Last week we were at an event in Singapore and a large hardware company asked this question about we were able to score some of the largest customers for our product. At the end of our explanation, they were quite excited to work with us.
The key challenge is how do we move a 5 feet tall, 50kg robot autonomously in rough conditions without risk of toppling even with a push, does it 12 hours in a single charge, make it easy to packed in a suitcase to take in plane/train/beauty rickshaw/lorry, be durable to be in events every day around the world with all sorts of software features (face recognition, NLP, recommendation). All while manufacturing in house, with no external funding or years in development time.
Autonomous navigation in an uncontrolled environment
Our robots have to move around in places like malls and banks with no human supervision.
This is a pretty hard problem to solve.
- We cannot change anything in the environment for a line follower like arrangement. This rules out the simplest of navigation techniques that is commonly used.
- The environment is uncontrolled unlike a factory warehouse. A robot moving in a warehouse will be well trained for the environment and people around are also trained for it. Our guests [who could be any random mall visitor] are not trained to be around robots.
- We cannot use LIDARs due to cost, power constraints but most importantly due to its failures in areas with a lot of glass surfaces [like malls or banks].
- We need a lot of precision – currently, we get about 2 cm precision.
- We need a global mapping and not merely local. We have location specific content fed from the cloud.
- The sensors have to be cheap and affordable. We sell the robots cheaper than the price of the high end LIDARs.
- It is a 5 feet tall object and has toppling risks. We had to design to make it very hard to get toppled as it moves around.
All of this in a production environment and not as a research project.
Easy to assemble and disassemble
Our robots are about 5 feet 50kg – that is quite big, but we designed it to be carried at a moment’s notice in suitcases and then quickly assembled. This agility is quite unique about ours. This has enabled us to do events around the world and we have done 45 major events in the past 1 year.
The ability to dismantle it all and carry in regular suitcases allows to go past all the customs and other hassles in air travel. Besides air, we have carried the bot in trains, buses, lorries, auto rickshaws and even two wheelers. And we are designing to be supported by people with very little training.
Durable, long running and under dynamic conditions
It goes through brutal transportation hurdles – such as being thrown into a bus or air cargo – and still survives it all. Very few electronics items, leave alone robots can survive such stresses. This is where the difference from a lab project show significantly.
We built a suspension system that enables our robot to run even on roads as poor as ours. And for a 5 feet tall object, it is quite stable and very noiseless and graceful in movement.
And our customers expect the batteries to run 12+ hours.
Most robots run out of juice in an hour or so. When even phones struggle to run the whole day, there is an expectation that a 50kg object with all the screen and speaker connectivity, with bluetooth support and navigation to go the whole day.
Key software features
- Contextual voice engine to have a multi level conversation even in a very noisy environment
- Face detection
- Face recognition and the ability to store 1000s of faces
- Cloud management of the robot
- Event specific features
- A basic recommendation engine
A lot of newbies would ask, I can say they can build those features quickly. They might actually do with an available framework. The problem gets tricky not with features, but how all those features work together [integration], building test cases to make it production ready and under real life constraints of cost and time. That is what differentiates newbie developers from seasoned – the former can build features, the latter can also integrate them and put them in a bigger product.
We built it all in India that made some of our processes harder, but we also have more control over the process. Since we built from scratch, we have the ability to go swift with the design changes.
What Invento excels at?
A strong hands-on culture. When we wanted a large conference table, our team went and built something from scratch in 4 hours. Extremely sturdy and cheap. We have also built a variety of things in our office, including a car and a semi-autonomous scooter.
Highly capital efficient. Hardware R&D companies drain money like anything.
We have not taken external funding and run with the money we make from events and sales. That means we have to be very scrappy and be super efficient with our resource.
A team of non-roboticists who learned robotics. Majority of the team comes from outside the robotics background. That was a disadvantage initially as we had a learning curve, but now we are able to truly utilize our other specializations and can run faster than some of the competitors.
We had to make some hard design challenges to make the product commercially viable and be scalable with logistics.
These woods are lovely, dark and deep,
But I have promises to keep,
And miles to go before I sleep,
And miles to go before I sleep.
—— Robert Frost