India needs to increase robot adoption, develop innovative robotic automation: CynLr

CynLr is a visual object intelligence platform that enables industrial robotic arms to see, understand and manipulate any object. It is striving to make robots smart and intelligent.

Here, Nikhil Ramaswamy, Co-Founder and CEO, and Gokul NA, Co-Founder and CTO, CynLr, tell us more. Excerpts from an interview:

DQ: Talk about the evolving industrial robotics landscape in India over the years.

Nikhil Ramaswamy: Ever since its inception, industrial robotics has been an integral part of manufacturing with automotive industry being the largest buyer. The first ever industrial robot was deployed by General Motors to automate welding operations. With improvement in machine vision, grasping technology and collaborative robots there has been a surge in the deployment of robots for a wide variety of applications, especially in electrical/electronics manufacturing.

Globally, electrical/electronics industry has become the largest buyer of robots in the last year with China being the largest buyer (close to 44% of robot sales - IFR). Robots have played a crucial role in bolstering China’s position as the world’s largest manufacturer.

A PwC study found that over 50% of CFOs surveyed expressed an interest in accelerating automation across their organizations. This increased interest to automate reflects in the growth projections for industrial robotics. As per MarketStudy Oct. 2021 report, the industrial robotics market is expected to grow to $87.79 billion at a CAGR of 10.35%. Rising labour costs, supply chain interruptions, political tension and fierce competition are driving global manufacturing leaders to diversify their manufacturing hubs. The political tension between US and China is forcing global corporations to actively look at alternatives.

India, one of the contenders, faces stiff competition from countries such as Vietnam, Thailand, South Korea, and Indonesia. Of the $31 billion in manufacturing output that moved from China to other Asian countries, India captured only 10%. Although the number of robot installations in India grew at a CAGR of 20% between 2013-2018, India’s current robot density (number of robots deployed per 10,000 employees) is only 4 (IFR), which is well below that of any of the competitors.

To put it in perspective, the robot density for China is 246 and the global average is 126. To take advantage of the current tailwind, we need to not only increase our robot adoption, but also develop a competitive advantage through innovative robotic automation that can help us leapfrog over global competition rather than simply play catch up.

DQ: How is the emerging integration of automation of technologies and innovation in AI in the robotics segment?

Gokul NA: Today’s robot automation systems are extremely precise and accurate but also unfortunately rigid. Although robots are programmable, they can perform only a specific task on a specific part that is placed in a specific location and in a specific orientation. This means there is significant customization that one needs to invest in to deploy a robot. The investment is often higher than the cost of the robot itself. As a result, traditional robotic automation systems require the product life cycles to be almost seven years to make the RoI for robotic automation work.

The world has changed leaps and bounds in the last decade. Product life cycles are getting shorter and shorter. Product designs are revised and changed frequently. If a manufacturing line cannot handle even minor changes, then it becomes a roadblock in today’s dynamic world. Manufacturers have resigned themselves to relying on manual labour to meet the required flexibility and versatility in manufacturing processes. The downside is that this approach is not scalable and has a heavy dependence on availability of skilled manual labour.

Today, even a simple task of putting a screw into a screw hole without slipping a thread remains non-automatable across the globe. Imagine what it would take for a robot to assemble a smartphone or a car by putting together 1000s of parts of varied shapes and weights presented in random orientations. An average car has 3500 fasteners and an aircraft such as Boeing 747 has three million fasteners. Enabling robots to perform picking, orienting, and placing of objects straight from a container has been a long-standing unsolved problem – touted as The Holy Grail of Robotics.

Another segment where there is considerable investment in robots is warehouse automation. Even in this segment, robots are mostly only deployed for moving goods between stations or humans. Picking, kitting, and billing tasks, that take up a large chunk of the operational cost, are still largely manual.

This problem statement has sparked the infusion of AI/machine learning and machine vision with the world of robotics. The field is going through a transformation where robotics is not just looked at from a hardware perspective, but from a holistic perspective – software and hardware going hand in hand.

DQ: What has been CynLr’s innovation in contributing towards visual object intelligence for industrial robots?

Nikhil Ramaswamy: Most of today’s AI based approaches and machine vision technology in robotics rely on predicting the location of an object based on static images, 3D or otherwise. The issue in this approach is that the orientation of an object and lighting can significantly change the color and geometrical shape of the object that the camera perceives. The situation is much more complex when we deal with metal objects with mirror finish. Thus, traditional AI methods that rely on color and shape to identify objects are not as universal.

To truly automate with the required flexibility and reliability, we need robots that can see, understand, and interact with objects the way humans do.

We at CynLr, build the missing pieces of visual intelligence for robotic arms that make them object aware. This enables robotic arms to manipulate objects with superior agility - adapting to the varying shapes, orientations, and weights of the object. We envision that this technology could simplify and standardize the large, rigid, and customized manufacturing lines into LEGO blocks of micro-factories.

DQ: How did Covid-19 accelerate operations of automated technologies?

Gokul NA: The Covid-19 pandemic gave a glimpse of how sensitive our manufacturing ecosystem and infrastructure are to disruptions in manual labor and supply chain. The pandemic posed important questions to manufacturers who assumed they were automated enough to remain unaffected. Companies discovered that owing to the rigidity of today’s manufacturing automation practices, there is still a large dependency on manual labor.

The bulk of tasks in manufacturing and individual piece-picking in warehouses are still manual. For example, manual labour is employed for the seemingly simple task of machine tending of CNC machines. Even if a manufacturer were to invest in an expensive customisation to automate machine tending, a human being is needed to prepare the structured pallet that a robot can operate on.

Manufacturers have now recognized that flexible automation is the need of the hour to overcome challenges posed by shortage of manual and skilled labour.

DQ: What is the company's roadmap for the next 3-5 years?

Gokul NA: Our aim is to add object intelligence to any computing system that may interact with an object. We are starting with industrial robots and automating tasks in a manufacturing plant that are currently not completely automatable using today’s technology. Our proprietary vision and grasp technology aims to make robots as dynamic and versatile as humans in performing any task involving object manipulation.

The first set of applications we are addressing are machine tending with random bin picking, kitting, packaging, and inspection. We believe that this will catalyze a disruption in manufacturing by helping build the intelligent factories of the future – adaptable, streamlined, and cost-effective.

In the longer run, we are working towards building dynamic neural models of objects. We envision the utility of such technology to be beyond manufacturing and at the juncture of any system that needs to interact with any object.