Kayal

Problem: Keeping urban waterways clean is essential for healthy cities and also preventing ocean plastic pollution. However, the current global methods for waterway maintenance are labour intensive, inefficient and don't scale.  

Solution: We’re automating waterway maintenance with AI and drone technology. We are building all-electric, autonomous aquadrones that collect floating debris and remove weed using computer vision. They also function as mobile data acquisition platforms, tracking waste aggregation trends while recording water and air quality data. 

Impact: WHO reports that improving water quality can reduce the global disease burden by 4%. Our drones make waterway cleaning cheaper, more effective and more efficient, helping mitigate disease outbreak, flooding, and water scarcity. Further, data acquisition helps measure chemical discharge, intensity of pollution, etc. thus informing sound intervention. Our drones will also significantly reduce the litter outflow to the ocean via waterways, protecting marine ecosystems and its dependent communities.

In the developing world, the urban poor live closest to highly polluted waterways and face the highest risk of waterborne diseases and flooding. The slums lining Chennai and Colombo’s canals house more than two million people.

Marine pollution impacts the ocean’s function as a CO2 recycler and a heat sink, accelerating climate change, which disproportionately impacts the poor. ~7.1 million fisherfolk in India face the threat of flooding due to global warming. Consuming plastic filled fish is a health hazard. 80% of ocean litter is transported via inland waterways. 

Keeping inland waterways clean is critical for the livelihoods of the urban and coastal poor; but this is a very difficult task. In Colombo, the government employs daily manual labourers to fish out more than 15 tonnes of trash from the city’s 100 km canal network. Despite US$2 million spent on wages, much waste is left uncollected and ultimately ends up flowing into the ocean. The problem is more severe in other South Asian cities. Currently, workers paddle in a pontoon and scoop up waste with weir skimmers. This is a major health risk for workers and is also inefficient. Mechanized substitutes are expensive, heavy, and have high operating and maintenance costs.

We serve two vulnerable segments of the population: 

1. Slum communities in densely populated urban areas, particularly those lining waterways
2. Coastal communities dependent on the marine ecosystem for livelihood and food

Urban waterways in developing countries are narrow, shallow, and heavily polluted. To perform cleaning in such conditions, the device needs to be compact but capable of heavy-duty operation and offer a large payload. Further, the pricing must be suitable for developing countries. Our trash skimmer encompasses all these considerations and will have 10x the carrying capacity of comparable solutions in this space. 

Further, we have observed and interviewed a large number of cleaners, who also live in the same slums. Our interaction with the cleaners led us to decide that we will primarily sell human operated versions of our trash skimmers in regions where there is a shortage of jobs. Thus, we will not cause a loss of employment: some of the existing cleaners will be retrained as operators while others will be absorbed by the ecosystem assigned to safely dispose of the collected waste. 

Our solution will ensure that the waterways close to slums are clean, flowing, and free of larval breeding locations. 

Hardware

Our primary product is an intelligent trash skimming aqua drone that uses computer vision to scavenge for floating waste and aquatic weed. Currently, we are finishing the design and engineering detailing of this new design. 

Key design features:

• All-electric and highly energy efficient
• Superior hydro dynamic performance
• A novel conveyor manipulation system for independent collection and disposal of trash
• Small form factor and lightweight

Technical specifications:

• Dimensions: 3 m (length) x 2 m (width) x 0.75 m (height)
• No-load weight: 400 kg
• Average speed: 10 km/h
• Run time: 8 hours
• Payload volume: 1,000 liters 
• Payload weight: 200 kg
• On-board sensors: RGB camera, GPS, underwater sonar, depth, temperature and pH-level sensors
• On-board computer: Jetson TX2

Software

In parallel, we are working on the robotics software platform using a smaller, proof of concept drone we have built (see below). 

Trialing existing prototype

The test drone's current workflow is:

1. An operator defines a target search area with GPS points 
2. The aqua drone then scavenges for floating waste using computer vision within that geo-fence. We use k-means clustering and Speeded Up Robust Features for identifying floating debris and optical flow based tracking to continuously track and collect the waste
3. The drone returns to the starting position once it completes the work. 
4. The operator can view the on-board camera's live stream and also intervene at any point during the mission via a base station (either mobile phone or a laptop) that is linked with the aqua drone. 

Notes: The communication is via 3G/4G (any distance, provided both the robot and the base station have strong GSM connections) or WiFi (100 meter radius with line of sight). Additionally, the drone also supports Radio Control (1000 meter radius with line of sight). The test drone is currently trialing alongside Colombo's canal cleaners.

We are innovating on two fronts.

On the hardware design front, when compared to other available mechanized solutions in this domain, our drone (i) is all-electric and highly energy efficient, (ii) of a smaller form factor and lightweight - making it amenable for a variety of use cases and easy transportation, (iii) can collect and safely dispose trash without additional infrastructure or support, using a novel single conveyor manipulation system, (iv) has built-in fail safes, (v) measures  waste quantity, water and air quality, waterway depth, and comes (vi) geared for complete automation. 

On the software front, we have already built an machine learning platform that enables our existing proof of concept to autonomously detect, localize, and collect floating trash using an on-board camera; coupled with underwater sonar and GPS for safe positioning and navigation. The initial product we launch will be human operated, due to safety, technology maturity, and regulatory concerns. But we are aiming to roll-out supervised autonomy in the next 9-12 months.

We use computer vision and machine learning in the following ways:

(i) segment floating waste and weed from water surfaces, make decisions on what to collect and avoid. 

(ii) 3D localize the trash to be collected using structure from motion. 

(iii) infer absolute size and depth of debris by fusing IMU measurements. 

(iv) probabilistic positioning (SLAM) using a combination of vision, GPS, and underwater sonar readings. 

(v) learn over time where most waste typically aggregates and pre-plan cleaning mission accordingly. 

(vi) continuous determination of best possible navigational course to extend battery life. 

We have designed a compact, lightweight novel conveyor manipulation system for infrastructure independent, safe collection and disposal of trash.

We are building web and mobile applications for mission planning, monitoring, and fleet management.

Problem: in developing countries, urban poor disproportionately live next to polluted inland waterways. This places them under severe risk from spread of vector borne disease and other illnesses such as diarrhea. 

Our drones perform the following activities: (1) effective and regular cleaning of waterways; (2) continuous monitoring of water quality parameters (pH level, total dissolved solids, etc.); (3)  identifying and disturbing larval breeding grounds (feature to be built). 

Direct outcomes of our activities are waterways with vastly reduced floating and semi submerged debris, less stagnant waters, and deep information and understanding on the quantity of waste, their sources and trends, location of larval breeding grounds, and health of urban water bodies.

The key short-term outcomes are cleaner waterways next to slum communities, reduction in vector breeding grounds, and execution of interventions that stop pollutants at source, progressive reduction in the amount debris outflow into the ocean. 

The key long-term outcomes are significant reduction in the incidence of vector borne diseases among urban poor population in developing countries. 

Currently, we are trialing our existing prototype in a 1 km stretch in Colombo North. As such we are currently enabling better surroundings for roughly a thousand people or 200 families. 

In the next 1-1.5 years, we will be operational throughout Colombo Metropolitan Area: our drones will cover the city's entire canal network, cleaning over a 1 million sq meters of waterway, stretching 100 kms. The total slum population adjacent to the canals is close to a million people. They will experience cleaner waters and significantly reduced risk from vector borne disease and floods. 

In the next 2-3 years, we will be operational in waterways across 10 major cities in India and Bangladesh. Our sales target over this period is to put least 300 drones to work in our urban waterways in the first 3 years. We estimate cleaner waterways in these cities will directly impact 50 million people from slum communities. 

In addition, we will be stopping an estimated 10 million kilograms of trash, mostly plastics, from entering the ocean just in Colombo upon complete launch. We will be directly preventing and reducing marine plastic pollution and indirectly protecting the lives and livelihoods of coastal communities. 

Year 1 - Jan 2019 - Dec 2019: Design and build winning hardware for electric trash skimming aqua drone in urban waterways.

Year 2 - Jan 2020 - Dec 2020: Launch aqua drone across Colombo's canals. 

Year 3 - Jan 2021 - Dec 2021: Launch aqua drone in 10 major cities in India and Bangladesh. Introduce aerial waste mapping drone technology to existing clients.  

Year 4 - Jan 2022 - Dec 2022:  Launch aqua drone + aerial waste mapping in additional 20 cities, including in Africa. Launch mosquito larval surveillance drones.

Year 5 - Jan 2023 - Dec 2024: Launch aqua drone + aerial waste mapping in additional 40 cities, including in Latin America.  Mosquito larval surveillance drones operational in Sri Lanka and India. 

Our vision is to tangibly aid at least 200 million urban poor in developing countries, lead lives that are enriched by cleaner waterways, at reduced risk from the spread of vector borne diseases, and climate crisis impacts such as flooding.  

1. Financial: ability to raise Series A funding in Sri Lanka within the next 2 years. The local startup environment has low liquidity. As an early stage deep tech, Kayal has a high capex requirement. It is difficult to source the depth of funding required for deep tech within the Sri Lankan startup ecosystem.

2. HR: A key barrier is hiring top quality robotics talent. 

3. Legal: We will file for a patent once the current design iteration process is over, guarding our unique IP. This is a barrier we will overcome in the next year.

4. Market: The problem of polluted waterways is most acutely felt in developing countries in South Asia, Southeast Asia and Sub-Saharan Africa. Kayal is launching in Sri Lanka and we plan to then expand into markets such as India, Bangladesh, Indonesia, and South Africa within the next three years. Each new market has a unique local business and regulatory environment. 

5. The main buyers for our products are central and local governments, ports and other government agencies with maintaining large water bodies. Governments are buyers with a long lead time; however, they can be sticky once the relationship is established. 

1. We are applying to international sources for both venture capital and grants. We are also constantly tapping into and growing our external network through startup competitions and cold emails to engage and connect with potential new funding sources. 

2. We are developing an HR plan that involves connecting with local graduates who will have the necessary skillset that we are looking for. 

3. We are working in stealth mode until we are able to file the IP. Additionally, we will work with the government's institution that files patents to minimize delays in the patent filing process. 

4. By applying to international competitions, we are expanding the base of individuals we connect with. We are strategically networking to find people who can facilitate introductions in the markets we plan to expand into. Currently ICTA and ClimateLaunchPad (detailed in sections below) have committed to connecting us with customers locally, in India and other South East Asian countries.  

5. We are working to bring relevant politicians or Ministers onto our board to champion the startup and support procurement processes. Recent 2019 budget initiatives encouraging government to favour startup procurement and increase exports is also encouraging Kayal’s position.

3 full-time

- Computer vision engineer (BSc in Electrical and Electronic Engineering)

- Electronic engineer (BSc in Electrical and Electronic Engineering)

- Industrial designer (MSc in Industrial Design)

2 part-time

- Overall technology adviser (PhD in robotics) - 20 hours a week

- Sales and finance (MBA, Wharton, incoming) - 10 hours a week

Currently, we have a core team of five people - 3 engineers (1 with a PhD in robotics)*, 1 industrial designer, and 1 business and finance person. Between us we have the following technical skills: electrical engineering; computer vision; robotics; software development; industrial design. We have a team that is conversant in 5 languages: English, Tamil, Sinhalese, Dutch, Tamil (all very strong), and Hindi (weak). 

We have spent more than a year studying and understanding the problem domain we’re working on. 

We have built a solid network of promoters in India and Sri Lanka who will help us take the technology into governments. We have a partnership with the Sri Lanka Land Reclamation and Drainage Corporation and they are ready to pilot our technology as soon as it is ready. 

The founders of Kayal Technologies, Elijah and Janagan, started working full-time on this project in January 2019. At the time, we had not raised funds, and only had a partially evolved idea. Today, we have raised $42,000, built a working prototype, established a partnership for a pilot, and in the process of reaching out to more end users. We are growing fast and are committed to making our technology the go-to waterway maintenance tool across the developing world.

Note*: Dr. Senthan Mathavan functions as our overall technology advisor. He currently spends 20 hours a week at Kayal. He will join full time as CTO starting from January 2020. 

Sri Lanka Land Reclamation and Drainage Corporation

- This is the government agency in-charge of waterway maintenance in Colombo. We are currently trialing our prototype with their manual cleaners as we continue to improve our computer vision technology. This institution has also agreed to pilot our first market-ready product in December. 

ClimateLaunchpad - Climate-KIC (EU) & ClimateStudio (India)

- We are national finalists at ClimateLaunchpad in Sri Lanka. We are currently going through their accelerator programme. ClimateStudio is ClimateLaunchpad's partner organisation in India. They are helping us discover customers in India. 

Information Communication Technology Agency - Sri Lanka

- We got funded on 17 July 2019 by the ICTA Sri Lanka, the government institution for IT, through their grant programme for the top 12 startups in the country. We are getting business mentoring now. They will also link us to foreign customers as their 6 month incubator programme progresses. 

We have a three-pronged business model. (i) We sell aqua drones to public and private entities involved in regular waterway maintenance at $20,000 per unit. Payment is received before the machine is shipped. (ii) We plan to charge a subscription fee of $200 per annually from customers who use our fleet management and data visualization web application. (iii) We contract out waterway cleaning services to those entities who require periodic waterway cleaning at $40 per hour (or per 10,000 sq. meters). Here, payment is received upon the completion of the mission. 

We estimated that it will cost roughly $7,000-$8,000 to manufacture and sell one drone. Similarly, it will cost $15 per hour (in Sri Lanka) to offer the cleaning service, including electricity for charging, payment for the operator, and transport. 

We will employ direct sales for exports to Europe and US markets and the mission as a service model locally in Sri Lanka and in India through a partner.

We estimate our first year expenses to be $48,600. We have raised $42,000 of this amount. In summary, we have a runway of 12 months. We are drawing low salaries and are running a frugal operation - we may be able to stretch the runway to 18 months. 

We will require another $3 million coming into the company at the end of the runway to cover the next 30 months. We start making revenue in 12 months from now (July 2020) and will break even in another 26 months (October 2022). This is so if we spend heavily on tooling in 2021 and ramp up our production throughput by 10x from 2020, and then aggressively pursue a 2x year-on-year revenue growth in revenue in successive years. 

Our path to financial sustainability is: build market ready product by December 2019, pilot with established partner in Sri Lanka in January and February 2020, source partly paid purchase orders during this period, raise $3 million using the purchase orders as evidence for traction. 

We have already set up a Delaware C Corporation in the USA to enable easier fundraising. 

Annualized estimates are as follows:

2019:

Revenue: 0

Expenses: $30,056

2020:

Revenue: $124,773

Expenses: $255,664

2021:

Revenue: $1,247,216

Expenses: $1,255,297

2022:

Revenue: $2,337,457

Expenses: $2,179,238

2023:

Revenue: $4,569,769

Expenses: $4,146,248

2024:

Revenue: $12,599,146

Expenses: $11,142,211

2025:

Revenue: $25,902,275

Expenses: $22,195,347

We are applying to Solve for the following reasons:

1. Receive mentoring and support on product and technology development. 

2. Recruit top AI talent to work with us or be on the board. 

3. Connect to US investors in the sustainability / clean-tech domain to help with further fundraising.

4. Discover customers in the US and other places across the globe, through participation in global water pollution related conferences and exhibitions.

5. To get coaching and streamline our media and social media outreach strategy.

Technology 

University of Washington:

• We have so far done only limited work on water quality monitoring. UW is building Adaptable Monitoring Platforms for cheap but efficient real time underwater sensing. We would like to learn from their expertise and/or partner with them to build this aspect of our product. 

MIT

• We have done considerable work on computer vision based detection and localization of floating debris. However, we would benefit immensely from guidance from MIT’s leading computer vision experts on further optimization to run on cheaper hardware, training lighter models, etc.

Business

UN or similar multi-national agencies working on clean tech and sustainability. We know that the SDGs specifically target healthier and cleaner cities and decrease marine pollution. We would like to reach out to new customers by partnering with international agencies that work to improve waterway maintenance in developing countries.

Additionally, we are keen to partner with private entities involved in waterway maintenance. 

Our prototype uses geometric computer vision techniques for detection and localization of trash. 

But, we are working on an end-to-end deep learning model for the whole operation cycle of the aqua drone - i.e. positioning, identifying and tracking trash, navigation, and servo actuation.  

We have made significant progress. (i) We are now able to obtain better positioning information by coupling visual SLAM with GPS and sonar readings; (ii) We have incorporated semantic scene understanding to distinguish between water surfaces, lining walls, boats, etc. (iii) We have trained a CNN model that can distinguish between five different types of commonly encountered floating waste with about 95 percent accuracy on the validation set. We are exploring ways it can perform faster at inference. (iv) We are able to better infer the absolute depth and size of to-be collected debris by fusing IMU measurements. (v) We have set up a data acquisition and training pipeline to continuously improve our models after each trial run. 

We will use the AI Innovation prize in the following way:

We have an aqua drone that can collect and dispose of trash independent of human or infrastructural support. But a human operator has to plan the mission. For complete autonomy, however, we need a robotic agent to plan the mission as well. We want to do this via a flying drone that takes aerial footage of waterways and geo-locates floating debris. We will use the AI Innovation prize to build this system and integrate it with our existing platform. 

Vector borne diseases such as malaria and dengue remain major health challenges in urban areas in developing countries. In 2017, there were 417,000 deaths from malaria in Africa. In 2017, there were close to a million reported malaria cases in India. In Sri Lanka, last year, there were 51,000 reported cases of dengue, most of this was from Colombo, Sri Lanka's largest city. 

We believe that our core computer vision and AI capabilities can generalize to adjacent domains. One area we want to specifically explore is combining aerial imaging - RGB and multi-spectral - and video and sensor readings obtained from our aqua drones to to detect potential malaria and dengue larval habitats. There are already encouraging signs for the potential of aerial imaging technology for the detection of larval habitats. Fusing ground readings will further improve the results and could fundamentally change the way in which health agencies do larval habitat detection.

We will use the Healthy Cities Prize to improve our solution in this direction.