Lourenco Castro, AI developer at Mobile Industrial Robots (MiR) and Andreas Boedenauer, CEO of AGILOX North America gave separate but related presentations on AI and autonomous mobile robots (AMRs) during RIA’s 2020 Autonomous Mobile Robot Conference.
AI advancements enable the robotics industry to take hold of new concepts of vision, machine learning and independent fleet management. Castro discussed the advancements of MiR’s AI camera, which offers object detection and classification. With these two camera features, a robot can avoid obstructions and notify the fleet of the obstruction.
“If the camera detects something marked as safe—for instance, a person—our robots can be configured to still attempt to drive through, given that the person should be able to easily avoid the robot,” said Castro.
He discussed that an AI camera has to be customized to the fleet and the facility in which the AMRs operate. To do this, trained algorithms are required, which give the robots several examples of objects and how it should or shouldn’t navigate around them.
Castro also gave his predictions on where the AMR industry is going with the help of AI. Now, AMRs have little to no understanding of their surroundings, but they have a surplus of safety features to prevent crashes and other hazards.
“By adding configured vision capabilities to AMRs, we’re empowering them to understand what they’re seeing and then react accordingly,” he said. “Overall, you’ll be able to map specific behaviors to specific objects, letting your robot make more efficient decisions without requiring your constant input.”
In the future, fleet management systems will automatically adjust robots’ operations and routes based on patterns and unexpected occurrences.
“Fleet management systems will essentially become sensor hubs, processing information from multiple robots and devices,” Castro explained.
He also discussed that AI-enabled robots will be a facility’s first line of support. They have access to troubleshooting guidance, predictive maintenance, heatmaps and current and past diagnostics.
“If you think about it, your robots have most of the data required for reporting and troubleshooting,” he said. “So why shouldn’t you use them for that purpose?”
In his presentation later that day, Andreas Boedenauer echoed several topics Castro touched on in respect to having a complex technology wrapped into an easy-to-use HMI. Boedenauer also discussed the idea of artificial swarm intelligence—mirroring a fraction of natural swarm intelligence we see in bees and birds—in which individually processed data is shared among a swarm (or AMR fleet). That data is then collectively used to complete tasks and prioritize objectives.
In a centralized fleet management system, information is stored on a server, and that server individually tells the robots what to do. An issue with this, as Boedenauer pointed out, is the special skills needed to operate the centralized computing system.
“[AGILOX] cut the cord from the centralized fleet management system server,” said Boedenauer. “We built a union—called the swarm intelligence union.”
In this system, each AMR has its own computing system and inter-process communication (IPC). For example, if a work order is entered, the units would decide which AMR will complete the task. This type of system also has the ability to recover quickly from malfunctions or breakdowns by automatically dividing the workload among the remaining operational units. If the Wi-Fi connection is broken, units would complete their assigned tasks and park until a connection is re-established.
In many places, AI is already enabling AMRs to act together. Advanced cameras, complex vision systems and ever-evolving object detection algorithms will empower the robot systems of the future.