When we think of how to combine AI, autonomous functioning and robotics, Hollywood and smooth-tongued screenwriters have us convinced that the future will land somewhere between “The Terminator” and “Her”. While flights of fancy are necessary to keep life interesting, the fact of the matter is that robots that can work autonomously will simply be too useful to take over the world. What they’ll really do is fill long-needed roles that ultimately give us more freedom as people than we’ve ever known.
According to MIT, Autonomous robots “are intelligent machines capable of performing tasks in the world by themselves, without explicit human control.” In this context, we know that human control is a limiter, but what are the limits?
In order to achieve robot autonomy, the first limit a robot must surpass is the ability to program itself – the ability to take goals that are set for it and figure out a way to achieve it without manual human intervention.
The next steps towards robot autonomy are, as it turns out, much more qualitative. It’s in these qualities – whether its a service provided, a process executed or data generated – where the idea of robot autonomy really comes to life.
Robot Programming Is Tiring, Tedious, Terrible
The truth is, robots today are generally very “dumb”. That’s not an epithet of course – considering what they can achieve when well-programmed, they are incredible powerful solutions. Every robotics company is working towards making the robots they sell easier to use, and while some robotics startups are working to do the same thing on top of existing hardware, every single company in the space is hoping that one day robots will be able to program themselves.
Why is this a common goal in the robotics industry? Because robot programming is a hugely tedious process this often takes up a significant budget within any robotics integration. For those that aren’t familiar with the process, robot programming – though fun for many professionals – is painful to get right, as it often entails up to 6 degrees of freedom, dozens of potential tool paths and end positions, and a variety of kinematic constraints that require deep consideration in order to generate the most efficient possible robot motion program.
At the same time, this process of motion generation must be done manually for each and every part and environment that a robot is expected to work in. This means that, up to today, robots have been practical for companies in industries like transportation and consumer electronics, but impractical (with the exception of specialized applications) for most companies that aren’t churning out tens of thousands of high-value SKUs per year, with an otherwise limited variety of products.
That’s where self-programming capabilities can become a game-changer for the manufacturing industry. The majority of the costs that come in using a robot come in programming it and then structuring a manufacturing space to meet its needs. By reducing programming to something that is automatable, it is then possible to allow robots to adapt to both the parts and environment around it. Achieving this means that robots could finally be used by all kinds of manufacturers, no matter the part mix or SKUs involved.
What does it take to achieve “Self-Programming”?
A variety of solutions have been introduced over the years (and decades) to make it easier to program robots. Teach pendants, kinematic instruction and offline programming are all ways in which robotics providers have made the process easier and more useful for engineers, programmers and integrators.
Ultimately, these all can automate some elements of the process or at least make them more intuitive, but each complete motion still takes hours, days or even weeks to generate and validate, and all of course within highly structured environments.
In order to think about what a robot needs to generate its own motion, one key answer is thinking about what could make robots more like us. Without getting too philosophical, we human beings rely on our senses to learn almost everything we know. At the same time, our senses are what allow us to interpret and respond to our environment – much of this happening unconsciously, reflexively, instinctively or even being learned as we age.
Giving robots the senses they need to then respond to their environment – the ability to interpret 3D space and textures in particularly – would then let them gather the information they need to program themselves in order to accomplish a goal.
A self-programming robot, on its own, could be achieved with the right set of algorithms, but to what end? The real challenge here is letting a robot program itself towards a conventionally achievable goal – taking the same ability of a human to effectively program a robot with existing tools and making it an automatic process.
The Process in Spray Operations
How does this happen? To take the example of spray processes, the essential need in order to achieve a quality output is, put simply, the ability to identify and interpret the part in space. By knowing this, with the right ability to actually generate a program, a robot can then achieve a goal of covering a surface with coating (or spraying it with a thermal gun or abrasive materials).
Generally, this accuracy needs to go down to one millimeter or less in order to achieve the same acuity that human beings have during a work process. Fortunately, this is a level of accuracy that is achievable with infrared sensors, which when combining multiple angles and positions are able to provide the information needed to render a Digital Twin of an object in 3D space.
From there, that kind of capability allows firms to automate tasks that have been seen as tedious and even dangerous for human workers. Spending 40 hours per week, 50 weeks a year for 30 or 40 years spraying and coating industrial parts isn’t the best thing you can do – whether for your health or your psyche – while adding a robot creates material benefits that make other jobs easier and more fun to do.
In comparison to traditional robot programming, the benefits aren’t even close. In most coating operations, there are a wide variety of parts that need work – whether in a given day or given production run. Many manufacturers in these circumstances have tried and failed to use robots, and automotive manufacturers are generally the vast majority of paint robot users today. With self-programming robots, however, that breakdown will shift drastically.
After Sense, What Comes Next?
Now that you understand the goals that a self-programming robot can achieve, it’s important to understand what goes into the process from front to back. After generating its sense of the environment, a self-programming robot then needs to process all the conventional requirements of an industrial robot, with the added need to understand the unique characteristics of a process or cell. This could be process know how (e.g. where should I step to get out of someone’s way or what pattern should I use to apply paint to a surface), or specific instructions input through a basic software interface to make sure exact quality goals are prioritized – all, still, of course with no programming required.
Below is just one example of how the self-programming process is organized. Omnirobotic’s AutonomyOS™ is the world’s only platform for truly autonomous manufacturing. Using 3D Perception with AI-based Task Planning and Motion Planning, manufacturing engineers and integrators can configure autonomous robotic systems for value-added processes. By being connected to process controllers in a factory setting, as well as being aware of the position and constraints of the robot itself in space, the robot can then use process knowledge, task breakdowns and motion ordinancing to engage in a parallel task planning process. Each outcome is assessed in the Digital Twin environment, which then leads to the most optimum possible outcome – sometimes on the basis of 10,000+ scenarios per operation.
By using this ability to prioritize tasks, a robot can effectively generate a series of motions for itself that adds up to a complete process. Because robots are inherently more consistent, predictable and can work faster for longer than humans (as should be expected from machines, of course), the ultimate output of a self-programming robot is much more consistent than any solution out there.
If You Make Something, You Have To Have A Purpose
AI may replace some jobs, but in that process of automating high-value bottlenecks in all our production and work processes, they will actually create demand for more creative, innovative and intuitive jobs from human beings. While one day there may be a way to automate these uniquely human traits, why would we want that to happen? And, of course, would we as consumers be as interested in the “creativity” produced by a machine?
Of course, self-programming technology gives us just as much of a kick in the imagination as a good movie, but its important to anchor ourselves in the practical and cost-effective benefits of any technology. There is a clear need for more efficient and effective production processes, and a shortage of skilled labor has persisted – even in the midst of record unemployment.
The fact is, using more robots is the best way to get people back to work. How, you ask? As you create more productivity (and hence profitability), that enables new jobs to actually be created, wages to rise, and more of the members of a given company or society to benefit from the fruits of their labors. If you’d like to understand more about why robots are going to be so much more necessary to both business and society in the future, check out this great Robo Busines Direct talk by famed roboticist Rodney Brooks.
And, of course, if you’d like to understand how self-programming technology can already benefit the spray processes of today, don’t hesitate to check out our main website.
With AutonomyOS™ and AutonomyStudio™, it’s never been easier to deploy an autonomous robotic system. Using 3D Perception with AI-based Task Planning and Motion Planning, manufacturing engineers and integrators can configure autonomous robotic systems for value-added processes that allow manufacturers to achieve more consistency and flexibility in production than ever before.