Table of Contents
Addverb’s AI control assistant reorients AGVs by translating requests from warehouse workers into AGV commands. Intel architecture-based server from Supermicro provides edge AI processing.
The new AGV control AI assistant from Addverb combines the power of AGV maintenance with generative AI and speech-to-text. This translation capability allows the operator to speak in any one of 98 languages and translate those words into commands, hence be understood by the AGV. This allows Addverb’s AGVs to fix the problem quickly with minimal lost time.
As a member of Intel Industry Solution Builders’ Industrial Builders Community, Addverb is working with Intel architecture processor-powered servers from Supermicro to make this voice-controlled industrial robot solution a reality for its Zippy family of sorting robots.
Sorting Robots Tackle Warehouse Jobs
A significant part of the work done in a warehouse or distribution center is sorting packages or products to match with orders and build inventory or to ensure products get sent on the right transport for overnight delivery. Till now, Sorting had been a manual process that is slow, expensive and mistake prone. Warehouse and distribution centers are now embracing automation to streamline their distribution centers, and adding sorting robots is a part of those initiatives.
Robotic Sorter: Zippy
Zippy Family of Sorting Robots
Addverb’s Zippy sorting robots are a family of robotic systems that sort items at very high speeds. The system is complete with obstacle detection and uses grid-based ground markers for direction. Zippy management software allows flexible adding or removing of robots and sorting destinations as package volumes increase and decrease.
The complete Zippy product family includes:
Zippy 6: Sorting robot with a payload capacity of 6 kg
Zippy 10: Sorting robot with a payload capacity of 10 kg
Zippy 25: Sorting robot with a payload capacity of 25 kg
Zippy 40: Sorting robot with a payload capacity of 40 kg
Zippy Tug: Advanced AGV with a payload capacity of 2,000 kg designed for material movement and tugging applications in various industrial settings
Zippy X: AGV with a payload of up to 15 kg that utilizes vertical lifting capabilities for flexible material handling and sortation, and to facilitate better ergonomics in manual operations.
AGV Control AI Assistant for Native Language Management
To help with managing the Zippy sorters, Addverb has developed an AGV maintenance with generative AI assistant based on generative AI large language models (LLM). It is an autonomous communication system for the Zippy robots to minimise downtime and increase productivity. A warehouse worker can communicate with the model in their native language via unstructured speech. The AGV maintenance with generative AI model is smart enough to translate that speech into commands that the Zippy AGV recognises.
Process used to translate verbal requests into AGV commands.
The system uses two large language model (LLM)-based AGV maintenance with generative AI technologies:
Llama 3 is an openly available and free-to-use LLM developed by Meta. It is cost-effective enough for Addverb to deploy it on edge servers for onsite AI inference workloads.
Addverb also utilizes ChatGPT, a commercial LLM for generative AI that resides in the cloud and can be accessed by the Addverb model when performance is needed. For this application, ChatGPT offers reduced time to generate the instructions (tokens) that are sent to the Zippy AGVs
Supermicro Server Provides Edge Computer
Addverb selected the Supermicro IoT SuperServer SYS-111E FWTR high-performance server to run the AGV control system. The SYS-111E-FWTR (see Figure below) is designed to meet the demands of AGV maintenance with generative AI and machine learning workloads. With a single socket design that uses 4th Gen Intel® Xeon® Scalable Processors, this server offers computing power with up to 32 cores (or up to 64 threads). The server supports up to 2TB of DDR5 RAM.
For I/O, the server features 1GbE and 10GbE ports as well as USB ports and a video port. These network connections are critical for handling large datasets and ensuring fast communication between nodes in AI clusters. This level of flexibility is essential for organizations looking to scale their AI infrastructure to meet expanding robotic fleets.
The system has three PCIe 5.0 expansion slots (one PCIe 5.0 x16 LP slot and two PCIe 5.0 x16 FHFL slots) for network devices and accelerator cards. These PCIe expansion slots can support graphics processor units (GPUs), such as the Intel® Data Center GPU Flex Series, for additional AI/LLM processing power.
Front view of Supermicro IoT SuperServer SYS-111E-FWTR server.
Conclusion
The growth in AGV fleets will mean more maintenance issues and more demand for maintenance engineers. But in many cases these issues can be resolved by untrained warehouse staff using Addverb’s AGV maintenance with generative AI assistant that allows anyone to speak into the system and have their words turned into commands that are understood by the AGV.
The system leverages the use of speech-to-text to translate words in many languages so that it can be understood by the Llama 3 and ChatGPT LLMs that turn those requests into commands for the AGVs. To deliver the compute performance needed for the AI models Addverb has turned to Supermicro for an edge server that is based on 4th Gen Intel Xeon Scalable Processors and is optimized for edge AI applications.
Founded in 2016, Addverb offers complete robotics solutions for warehouse and industrial automation, with a strong global presence through its subsidiaries worldwide. The company provides a range of in-house automation products, including autonomous mobile robots, ASRS, and sorting technologies. It serves over 350+ clients, including well-known companies such as Coca-Cola, Amazon, and DHL
