How Does Kollmorgen Utilize Data Analytics for AGVs?
The purpose of an AGV system is efficient transportation of goods in a specified environment. The AGV system is a big investment, so it is desirable to have a high degree of utilization as well as a high ratio of deliveries per hour to get a good ROI.
Searching for clues
Kollmorgen measure the performance and behavior of AGV systems, both on a system level and on board individual AGVs. We collect information on motor drives, laser scanners, localization, traffic, and obstacle interference.
By collecting data Kollmorgen finds trends and patterns, which can be used to improve the products and develop the best possible configuration for any given AGV system. Highs and lows in productivity, daily utilization trends etc. is measured, and the data is used to optimize routes, increase throughput and calculate smarter resource utilization. To the end-user, this translates into lower costs and higher revenue.
The collected data can also give important clues to external processes that might disturb AGV performance. For example, pedestrians getting in the way of the AGVs, or manual forklifts driving in areas originally planned primarily for the AGVs.
Leveraging data and creating benefits
Analysis of the collected data helps reveal inefficiencies of in the AGV route network of a particular site. By aggregating data from multiple sites, Kollmorgen system users can identify deviating individuals, models or components in their AGV fleet. The collected data is stored in a SQL database, accessible in NDC8, offering a dashboard with visualizations of system level KPIs to deliver insights. The database is also made available to users to run their own data analysis.
The benefits of Data Analytics and Machine Learning are many for the end customers. Firstly, revenue can be increased by improving the throughput of the AGV system, i.e. moving more goods more quickly. Secondly, costly unexpected downtime is minimized by predictive maintenance. And finally, hyper-automating the configuration of the system contributes to reduced installation time.
Long time AI experience
Kollmorgen has been in the AI business for longer than most companies. The concept of AGVs has always been about teaching robots to solve complex task without human supervision. This is AI in its classical sense. For example, the route optimization that coordinates the AGVs is based on the A* algorithm, which has been a staple of Artificial Intelligence for decades.
Kollmorgen’s programmers develop and implement the AI algorithm, giving the computer explicit, step-by-step, instructions on how it should think. It this kind of AI that is behind our success. Be it disrupting the flexibility of AGVs by pioneering the laser navigation technology in the 1990s, or be it to safely handle a fleet of over 100 vehicles at a single site in the 2000s, AI continues to be the backbone of the technology we offer today.
The increased buzz around Artificial Intelligence in recent years has very much been focused on visual perception. The reason for this is not the progress in the theory of Artificial Intelligence per se, but has more to do with improvements in the processing power of computers. This has enabled the use of complex machine learning models in real world applications, such as object detection. From a mobile robot (AGV) perspective, having more ways to perceive and understand the environment allows the robot to make more intelligent decisions. For example, Kollmorgen uses machine learning in the form of neural networks to detect humans and estimate their pose, so that the robot can autonomously determine which way to drive to best avoid the human.
Machine learning is also an important part of data analytics, for example when building statistical models to understand what is normal and abnormal in the daily operations of a facility, or when predicting the remaining lifespan of an AGV component.
Creating a positive circle
Machine learning essentially rests on two pillars: automation and data. Since the two amplify each other, this combination is self-perpetuating. More automation yields more data, and more data powers more competent machine learning models, which in turn enables more automation. This offers an intuitive approach to artificial intelligence, where you can start with a low degree of automation, focusing on simple problems with clear boundaries. At the same time, you are collecting data, enabling more challenging applications. In this manner, the user gets the chance to gradually entrust the AI with increased responsibility as it learns. The increasing degree of automation means that more and more decisions are taken by the machines, saving time and resources.
About the Author