27th Workshop: Control of Mobile Units in Industrial Environments

• This presentation discusses the state of the art progress in sub-meter accurate localization solutions for challenging industrial environments. It provides an overview of the basic concepts of RF indoor localization, discusses how different technologies compare to each other and finally discusses the realistically attainable performance (accuracy, scalability, coverage) of UWB localization. Several outcomes from localization research projects (LUNAR, nano4sports) are also demonstrated.
  

Company testimonial:
Challenges for autonomous navigation in horticulture and outdoor operation
Andreas De Preter, Research engineer, Octinion

• Autonomously guided vehicles (AGV) are nowadays well established in indoor factories and warehouses. Vehicles in the horticulture industry and outdoor robots (like lawn mowers) are however lagging behind: they are still driven manually or driving around in a random, uncontrolled way. This presentation will provide an overview of the technological challenges for the automation of these applications. A critical difference between the indoor and outdoor industry is the observability and variability of the environment. These characteristics have a serious impact on the accuracy and robustness of the technologies that are currently being used for indoor AGVs.

• This presentation is focused on how to navigate through a factory environment, and how to avoid moving obstacles in the local neighbourhood. The focus of this presentation will be on 2D use cases (mobile robots). Extensions of this methodology to 3D to suit path planning for drones will be presented. Some results of previous and running Flanders Make projects such as Sitcontrol and Avcon will be presented.
  

Company testimonial:
Shared situational awareness and navigation for autonomous vehicles
Peter Staelens, Co-Founder & Technical director, dotOcean
(slides not available, contact us for more info)

• Autonomous vehicles with a digital twin on fog and cloud level are capable of real-time sharing situational awareness. Distributing situational awareness sensors in the vicinity and on the autonomous platforms creates the data necessary for augmented reality for robots and humans. Services on edge, fog and cloud level fuse and render the data into new data products that allow robotic systems to navigate in confined spaces. These services can solve complex optimization problems and allow robots to anticipate rather than to react and avoid obstructions.

• Since WiFi connectivity to mobile units is not always available and/or reliable in large industrial locations, this talk provides an overview of alternative wireless technologies such as BLE long range and sub-GHz that provide larger range and/or lower latencies, and discusses how devices can intelligently switch between these technologies. The talk includes results from the SafeDroneWare & Hycoware project.

Company testimonial:
The fifth network –  the drone network. Industrial scalable drones must be secure connected for safety requirements and social acceptation reasons
Francis Knudde, Director of Business Development, DroneMatrix

• Drones are after the internet revolution, the next new technology going to change existing business models and today’s modern society.  We are facing 3 important challenges to realize this next big technology evolution:
  • Flight time of the drone is a power technology issue.
  • Connected drones give us more safety and control what we absolutely need to deploy the drone technology.
  • Non social acceptance (privacy, sound, …) for drone deployment is a huge showstopper.
• Connected drones are all about wireless communications, their capabilities, the security and the redundancy.

• Real-time object detection and recognition is crucial for many applications of Unmanned Aerial Vehicles (UAVs) such as surveillance, warehouse inventory and infrastructure inspection.  In this talk, we give an overview of the progression of object detection and recognition field.  The talk will cover visual recognition from the early 90’s, to the current state-of-the-art in deep learning applied to object categorization. Some results from the previous projects conducted at Flanders Make will be presented e.g. EMDAS, SitControl, AVCON,…

Company testimonial:
From vision-based AI to real-time flying Robot
Kristof Beenders (Founder) & Michael Wellens (Drone software engineer), Airobot
download slides (members only, login required)

• We will explain our strategy for the implementation of AI algorithms on a small scale mobile device.

• Demo 1: Autonomous drone @ industrial IoT testbed
  During the demo, a drone will fly autonomously throughout a warehouse environment. The drone is localized using UWB technologies and is capable of independently doing stock counts using QR codes.
• Demo 2: Continuous TSCH communication with mobile units
  Existing industrial monitoring solutions such as WirelessHART and ISA100.11a deliver determinism, but do not support mobility well. This demo shows a novel network architecture able to deliver both determinism as well as seamless mobility.
  
• Demo 3: Wireless foot pedal
  LVD’s sheet metal bending machines are pedal operated. However, with the ever growing need for ergonomic design and production efficiency, they were looking into going wireless. A wireless foot pedal eliminates the hassle of vulnerable cables on the floor and allows for a more lightweight design. When the sheet press machines are used in a tandem configuration, several (smaller) machines can work simultaneously, increasing the overall capacity.
  This decision is not without challenges though. First of all, the wireless technology needs to achieve very low latency and it need to be implemented in a redundant design. Additionally, it should provide positioning of the pedal relative to the machine with sufficient accuracy for indoor use. Indeed, when one pedal can be used for several machines, you will want to make sure that the pedal cannot accidentally trigger the adjacent sheet press, potentially causing an unsafe situation.
  The demonstrator will show the use of UWB wireless technology for controlling the latency-critical operation of the press brake machine while at the same time localize the position of the foot-pedal.
• Demo 4: ORCHESTRA – Enabling inter-technology network management
  Modern wireless devices have a plethora of technologies available for communication. Currently, all technologies are managed isolated from each other, which leads to performance and Quality of Services decreases. ORCHESTRA allows intelligent inter-technology management by abstracting technologies and introducing a central SDN based controller. It enables smooth packet-level handovers between technologies, load-balancing to increase throughput, and duplication to increase reliability.
• Demo 5: Deep learning on FPGA, embed AI in your hardware application
  The Easics deep learning demo is based on FPGA technology. The application recognizes, localizes and tracks several classes of persons, animals and objects in real-time on live streaming video. The solution can be embedded in industrial machines, robotics, autonomous vehicles and others that benefit from object recognition, localization and tracking in images or live video.
• Demo 6: Smart high-resolution radars for people and object detection and classification
  The demonstration shows detection and tracking of people and objects using off the shelf radar hardware.
  Signal processing algorithms detect the (non-)movement of people and label tracked targets accordingly.
  For harsh environments where radar alone might not be sufficient, sensor fusion algorithms that combine radar and camera and increase the level of confidence, will be demonstrated.
  download slides (members only, login required)