Automated guided vehicles (AGVs) recognize people with the help of artificial intelligence.

8.2.2021
February 8, 2021

Today, AI (artificial intelligence) enables autonomous machines to reliably assess complex scenarios in real time. Data Spree's AI-based image processing software and BECOM's Time-of-Flight(ToF) camera technology work together to demonstrate the possibilities for automated guided vehicles (AGVs).

In logistics in particular, efficiency along the entire process chain is a decisive factor in competition. Reliable and, above all, safe automation of goods transport is crucial for modern and efficient logistics. Today, employees and autonomous systems must be able to work together safely in warehouses and production facilities, and avoiding accidents is a top priority. With artificial intelligence from Data Spree and 3D camera technology from BECOM, autonomous vehicle scan recognize people and obstacles quickly and reliably and react accordingly.

 

3D vision and artificial intelligence make it a valuable helper

The 3D ToF camera generates a depth image and grayscale image that can be used to localize objects and distances to objects in three-dimensional space. The ToF technology offers advantages over other methods of 3D imaging, such as extensive independence from lighting conditions and high robustness to visual variations of the objects. Using the 3D information, the AI reliably detects the presence, position, and size of trained obstacles, such as people in this case. Then, based on the depth information, the three-dimensional position of the obstacle is detected with centimeter accuracy. Furthermore, the AI can be additionally trained according to other obstacles, such as other AGVs, pallets, packages, etc. The training and creation of the AI can be implemented very quickly and without programming effort using Data Spree tools and services, such as Deep Learning DS.

The AI neural network reliably recognizes employees in different situations.

With the help of this AI-based 3D camera approach, a very robust and safe orientation logic for autonomous vehicles for logistics and production (AGVs) can be implemented. The real-time AI execution required for this takes place on the BECOM Multi-Cam HUB, thus providing a fully integrated interaction of imaging, AI logic and communication. Thanks to their compact design, the ToF sensors used are ideally suited for mobile applications and can, through a combination of several sensors, provide a complete 360° all-round view of the environment. Here, the AI can handle a wide variety of scenarios and situations, since the logic is based on real existing data and is not freely programmed by hand. With the software tools of Data Spree, the AI is also flexibly extendable at any time and can be trained on new definable objects or classes. Thus, complex orientation patterns for AGVs can be implemented, which are not conceivable with conventional methods.

 

Traditional algorithms - complex, inflexible and expensive

Traditional image processing requires very complex programming from scratch. Algorithms are developed by hand by experts, which requires a lot of know-how and time. At the same time, these classically programmed algorithms function according to hand-crafted rules and decision criteria. If deviations from these rules occur during operation, risks arise. These deviations can be, for example, changing lighting conditions, environmental parameters, changes in paths, workflows or processes. Artificial intelligence can correctly assess a wide variety of situations and environmental changes and react correctly by training with real environmental data.

 

Why logistics must use artificial intelligence (AI) in the future

With AI-based image processing, as shown in the example, autonomous vehicles can better interpret complex work situations and processes and make human-machine interaction safer. Depending on the training, AI networks can be taught to different obstacles and objects, such as packages, pallets, high shelves, containers, etc. This opens up limitless possibilities for autonomous vehicles, but also for collaborative robots or automation in general. Smart processes automated with AI will be a key element in modern logistics in the future. Thanks to artificial intelligence, the many lines of programming code are history for the most part. Automated processes for transport, counting tasks or sorting can be implemented quickly and reliably with AI-based image processing and with the corresponding tools from Data Spree. An insight into this new type of image processing is therefore particularly worthwhile in logistics, where time and costs are the decisive factors for success.