pipe diameters. The SubMerge prototype won the Aquatech Innovation Award in 2023.
Deep Learning for Pipe Defect Classification
Independent of the robotics platforms, computer vision researchers have developed increasingly accurate defect detection models for pipe inspection imagery. A 2025 study demonstrated a YOLOv11-based framework trained on 53,486 pipe images with 27,000 annotated defect instances, achieving precision of 0.90, recall of 0.80, and [email protected] of 0.90 for automated detection of cracks, corrosion, joint displacement, and root intrusion. A hybrid ResNet50-Swin Transformer approach published in Scientific Reports (2025) achieved 90.28% defect classification accuracy. These algorithms currently process imagery from tethered CCTV crawlers; their adaptation to autonomous robot platforms operating in turbid, pressurized environments is an open research problem.
The field is converging toward deployment. What remains absent is: (a) robust multi-sensor fusion algorithms that combine visual, acoustic, and ultrasonic data for comprehensive condition assessment in turbid flow conditions, (b) standardized miniaturized robot hardware that water utilities can procure and deploy using existing maintenance crews, and (c) integration with utility asset management and GIS systems for network-scale condition mapping.
3. Foundational Research
Barfuss SL, Fugal M. (2025). “Water Main Break Rates in the United States and Canada.” Journal AWWA, 117(2), 22–33. DOI: 10.1002/awwa.2401.
Conducted at the Utah Water Research Laboratory, this study surveyed over 800 water utilities representing 30.1% of the US and Canadian population and 17.1% of the estimated 2.3 million miles of installed water mains. The study found 260,000 annual water main breaks costing $2.6 billion in repairs, with 33% of all mains (770,000 miles) exceeding 50 years of age. The replacement deficit was estimated at $452 billion. This is the most comprehensive statistical characterization of US water infrastructure condition to date, and it quantifies the scale of the inspection problem: utilities cannot prioritize replacement spending without condition data for individual pipe segments.
Worley R, Yu Y, Horoshenkov KV, Anderson SR. (2024). “Acoustic Echo Sensing for Robot Localization in Buried Pipe Networks.” IEEE Sensors Journal, 24(16). DOI: 10.1109/JSEN.2024.3423040.
Developed within the Pipebots program at the University of Sheffield. This work addresses the critical challenge of localizing an autonomous robot inside a buried pipe network where GPS is unavailable and inertial navigation accumulates drift. The authors demonstrated that acoustic echoes — sounds generated by the robot and reflected by pipe features (joints, bends, tees) — can be processed to determine the robot’s position within the pipe network with sufficient accuracy for mapping inspection data to utility GIS coordinates. This is foundational because inspection data is only useful if it can be geolocated to specific pipe segments for maintenance scheduling.
Yu Y, Shi P, Krynkin A, Horoshenkov KV. (2024). “An Application of a Beamforming Technique, Linear Acoustic Array and Robot for Pipe Condition Localization.” Measurement, 238, 115361.
Also from the Pipebots program. This paper demonstrated that a linear acoustic array mounted on a miniature robot platform can localize pipe condition anomalies (cracks, joint degradation, wall thinning) along the pipe axis using beamforming techniques adapted for the cylindrical waveguide geometry of water-filled pipes. The acoustic method provides condition assessment data complementary to visual inspection, detecting subsurface wall degradation invisible to cameras — addressing the key limitation of camera-only pipe inspection systems.
Cha YJ, Choi W, Suh G, Mahmoudkhani S, Buyukozturk O. (2017). “Autonomous Structural Visual Inspection Using Region-Based Deep Learning for Detecting Multiple Damage Types.” Computer-Aided Civil and Infrastructure Engineering, 33(9), 731–747. DOI: 10.1111/mice.12334.
A foundational study from MIT demonstrating that region-based convolutional neural networks (Faster R-CNN) can simultaneously detect and classify multiple structural damage types — concrete cracks, steel corrosion, bolt corrosion, and delamination — from inspection imagery with mean average precision (mAP) exceeding 0.87. While developed for above-ground structural inspection, this work established the viability of multi-class defect detection from robotic platforms, and its architecture has been adapted by subsequent in-pipe inspection research.
Chatzigeorgiou D, Youcef-Toumi K, Ben-Mansour R. (2015). “Design of a Novel In-Pipe Reliable Leak Detector.” IEEE/ASME Transactions on Mechatronics, 20(2), 824–833. DOI: 10.1109/TMECH.2014.2308145.
From MIT’s Mechatronics Research Laboratory. This paper presented the design and validation of a tetherless in-pipe leak detection robot that exploits the pressure differential near leak points in pressurized water mains. The robot detected leaks using instrumented flexible fins that measure suction forces against the pipe wall, achieving detection at any circumferential position. Critically, the system operates in live pressurized mains without requiring dewatering or service interruption — the fundamental operational requirement for utility-scale deployment. Field testing with Saudi Arabian water utilities validated detection of leaks as small as 0.5 GPM in 4-inch and 6-inch PVC and ductile iron mains.
4. Competitive Landscape
No company currently sells an autonomous, tetherless in-pipe inspection robot for live pressurized drinking water mains.
Motmot Inc. (Detroit, MI) is the closest entity to commercialization. Founded to develop the MotBot autonomous underwater robot for pressurized water main inspection, Motmot received a $1.555 million NSF SBIR Fast Track grant (the PIPELINE project). The robot enters live mains through fire hydrants, navigates autonomously, and uses cameras and acoustic sensors for inspection. Commercial pilot programs were announced for Summer 2025. However, as of early 2026, MotBot remains pre-commercial with no publicly reported utility deployments or revenue.
SubMerge (Netherlands) has demonstrated the most mechanically sophisticated prototype — a 19-module articulated platform — but operates as a consortium of Dutch water utilities developing technology for their own networks, not as a product company selling to external customers. SubMerge has not announced plans for international commercialization or US market entry.
Existing pipe inspection companies — RedZone Robotics (Pittsburgh), CUES Inc. (Orlando), and Envirosight (Randolph, NJ) — sell tethered CCTV crawler systems for gravity sewer inspection. These systems require dewatering and do not operate in pressurized drinking water mains. Their product architectures — tethered, requiring service interruption, designed for gravity flow — are fundamentally incompatible with autonomous operation in pressurized systems.
The competitive gap is structural, not temporal. Tethered sewer inspection companies cannot enter the pressurized water main market by extending their existing products; the physics of pressurized flow, the requirement for tetherless operation, and the need for multi-sensor fusion in turbid water demand a different system architecture.
5. Total Addressable Market
Bottom-Up Calculation
The American Water Works Association reports approximately 50,000 community water systems in the United States, of which roughly 4,200 serve populations exceeding 10,000 — the utilities large enough to justify robotic inspection technology. These 4,200 utilities operate an estimated 1.5 million miles of distribution mains.
Conservative deployment assumption: 4,200 utilities inspecting 2% of their pipe network annually (the minimum recommended by AWWA guidelines) at $6.50/linear foot average:
- Annual pipe length inspected: 1,500,000 miles × 2% = 30,000 miles = 158.4 million linear feet
- Annual inspection service TAM: 158.4M ft × $6.50/ft = $1.03B/year
- Robot hardware TAM: 4,200 utilities × 2 robots × $150,000/robot = $1.26B (capital)
- Annual software/data platform: 4,200 utilities × $80,000/year = $336M/year recurring
- Annual recurring TAM (services + software): ~$1.37B/year
- Global multiplier (2.5× US): Annual recurring ~$3.4B/year
Top-Down Cross-Check
The global in-pipe inspection robot market was valued at $2.57 billion in 2024 and is projected to reach $6.96 billion by 2031 at 15.3% CAGR (Reanin, 2024). The pipe inspection robot market broadly was valued at $3.87 billion in 2025, projected to reach $15.3 billion by 2035 (Future Market Insights, 2025). Both cross-checks are consistent with the bottom-up estimate.
Revenue Model
This is a non-medical infrastructure market. Revenue flows through: (a) robot hardware sales or lease-to-own programs, (b) inspection-as-a-service contracts (per-foot pricing), (c) annual software licensing for AI defect classification, (d) integration services with utility GIS/asset management systems (IBM Maximo, Cityworks, Innovyze), and (e) federal and state infrastructure grants (IIJA, EPA DWSRF) that fund utility adoption.
6. Research Gap & HHA Contribution
The research prototypes described above validate that autonomous in-pipe navigation in drinking water mains is physically achievable. Three specific integration gaps prevent commercial deployment at utility scale:
Gap 1: Multi-Sensor Fusion for Comprehensive Condition Assessment
Each published system relies on a single primary sensing modality — Pipebots on acoustics, MIT on pressure-based leak detection, SubMerge on cameras with supplementary ultrasonics. No published system fuses visual, acoustic, and ultrasonic data streams in real time to produce a unified condition assessment of pipe wall thickness, internal corrosion, joint integrity, and leak status simultaneously.
HHA contribution (Haedar Hadi): Design and implement the sensor fusion architecture — a multi-modal deep learning pipeline combining convolutional neural networks for visual defect classification (YOLOv8/v11 backbone), signal processing models for acoustic echo localization, and ultrasonic waveform analysis for wall-thickness estimation. The key intellectual challenge is learning joint representations across heterogeneous data streams with different sampling rates, noise characteristics, and spatial resolutions. Architecture: multi-task neural network with shared encoder and modality-specific decoder heads, trained with uncertainty-weighted loss balancing. Evaluation: benchmark against single-modality baselines on paired multi-sensor datasets from pipe test facilities.
Gap 2: Miniaturized Multi-Module Robot Manufacturing at Utility Procurement Volumes
SubMerge’s 19-module articulated platform and Pipebots’ 40 mm crawlers are hand-assembled research prototypes. Scaling production from single units to hundreds or thousands requires Design for Manufacturability (DFM) analysis of every module — propulsion, sensing, power, communication, articulation joints — to identify materials, tolerances, and assembly processes compatible with serial production.
HHA contribution (Ahmed): Embed DFM constraints into the robot design process from the earliest prototype iteration. This means: injection-molded housings replacing custom-machined aluminum, standardized waterproof connectors (IP68) replacing epoxy-sealed research assemblies, automated optical inspection for PCB quality, and tolerance stack-up analysis across the articulated joint chain to ensure consistent actuation under pressure loading. Ahmed’s specific deliverable is a manufacturing readiness assessment for each robot module — identifying bill-of-materials cost at 100-unit, 500-unit, and 2,000-unit volumes, lead times for specialized components (miniature rim-driven propellers, pressure-rated camera modules, piezoelectric acoustic transducers), and assembly sequence optimization for production line throughput. This capability is absent from every research group in the field — academic labs do not employ manufacturing engineers, and the adjacent sewer-crawler companies manufacture tethered systems with fundamentally different form factors.
Gap 3: Utility Data Integration and Fleet Management
Inspection data is valuable only when integrated with utility asset management systems (IBM Maximo, Cityworks, Innovyze) and GIS databases that track pipe material, installation date, soil conditions, and service history. No published system includes this integration.
HHA contribution (Hass Dhia): Design the data integration architecture that maps robot inspection output — geolocated condition scores per pipe segment — to utility asset management workflows. This requires: a standardized inspection data schema compatible with ESRI ArcGIS and utility GIS platforms, RESTful API endpoints for real-time data ingestion from deployed robot fleets, automated report generation matching AWWA condition assessment standards, and a fleet management dashboard for multi-robot deployment scheduling across pipe network zones. Hass’s experimental design background from biomedical research maps directly to designing the field validation protocol: sample size calculations for pipe inspection campaigns, statistical methods for comparing autonomous robot assessments against ground-truth excavation data, and systematic bias detection across pipe materials (ductile iron, PVC, cast iron, asbestos-cement).
Why Originating Labs Have Not Closed These Gaps
Academic research groups (Pipebots, MIT, SubMerge) have not closed these gaps because their funding comes from research grants with publication mandates, not commercialization timelines. The PI publishes a demonstration paper and moves to the next research question. Manufacturing optimization, data platform engineering, and utility integration are not publishable research — they are product development activities that fall outside the academic incentive structure. Existing pipe inspection companies (RedZone, CUES, Envirosight) cannot close these gaps because their product architecture is fundamentally incompatible with pressurized autonomous operation — entering this market would require developing entirely new platforms while maintaining their profitable tethered-crawler businesses.
7. Comparable Funded Projects
| PI / Institution | Funder / Program | Amount | Year |
|---|---|---|---|
| Kirill Horoshenkov, University of Sheffield | EPSRC Programme Grant + Ofwat Water Breakthrough Challenge | £18M (~$23M total) | 2019–2026 |
| Motmot Inc. (Detroit, MI) | NSF SBIR Fast Track — PIPELINE Project | $1.555M | 2024 |
| Nuno Martins, University of Maryland | NSF Cyber-Physical Systems / National Robotics Initiative | ~$850K | 2019–2022 |
| US DOT / FHWA | Culvert Autonomous Inspection Robotic System (CAIS) | Not disclosed | 2023–present |
| Carnegie Mellon University | ARPA-E — Confined Space Mapping Module | Not disclosed | Active |
These awards demonstrate sustained, growing investment across US (NSF, DOT, ARPA-E) and international (EPSRC, Ofwat) funding agencies. The funding trend is moving from basic research grants toward deployment-focused infrastructure programs — a signal that the technology is approaching commercial readiness and that funders see water infrastructure inspection as a national capability gap.
8. Opportunity Assessment
TRL Evidence Chain
TRL 4. Multiple independent systems have demonstrated autonomous or semi-autonomous navigation in pipe environments that simulate or replicate real water distribution conditions. SubMerge has operated in live pipe segments within utility networks; MIT has validated leak detection in field tests with Saudi Arabian water utilities; Pipebots has demonstrated acoustic sensing in buried pipe test facilities. The step to TRL 5 requires integrated demonstration of autonomous navigation plus multi-sensor condition assessment in an operational water utility network over multi-day campaigns — which Pipebots’ Ofwat-funded Phase 2 is scheduled to deliver by June 2026.
Technical Risks
Risk 1: Navigation Reliability in Pipe Network Junctions
ModerateDescription: Pipe networks contain tees, crosses, valves, and service connections that present navigation decision points. Current prototypes have demonstrated navigation in straight segments and simple bends; junction navigation requires robust localization and path planning.
Mitigation: Acoustic echo localization (Worley et al. 2024) provides junction detection capability; reinforcement learning-based navigation policies trained in simulation and transferred to hardware (sim-to-real transfer) can handle the discrete decision space of pipe network topology. Architecture: proximal policy optimization (PPO) with a state space defined by acoustic echo patterns, IMU data, and pressure readings. Go/no-go at M6: successful autonomous navigation through 10 consecutive junctions without human intervention in a pipe test facility.
Risk 2: Sensor Performance in Turbid Water
ModerateDescription: Drinking water mains contain sediment, biofilm, and variable turbidity that degrade camera image quality.
Mitigation: Multi-modal sensing reduces dependence on visual data alone. Acoustic wall-thickness measurement and leak detection are unaffected by turbidity. For visual inspection, adaptive illumination (structured light projection) and image enhancement algorithms (dehazing CNNs) can partially compensate. Go/no-go at M9: defect detection accuracy (mAP) exceeds 0.75 at turbidity levels up to 10 NTU.
Risk 3: Robot Retrieval and Reliability
HighDescription: A robot that becomes stuck inside a live water main creates a service disruption — the opposite of the system’s value proposition.
Mitigation: Passive buoyancy design ensures the robot is carried to the nearest downstream access point by water flow if propulsion fails. Modular architecture allows individual module replacement rather than whole-robot disposal. MTBF analysis and accelerated life testing applied to every mechanical subsystem. Go/no-go at M12: MTBF exceeds 200 hours for all propulsion and articulation subsystems under simulated field conditions.
Regulatory Context
In-pipe inspection robots for water infrastructure are not medical devices and do not require FDA clearance. Relevant frameworks: NSF/ANSI 61 certification for materials in contact with drinking water (all wetted robot surfaces must be certified), AWWA standards for utility equipment procurement, OSHA confined space entry regulations (29 CFR 1910.146), and state public utility commission regulations governing water service interruptions. NSF/ANSI 61 certification functions as a competitive moat: the materials testing and certification process takes 6–12 months per material, creating a durable barrier for fast-followers.
Algorithm Architecture
The defect classification algorithm would be locked after training for initial deployment — a YOLOv8/v11 model trained on annotated pipe inspection datasets, validated against utility ground truth, and deployed as a fixed inference model on the robot’s edge compute hardware. Over-the-air model updates would follow a staged validation process: new model trained on expanded dataset, validated against a hold-out test set, A/B tested against the production model on non-critical inspection runs, and deployed only after statistical equivalence or superiority is demonstrated.
9. Team Fit
Hass Dhia
MS Biomedical Sciences, medical school background (anatomy TA), AI infrastructure architect. Hass brings experimental design methodology from biomedical research — where randomized controlled trials and systematic protocol development are standard practice — directly transferable to designing field validation protocols for autonomous pipe inspection. His physical sciences breadth (physics, fluid dynamics, thermodynamics) maps to understanding the physics of pressurized pipe flow, acoustic wave propagation in cylindrical waveguides, and sensor behavior in turbid environments. His AI infrastructure experience (multi-agent orchestration systems managing parallel autonomous processes) maps directly to fleet management of multiple deployed inspection robots: scheduling inspections across pipe network zones, aggregating multi-robot data streams, and integrating condition assessments with utility asset management platforms.
Haedar Hadi
MS Computer Science (Boston University, Information Systems focus), cloud and database architecture. Haedar’s ML expertise maps to the core technical IP: multi-modal sensor fusion algorithms, reinforcement learning navigation policies, and deep learning defect classification systems. His evaluation methodology and benchmark design experience is essential for creating the standardized performance metrics that autonomous pipe inspection needs — measuring defect detection accuracy, false positive rates, localization precision, and cross-material generalization (ductile iron vs. PVC vs. cast iron). His cloud infrastructure background maps to the data platform architecture: real-time ingestion of inspection data from deployed robots, scalable storage and query of multi-modal sensor streams, and API design for integration with utility GIS and asset management systems.
Ahmed
Director of Manufacturing specializing in Design for Manufacturability (DFM), production scaling, quality systems, and process optimization. Ahmed represents the capability most absent from the in-pipe robotics research community. Academic groups (Pipebots, MIT, SubMerge) hand-assemble prototype robots from custom-machined components; no research group has published on manufacturing these systems at volume. Ahmed’s contribution is structural: analyzing every robot module (propulsion, sensing, power, communication, articulation joints) for DFM compatibility — identifying injection-moldable housings, standardized waterproof connectors, automated test procedures, and supply chain sources for specialized components (miniature propellers, pressure-rated camera modules, piezoelectric transducers). His manufacturing readiness assessments at 100-unit, 500-unit, and 2,000-unit volumes provide the production cost data that grant reviewers and utility procurement officers require. This addresses the valley of death between TRL 4 prototypes and TRL 7+ deployable systems — the gap where most funded robotics research stalls because no one on the team has ever operated a production line.
The team does not include a water infrastructure engineer with utility operations experience. This gap is intentional and addressed by hiring: the first hire funded by a Phase I grant would be a field engineer recruited from a water utility or pipe inspection firm (RedZone, CUES alumni) who brings operational knowledge of utility access protocols, permitting requirements, and crew deployment logistics. HHA’s comparative advantage is the integration of AI + manufacturing + experimental design — the three capabilities that no individual pipe robotics lab possesses.
10. Recommended Next Steps
Target Funder Programs
- NSF SBIR/STTR Phase I — Smart and Connected Communities (S&CC) or Cyber-Physical Systems (CPS). Water infrastructure inspection falls squarely within S&CC’s scope. Phase I: $275K/12 months. Phase II: $1M/24 months. NSF has funded MotBot ($1.555M SBIR Fast Track) and UMD miniature robots ($850K CPS) in this exact domain — direct precedent for HHA’s sensor fusion and fleet management proposal.
- EPA Water Infrastructure Finance and Innovation Act (WIFIA) — Technology Innovation Program. EPA administers $11.7B in Drinking Water State Revolving Fund (DWSRF) loans. A pilot program pairing HHA’s inspection technology with 2–3 municipal utilities would qualify for DWSRF supplemental funding. Estimated: $500K–$2M.
- DOE ARPA-E — Water-Energy Nexus. ARPA-E has funded CMU’s confined space mapping module for in-pipe robots. Non-revenue water losses (11% nationally) represent embedded energy waste — framing autonomous inspection as energy infrastructure qualifies for ARPA-E’s water-energy nexus programs. Estimated: $1.5–$3M.
- DOT FHWA — Exploratory Advanced Research (EAR) Program. DOT funds the CAIS culvert inspection program. HHA’s proposal extends autonomous inspection from culverts to pressurized water mains — a related infrastructure class. Estimated: $500K–$1.5M.
- AWWA Research Foundation (Water Research Foundation) — Infrastructure Monitoring. Industry-funded research targeting utility needs. Smaller awards ($200K–$500K) but include built-in utility partnerships for field validation. Provides the water industry credibility that pure-robotics proposals lack.
Estimated Funding Range
Based on comparable awards: $1.5–$4M for a 24-month research program developing multi-sensor fusion algorithms, DFM-optimized robot modules, and utility data integration. Phase I target: NSF SBIR ($275K) or Water Research Foundation ($300K). Phase II escalation contingent on milestone achievement and utility partner validation data.
24-Month Milestone Timeline
- M1–6 R&D (Haedar): Develop multi-modal sensor fusion architecture — joint CNN+signal processing pipeline for visual, acoustic, and ultrasonic data streams. Train on public pipe inspection datasets (CCTV defect images, Pipebots acoustic data). Benchmark: single-modality baselines vs. fused model on paired test data. Manufacturing (Ahmed): DFM audit of Pipebots and SubMerge published designs — identify injection-moldable components, standardized connectors, and critical custom parts requiring specialized sourcing. Bill-of-materials cost model at 100-unit volume. Go/no-go: Fused model mAP exceeds best single-modality baseline by ≥10% on benchmark dataset.
- M7–12 R&D (Haedar): RL navigation policy — PPO agent trained in pipe network simulator (Unity/Isaac Gym), state space from acoustic echoes + IMU + pressure. Sim-to-real transfer validation on physical pipe test rig. Experimental Design (Hass): Field validation protocol — sample size calculations, statistical comparison framework (autonomous assessment vs. excavation ground truth), bias detection across pipe materials. Utility partner recruitment (2–3 municipalities). Manufacturing (Ahmed): First DFM-optimized module prototypes (propulsion unit, sensor pod). Automated test fixture for waterproofing verification. Go/no-go: Autonomous navigation through 10 consecutive junctions in pipe test facility; NSF/ANSI 61 materials testing initiated for wetted surfaces.
- M13–18 R&D: Integrated system test — DFM-optimized robot with sensor fusion pipeline deployed in partner utility pipe network (live pressurized mains, supervised operation). Data integration with utility GIS (ESRI ArcGIS). Manufacturing (Ahmed): Production-intent design freeze for 3 core modules. Supplier qualification for critical components. Assembly sequence optimization. Go/no-go: Defect detection mAP ≥0.75 at turbidity ≤10 NTU in live mains; MTBF >200 hours for propulsion subsystem; NSF/ANSI 61 certification complete for all wetted materials.
- M19–24 R&D: Multi-site validation — deploy at 2 partner utilities with different pipe materials and network topologies. Publish benchmark results (defect detection accuracy, localization precision, false positive rate). Manufacturing (Ahmed): Cost-of-goods projection at 500-unit and 2,000-unit volumes. Supply chain risk assessment. Pre-production pilot run (10 units). Deliverables: Phase II proposal (NSF SBIR or ARPA-E) for commercial-scale robot production and multi-utility deployment. Water Research Foundation membership application for industry validation partnerships.