Implementation of Mobile Robots for an Autonomous Scalable Smart Factory

How did the experience of using mobile robots in Amazon's, Ocado and other's warehouses spur the evolution of transportation systems for autonomous smart, scalable factories?

Speculation was rampant that Amazon was replacing people with robots. But 10 years on, the facts tell a different story. They have more than 520,000 robotic drive units and have added over a million jobs, worldwide. They have more than a dozen other types of robotic systems in facilities around the world, including sort centers and air hubs.

Mobile robots have become an increasingly popular solution for automated warehouses and distribution centers. These robots have the ability to navigate warehouse environments, allowing them to perform tasks such as picking and transporting items, restocking shelves, and more. We all saw videos of hundreds of identical robots operating inside the facilities of OCADO, Amazone, etc. They are very efficient and successful, which stimulates R&D to launch new models, etc.

The possible area of applications in traditional assembly factories for mobile robots is natively narrow and very heterogeneous. On the one side, due to the wide range of conditions: indoor/outdoor, vast variety of mass and sizes, unpacking, loading, etc. On the other hand, the majority of the logistic is performed by conveyer belt. This makes mobile solutions sophisticated and expensive, making them even less attractive to the factories - a vicious circle. As a result, mobile robots, very prospective technology with great theoretical potential, have become niche products. Eventually, we have many small brands with high prices, different APIs, software, requirements to the infrastructure, etc., on the market - "zoo". The technology has no incentives for development and no ability to implement.

Between these two extreme cases, traditional factory and warehouse, there are some key differences:

• In a warehouse, all objects are very typified, with a little controlled variation: packaging, weight, size, etc.

• Environment for movement is very restricted - it is a matrix, in fact

• There are just a few simple types of logistic operations in a warehouse.

  
  

Due to this, mobile robots have become the main transportation system in the warehouse case.

Understanding the essence of fundamental causes of the success of warehouse mobile robots and the problems of ones for assembling factories has helped us, the Robofacturing Mobile robotics team at Arrival UK, develop an effective transportation system for our autonomous smart, scalable factory.

What if we study the success of mobile robots in a warehouse transport system and use successful solutions for a transportation system in an autonomous factory?

Autonomous scalable smart factory

To get rid of the complexity and rigidity of traditional production, we developed a modular,

software-friendly architecture and a new type of factory that should:

• be product agnostic
• have a low entry threshold - affordable for small and medium business
• be not product-defined, but joint types and materials defined (as “3D printer”)
• have substantially shorter time-to-market
• be easily scalable and fast, cheap updatable
• achieve higher utilisation and efficiency by design
• AI-driven
• modular and scalable as Data Center or Data Cloud

How did we do it? Step-by-step story

STEP 1. Change the architecture. Smart Cells

The key block of this factory is Smart Cell with fixed, agnostic, off-the-shelf equipment Chipset, Cell Operation System and API for Plug-and-Play “Applications”- like a blade server in a Data Center. We moved everything related to the uniqueness of assembled products to grippers/jigs/fixtures that move in/out of a Smart Cell and between them. Then packed every Tool, Gripper, Fixture, etc., as an Application with a Plug-and-Play interface and plugged in all needed Applications to every Smart Cell - to complete our “blade servers”:

STEP 2. Scaling Smart Cells and AMR

Smart Cells are unified software-driven cells that can do various typical atomic operations using flexible Artificial Intelligent Applications (AI App). To create a factory, we scale it up (multiple Smart Cells) into Grid, depending on products that we need to assemble (a load for “Data Center”) with many-to-many on the input/output between cells. To complete our “Data Center”, we need to add a data bus, in our case, a transportation system for parts - autonomous mobile robots (AMR):

As you can see, by superseding a physical conveyor with a virtual one, which can be rebuilt in real-time, and typifying cells and operations, we made our factory environment closer to warehousing, thus simplifying it for mobile robots. On the other side, now, our robots must be capable of acting synchronized and independently, working at pretty high speed (to keep a cycle-time of the factory), moving without predefined trajectories and being able to dodge different types of obstacles, including people.

STEP 3. Factory OS and Digital factory create

For orchestrating our AMRs, we developed Factory OS that manages and controls all resources (smart cells, AMRs, etc.) and consists of components such as the Autonomous execution engine - real-time, multi-agent, AI-based system; specialized and high-loaded Data platform; active Scheduler, APL studio - IDE to create all factory rules used by the autonomous engine, etc.

Digital factory create - an asset of data sets and software tools (Digital twin, Application Hub, Playbook, etc.) that provide processes of designing, executing and changing all objects of the autonomous factory. This asset can build itself digitally and run the implementation/rollout of itself.

So, we have simplified the factory's architecture but increased the requirement for our AMRs - intelligence and autonomy. In addition, we are still required to work with different ranges of masses and sizes. To solve these problems, we had to develop our eco-system of mobile robots.

STEP 4. RoboHive

RoboHive is a flexible, modular, reconfigurable, and cost-effective ecosystem of autonomous mobile robots designed to perform various manufacturing and logistics operations inside and outside autonomous scalable smart factories. The hive is made of plug & play active and passive modules similar to “Lego'' bricks; all components inside the robots are modules (e.g. sensors, actuators and payloads can be easily interchanged); robots themselves are collaborative modules that can be combined to create “cluster” systems, which can perform complex tasks in coordination.

RoboHive is a highly versatile ecosystem of mobile robots, adaptable to a range of use cases and product configurations. The robots are primarily designed inside factories for indoor logistics (lightweight/fast and heavyweight/ultrasafe). Other uses included automated (quality/site) inspection and service operations of assembled products (e.g. vehicle automated charging, vehicle batteries replacement, etc.).

Besides the mobile robots (i.e. HiveBots), the team developed robot motors, robot motor controllers, robot electronics and sensors, and a range of software systems for robot control, sensing and perception, simulation, and autonomous navigation.

RoboHive components:

• HiveBots - any AMR part of RoboHive (e.g. WeMo)
• Core Software - onboard & ground AMR control systems
• AMR Elements - сontrollers, motors, actuators, charging system
• RoboSLAM - platform-agnostic AMR vision-based navigation system & spatial AI
• Simulators & Visualisation Systems - single/multi-robot simulators, diagnostic tools, physics simulators, fleet/traffic simulators
• Motor/Motion Controllers - multi-modal state estimators, high-frequency motor controllers, 3D motion planning & control systems
• Infrastructure: power charging systems, network systems, safety systems, development and diagnostics tools, CI/CD

STEP 5. WeMo - Wheeled Mobility

WeMo is a modular autonomous mobile robot primarily designed to cover all the logistics use cases of the smart factory. This is a crucial block of the transportation system - a “working horse”.

• Exteroceptive sensors: sensing the environment around the robot. Interoceptive sensors: sensing the internal conditions of the robot.

• The software stack includes onboard software but also ground, development and testing software, including simulators.

• Omnidirectional wheels

• Max payload: 2,000 kg.

• Max speed: 1 m/s (safe speed 0.5 m/s for humans).

• Positioning accuracy during the cruise: 100 mm.

• Precise positioning: <5 mm.

  
  

Key features of this technology:

• Horizontal Modularity combined with Vertical Modularity: the robots connect to each other to create larger platforms (horizontal); the robots can carry active/passive payloads alone or in cluster (vertical).
• AMR is designed for adding Active modules on top of it (robotic arm, forklift module, mobile charging system, etc.)
• Data and safety comms combined over the safe wireless connection: a truly unique solution in the market. That allows AMR to be added and removed dynamically in the network in a safe way.

STEP 6. Cluster mode

To address the problem of carrying payloads of different sizes and weights, conventional mobile robot fleets have to be made of robots of different sizes and even different types, considering that most of them are designed and produced focusing on one specific area of tasks because companies can't afford a wider range of models and don't have sufficient demand from traditional factories, for the reasons mentioned at the beginning of the article. This leads to a large footprint on the factory floor, issues and costs related to servicing several types of robots (including personnel training), poor sustainability and technology reuse, increased time and cost of integration, higher maintenance cost and poor power efficiency.

WeMos are designed to act together without a physical connection to create larger transport platforms of different sizes (i.e. Cluster Mode) adapting to the dynamic needs of the factory (e.g. one robot to transfer payloads up to 2 tons of small/medium size, multiple robots synchronized and connected by software together to transfer payloads heavier than 2 tons or large in size.)

We developed special software and controllers to create "swarms" of AMRs of different sizes in real-time.

Cluster mode saves up to $7,65mln straightforward savings for a fleet of 150 mobile robots (typical size for a factory with throughput 10k jobs per year )

Open tasks

We launched our first autonomous scalable smart factory with 100+ AMRs fleet, in preproduction mode, in Sep 2022, in Bicester, the UK. The fleet has successfully completed all basic tests and validations and supported the production process at 80% of the target speed (cycle time), including Cluster mode.

However, some critical tasks must be resolved to achieve a fully autonomous capability, better accuracy and higher cost efficiency:

• Even floor - Enabling deploying the factory in warehouses without an ultra-flat floor.

• Robo SLAM - Next-gen software navigation system for HiveBots. It can save $350k per factory.

• Support 5G - Increases safety and reliability of the infrastructure. Allows significantly boosted the amount of data traffic, which will enhance AMR's capability of making sophisticated maneuvers and will give us the ability to use AMRs for comprehensive factory robotic operations: quality inspections, EOL, measurements, etc.

  
  

Summary

As you can see, by changing and loosening towards to each other, the main constraints - decreasing the complexity of the factory environment by using the "Data Center" concept and increasing the capabilities of mobile robots by creating an advanced modular eco-system we created an efficient transportation system that can be as flexible almost as a data bus:

• It doesn't rely on infrastructure and is fully autonomous
• Our AMRs are going to be free to navigate autonomously without path predefinition. This will enable rapid reconfiguration of the whole factory production flow and rapid switch of production of different products in the same site without impacting the layout cells because the AMRs will take care of the logistics changes autonomously.
• AMRs can travel to different places to deliver parts and, if necessary, be re-routed due to blockages etc. So we can produce products with different numbers and sequences of operations in one factory simultaneously.
• It can autonomously adjust to a dynamically changing load, sizes, weights, etc.