ACROS

ACROS

» Robots of the future need a brain!

Magazino builds it! «


In order to bring cutting-edge perception capabilities and artificial intelligence into robotics,
it‘s key to have an advanced cooperative robot operating system.

 
 

» Robots of the future need a brain!

Magazino builds it! «


In order to bring cutting-edge perception capabilities and artificial intelligence into robotics, it‘s key to have an advanced cooperative robot operating system.

 
 
 

Advanced Cooperative Robot Operating System

The brain of perception guided robots

With ACROS Magazino is building a software framework, which operates perception guided robots, manages their task execution, coordinates the robot fleet and gathers knowledge through a
cloud based artificial intelligence.


 

Conventional vs. advanced cooperative robotics

Conventional robotics

Characteristics:

repeat  Repetitive

crop_square  PLC-based

av_timer  Fast

lock  Separated



Raising the degrees of freedom results in high costs


 

Advanced cooperative robotics

Characteristics:

shuffle  Dynamic

visibility  Perception driven

cloud  Cloud based

account_box  Safe alongside humans



Increase in the degrees of freedom through perception driven machine learning

 

Conventional vs. advanced cooperative robotics

Conventional robotics

 

Characteristics:

repeat  Repetitive

crop_square  PLC-based

av_timer  Fast

lock  Separated

Impact:

done  High precision and speed

done  Cost effective automation of repetitive tasks

close  Cannot adapt to changing environments

close  No perception and mobility capabilities

close  Unable to handle uncertainty



Raising the degrees of freedom results in high costs


 
 

Advanced cooperative robotics

 

Characteristics:

shuffle  Dynamic

visibility  Perception driven

cloud  Cloud based

account_box  Safe alongside humans

Impact:

done  Fully autonomous and safe mobility

done  Automation of manual tasks in complex environments

done  Perception-driven adaptability to changing situations

done  Ability to handle uncertainty with safe behavior

done  Applicability of Artificial Intelligence to enable advanced self-learning robots



Increase in the degrees of freedom through perception driven machine learning


 

Key elements of advanced cooperative robotics

visibility   Advanced perception

 
 

explore   Manage uncertainty

 
 

school   Adaptive learning technology

 
 
 

Key elements of advanced cooperative robotics

visibility   Advanced perception

Autonomous and safe operation among humans requires superior handling of vast volumes of visual sensor data.

explore   Manage uncertainty

Open and uncertain environments demand close feedback loops to actively raise confidence to a level that allows 24/7 task execution. Cooperative robotic solutions require holistic crowd sourced models of the environment, object, tasks and the robot.

school   Adaptive learning technology

Applying new software approaches leveraging modular and generic architectures, Artificial Intelligence methods, and machine learning to autonomously analyze and react to changing environments.


 

ACROS framework architecture

 

ACROS framework architecture


 

Semantic models inside ACROS

The semantic models inside ACROS abstract away from aspects such as robots, environments, objects and tasks in order to share them with the cloud database. This makes it easy to transfer the knowledge to all robots of the same type, in the same environment, acting on the same objects, or performing the same tasks.

The modular structure of the models itself creates several advantages too: On the one hand, it allows to easily exchange individual parts, such as the environment model, when adapting the robot to a new problem. On the other hand, it keeps the remaining parts stable such that all robots can contribute data and benefit from improvements of the shared models.

Robots become able to work with new hardware components, environments, objects and tasks they‘ve never dealt with before, so their speed and reliability is increasing. The models are shared across the software layers, i.e. the grasping and perception algorithms use the same language for describing their environment and the objects they interact with as the cloud layer does. This way, data created at the lower levels can easily be passed to the Cloud, and models learned in the Cloud can easily be used on the algorithmic level.

Last but not least an efficient programming is enabled by this high level of abstraction. Moving the robots between customers is an easy step.