The workforce of the industry of tomorrow 

New workers, new colleagues

Over the years, industry efficiency has been focused on stock optimization and production space. To remedy the evolution of market demand and, consequently, the growth of warehouses and factories (in numbers and surface), we had to reconsider the place of the human workforce in space and the production process. Automated robots are at the center of this transformation of Industry 4.0 by helping the workforce answer this ever-growing demand.

Industrial robots are in charge of different tasks: sorting, manipulating, storing, or transporting objects. These are mainly low value and sometimes dangerous for workers. At Visual Behavior, we improve the collaboration between workers and robots to optimize the efficiency and the productivity of the workflow across the industry. To this end, we believe that we shall understand human behavior, the environment, the workspace, and the objects workers use on a daily basis. On the one hand, those capacities allow robots to act collaboratively alongside workers instead of just sharing the space with them. On the other hand, those behaviors enable robots to do more sophisticated tasks that help the workforce be more efficient.

From automated to autonomous robots

According to IFR, the integration of robots in the industry grew in Europe ‘from 543 000 unity in 2018 to 580 000 unity in 2019′. This constant growth creates opportunities for the industries of tomorrow. Picking and filling robots and AGVs are the most frequently seen solutions in the industry today. These robots execute repetitive and recurring tasks, resulting in more productivity, flexibility, and fewer human errors. They contribute to the workers’ jobs but could not entirely replace them to perform specific, non-generic, and complex tasks. This is why to create a complete collaboration between workers and robots, we need to support the transition from automated robots (i.e., with permanent human assistance or elementary tasks) to autonomous robots with limited human assistance.

Current technologies that participate in this transition rely on costly sensors, high development costs, low adaptability, and limited use-cases. Our artificial vision system based on human vision can solve this problem.

Our technology aims to give robots the capacity of scene understanding to make their own decisions, as humans do. To successfully collaborate with them without a high dependence on their supervision, robot scene understanding is key. Those decisions rely on a set of capabilities we often refer to as common sense for humans. Robots need to have a live description and comprehension of the surrounding scene, including ground detection, navigation possibilities, objects detection, and human behavior. Those capabilities are needed to solve the robot autonomy challenges.

Aloception: Our robotic visual intelligence technology

Artificial visual cortex

At Visual Behavior, we created Aloception, a new generic technology based on artificial vision designed to fill the gap of missing possibilities in the existing market mentioned before. We aim to provide robots with this ‘common sense’ so they can adapt to any real-world situation based on a generic but steadily growing understanding of the space around them. 

The notions of environment perception, objects, and people’s detection come at an early age as humans. This is why child development is the primary source of inspiration for our AI experts who create, test and train our technology to grow and pass the next perception level.

As children make assumptions based on what they learn by themselves and their parents, robots equipped with our technology can evolve in an unstructured environment thanks to semi-supervised learning. This enables it to adapt to new situations and usage without costly and time-consuming hardware or software modifications. We believe that advanced reasoning requires complete access to information provided by a unified multimodal architecture needed by humans to make the best decisions. 

To solve these challenges, Visual Behavior is at a crossroads of different domains: 

  • Artificial intelligence:We work with artificial vision experts working on transformer-based architecture and semi-supervised learning.
  • Robotics: Beyond image understanding, our technology explicitly addresses the robotics challenge that prevents robots from being deployed massively in the industry.
  • Neuroscience: The core of Aloception takes inspiration from the human visual system, it is symbolic reasoning, and its visual learning capabilities to design and structure Aloception’s components.

Simple tools

Sensors are often the predominant cost of a robotic platform (3D, Lidar, Radar, etc.). We praise that a set of required innovations can be more beneficial for autonomy than a comprehensive set of sensors. Emancipating a robot from sensors to let Aloception rely on a standard mono/stereo camera and visual intelligence has many advantages: a large diversity of outputs, an excellent temporal consistency, genericity, and adaptability. And again, all this technological capacity is gathered in one unified and consistent network: an artificial visual cortex.

… for complex situations

To find solutions to more specific cases, we created the Alo program. It is composed of multiple modules from our generic technology Aloception that can be implemented according to the particular needs of our clients. Visual Behavior is at the top of research and science in artificial vision, materializing it in a patent-pending technology to guarantee our product has a high fiability.

AloforAGV: a new specific technology

One use case kept at the heart of our research is ‘AloforAGV’, our product designed for automatic guided vehicles. The modules this product needed by to enable the robots to make clear decisions in unstructured and dynamic environments are:

Human behavior understanding

Alosym

Entity level detection and reasoning

Ground detection

Alopix 

RT Pixel level detection and segmentation

Object detection

Alodepth

RT depth estimation from stereo and monocular cameras

Navigation

Aloflow

RT motion & occlusion estimation from video

The robot understands the scene and keeps the information needed based on a simple sensor, a stereo camera. The acquired video stream is supplied to the system and is processed to answer the essential questions. (What is this? Who am I? Where am I? How do I move?)

Their integration provides a real game-changing solution, opening new use-cases and existing ones with new levels of precision. It allows customers to make a difference by improving navigation accuracy, obstacle avoidance prediction, and human behavior understanding.

Our solution offers the possibility for any automated guided vehicle to move from automation to autonomy.

Robin Jacqueline from Enjoy Automation talks about “AloforAGV” : 

‘The product provided by Visual Behavior is very appreciated for its simplicity and adaptability to any use-cases (ADAS, Picking robot, etc.). Here products can respond to industry-specific requests and automated driving vehicles: from generic to specific demands. The different features respond to your specific needs in efficient ways. We are very convinced by the unified architecture, semi-supervised learning, and real-time with latency awareness.’

Newsletter

We focus on the most important challenges of robotics: the lack of intelligence, often compensated by complex sensors at high prices, and the lack of rich interaction between robots and their environment or humans working alongside them.

‘Aloception’ offers a generic visual solution extended with specific features to answer complex problems such as ‘AloforAGV’. Our Artificial Visual Cortex makes the robot a good collaborator, passing the next levels of perception one by one with a promising and scaling intelligence. This way, customers can create a collaborative robot for a wide variety of tasks.

If you want to know more about the adaptation of our product Aloception in a specific robot or learn more about who we are and understand our artificial intelligence vision technology.

Contact us and we’ll be happy to help you find a solution to make your robots more autonomous.

Contact Us

More articles

Understanding Human Behavior

In this article, we introduce another aspect of our visual system. During the first Three Cup Monte Test for visual perception, we demonstrated the software capabilities to track objects and understand their movements and interaction in the outside world….

View article