The waste and recycling industry is becoming smarter every day. From the integration of artificial intelligence (AI) to the growing use of robotics, advancements in technology are helping to improve operations and make jobs safer.
One company employing its technologies in recycling facilities is Denver-based AMP Robotics, which creates innovative technologies to help facilities sort materials more accurately and efficiently and collect and understand data to better measure performance.
We recently sat down with Mark Baybutt, head of research and development at AMP Robotics, to discuss the impact artificial intelligence and robotics could have on the future of recycling.
Waste360: Looking back at 2019 and years prior, what advancements have you seen in AI and robotics?
Mark Baybutt: One of the biggest advancements has been the evolution in the field of deep learning within AI. Modern AI techniques are allowing machines to quickly identify and recognize objects on par, and in many cases, better than humans.
We’ve seen this progression from traditional machine learning to advancements in deep learning techniques, which means faster classification of items in the material streams and a maturing of our neural networks that allow deeper knowledge or inference of those items. Historically, the networks could generally tell the difference between two objects—say PET [polyethylene terephthalate] and OCC [old corrugated containers]—but now, they can learn finer details and feature associations enabling capability like specific brand recognition.
The field of deep learning is accelerating. That has been one of the biggest changes—pairing modern compute power with research and academia allows us to apply very powerful deep learning to the recycling industry.
On the hardware side, we’re seeing a parity in the natural evolution of robotic platforms. They are being added to more industrial environments, and they are becoming more robust and reliable.
Waste360: What trends are growing right now?
Mark Baybutt: Broadly, with robotics, a lot of focus has been on the challenge of manipulation. The holy grail is really being able to have grippers that can manipulate objects with human-like dexterity, solving the “one pick” problem. A lot of focus across the robotics space is giving the robots a pile of items to pick through quickly. Unfortunately, with that level of sophistication comes cost and complexity.
Fortunately for our industry, the material stream lends itself well to a vacuum-based system, which has been our approach, so we’re able to pick and move material in a very efficient way, eliminating the need for complex and pricey grippers.
Waste360: Looking to 2020 and beyond, what advancements in technology is AMP exploring?
Mark Baybutt: We’re certainly continuing our commitment to utilize deep learning and other emerging techniques as well as keeping a finger on the pulse of how to be a thought leader in the space. We’re also developing techniques to increase our pick rates, while preserving our pick success, to get more material off the belt more quickly and accurately.
We’re also finding new ways to present the data that we capture from the neural network’s material classification to the end customers—from material counts over time to visualization of material trends—so they can really start understanding moment by moment plant operation and turn this data into actionable information to run their facilities as efficiently as possible. This breaks down tribal knowledge and provides tools to assess KPIs [key performance indicators] and overall plant operations.
We have developed a dashboard where users can log in and see various time series trends, which show a plot of data over time and the different types of material that pass through the vision and robotic systems. They can then extract this information and export the data to do a more sophisticated data analysis.
We’re currently in a phase of incorporating user feedback from sites that have been using this product and rolling out new features to meet the requests of users. Being a nimble, fast-paced company allows our product development to quickly iterate these new features and quickly deploy to all users.
Waste360: As these technologies get smarter, what will that mean for the waste and recycling industry?
Mark Baybutt: It means that we’ll have the ability to be even more specific about the materials that we are either extracting or leaving in the various streams. As our deployment base increases, we continually gather examples of materials and aggregate to improve our neural network understanding of materials. We anonymize this data and share the knowledge across all facilities.
Imagine, for instance, taking all the sorters and all the knowledge they have learned in terms of the materials and the differences depending on geographic location and rolling that into one brain. We then have the ability to deploy that knowledge to all of our systems so that every facility and every customer can benefit from that information.
That is how we have the most impact—by keeping the systems refreshed with information and updating them to make them more intelligent.
Waste360: How can companies and industry employees prepare for the next generation of technologies?
Mark Baybutt: We can look at how recycling facilities are fundamentally constructed and how these new technologies can best be incorporated and utilized from the ground up. We have found success in being able to deploy our robotic solution where human sorters are—without requiring a retrofit. I think this trend will continue for existing facilities, but as new facilities are being conceived, we must keep an eye toward not only adopting robotics in their current form but exploring other emerging technologies that can have impacts on other areas within a facility. This will create a holistic solution that will make the most efficient facility possible.
Waste360: You mentioned designing facilities for the future. What should companies be focusing on to ensure they can accommodate these technology advancements?
Mark Baybutt: It may sound silly, but one issue we commonly run into is the ability to have decent internet throughout a facility. As these systems become more and more connected, provisioning high-speed connections is essential. The biggest impact for us is our ability to adequately monitor and troubleshoot.
In addition to internet, it will be beneficial to consider how technology will work in harmony as opposed to just having a number of standalone systems. We’re working to partner and scope more connected systems to inform maintenance at a facility level where adjustments may be needed so that all of these systems of equipment and machinery can operate in unison.