Sign In


Smart Drain Inspector

Challenge Statement

How might we deploy a cost-effective solution to autonomously inspect our drainage network to identify drain defects and map out the drain alignment/dimension?

Challenge Owners

Catchment and Waterways Department – Drainage Operations Division

Background & Current Practice

In Singapore, rainwater is collected through 8,000 km of drains, canals and rivers. As part of the maintenance regime in PUB, public drains are inspected regularly (typically once every 5 years) for defects and subsequent rectification to ensure that the drainage network is structurally safe and functional. While we would like to increase the frequency of inspection, it is not sustainable to do so with current methods.

Currently, PUB conducts manual inspections for most open and closed drains, which is labour intensive. In addition, closed drains with confined space pose potential atmospheric and physical hazards. Man-entry into closed drains is potentially hazardous and requires precautions, which reduces inspection efficiency and increases costs.

Recently, PUB has deployed autonomous drones that can navigate in GPS-denied and unilluminated closed drains to provide visual images of the drains without requiring man-entry. This has improved productivity by reducing the time and manpower required and enhanced safety by eliminating the need to access and work in a confined space environment. However, the current drone solution can only be used to inspect drains that are more than 2m wide. It is unable to inspect smaller drains (i.e., less than 2m wide), which form a significant portion of the drainage inventory, due to the bigger drone size required to house a battery of sufficient capacity. 

During the deployment of drones in drains more than 2m wide, there are also several challenges encountered due to their relatively big size and water-phobic nature. For instance, obstruction due to underground services or animals makes it difficult for the drone to manoeuvre around. As the drone is water-phobic, the drone operator needs to take additional measures (i.e. manually follow the drone) w​hen flying the drone in a partially submerged drain to prevent it from coming into contact with water.

In addition, beyond visual information, we are seeking solutions that automate the detection of defects with high accuracy and pinpoint exactly where such defects are within the network. We also require the solution to map out the drain alignment and dimensions into a geospatial database as it performs the inspection. This will help to automatically update the drainage records to confirm as-built drawings and provide new digital mapping data, particularly for older drains.  

While PUB has deployed autonomous drones to replace man-entry for a small number of closed-drain inspections, we are open to both drone and non-drone solutions that can achieve the above. 

Areas of Opportunity

We are interested in solutions that can autonomously

  • inspect our drainage network which consists of drains of various types and dimensions (refer to Resources, under “Dimensions and Types of Drains”);
  • identify the drain defects and anomalies (refer to Resources, under “Examples of Drain Defects”) automatically using AI-based video analytics; and
  • map out the alignment and dimensions of open and closed drains in both GPS-enabled and denied environments, to an accuracy of 0.1m for horizontal and vertical displacement and upload this information into a geospatial database.
Key Considerations and Challenges

The solution shall be:

  • Able to inspect drains that could be under various conditions, for example partially or fully submerged conditions due to tidal influence, turbid waters, presence of animals (e.g. bats, snakes), obstruction by services, etc.
  • Cost-effective to enable large-scale implementation across the drainage network.
  • Able to fit through access openings (refer to Resources, under “Dimensions of Access Openings”) that lead to the drains.
  • Able to automatically and accurately detect and identify defects, ideally providing measurements to determine the extent of the defect.
  • Able to automatically generate inspection reports with location data, images and video timings to support subsequent investigation and rectification works.

The inspection and data processing methods, e.g. data transfer and report generation should adopt innovative methods that require minimal manual interventions in order to achieve smarter operations

In addition, the need for personnel to follow the autonomous robot/drone during the inspection to retrieve it in the event of failure would be a strain on manpower resources and is disadvantageous
Current Technology Status

An autonomous drone is currently used for the inspection of closed drains that are wider than 2m.

Expected Timeline

Total project period – less than 18 months

Expected Outcomes

A cost-effective site-tested prototype system that can autonomously inspect drains of various types and dimensions, automatically identify drain defects in real-time during an inspection and map out the drain alignment and dimensions to an accuracy of +/- 0.1m for horizontal and vertical displacement into a geospatial database.

If the pilot is successful, the solution would be provided to PUB through a service model where the equipment is owned, operated, and maintained by the company.


Q: What is the distance between drain access openings?

A: The distance between access openings for the closed drains ranges from 6m to 50m. There may be outlier locations where the distance between openings have changed due to nearby developmental works. These are however few and localised.

Q: Does PUB have information regarding the location of drains with obstructions, e.g., service crossings?

A: PUB does not have all the information regarding the location of drains with obstructions. Therefore, we are looking for agile solutions that can overcome obstructions during the inspection process.

Q: What is the maximum liquid flow rate in the drains?

A: Inspections are to be carried out during dry weather conditions. Drains are designed with flow velocity between 1-3m/s during dry weather.

Q: Are there any sudden drops in depth along the drainage line?

A: The drains are built to have a gradual gradient, but there may be localised sections with sudden drops in depth (e.g., cascading profile) due to the ground profile.

Q: When is the inspection expected to be conducted?

A: The inspections are expected to be conducted during daytime and on days without heavy rainfall. However, it is important to note that there could be a sudden downpour of rain, and there needs to be a plan to retrieve the equipment quickly from the drains during such events.

Q: Is it a requirement that the inspection conducted by the robot covers all the internal surfaces of the drain?

A: Yes, the required inspection areas entail all the drains' internal surfaces, including the base slab, walls & covering slab (also known as the top slab).

Q: What are the required inspection intervals to assess the drain measurements and alignment, e.g., assessment every 0.1m?

A: We do not have specific requirements for the inspection intervals as long as (1) any visible defects present along the continuous length of the drain are detected and (2) the alignment and measurement of the drains are accurate to +/- 0.1m.

Q: What are robotics systems that have been trialled? What are the improvements needed?

A: We have trialled indoor drones, and they can only inspect drains of more than 2m in width. The needed improvements include the ability to inspect drains of less than 2m in width, the ability to navigate or manoeuvre partially submerged drains and the ability to overcome obstructions such as service crossings. More details and examples can be found in the presentation. 

Q: Is PUB open to other types of inspection or sensing technologies besides visual-based inspection technology?

A: Yes, we are open to better alternatives that can meet our requirements while remaining cost-effective.

Q: Would PUB consider a solution comprising several robots of different types to perform the inspection?

A: We are open to trying different robots as we understand that different robots could better suit different operating conditions. We are looking for robots that can perform inspection for the different drain types to cover the entire drainage network.

Q: Does the solution need to be compatible with any existing database or software platform?

A: We are currently using ArcGIS to store our drainage data. Ideally, the new robotics platform is compatible with ArcGIS.

Q: Is there any specific format requirement related to the geospatial database?

A: We are keeping it open for our applicants to suggest the data that are more useful to display on the interface as long as the information is in ArcGIS compatible format.

Q: Do solution providers need to use any specific cloud platform?

A: For information, PUB uses a cloud platform that abides by the requirements of the Government Technology Agency (GovTech). Visit this page for more information regarding the Government on Commercial Cloud Framework. However, for this pilot project, solution providers need not implement solution via GCC.