A digital twin is a virtual representation of telemetry sensed from a remote system and rendered either in real-time or via offline playback.
While digital twins don’t control hardware, they do provide valuable data on actual motion vs. commanded motion. This information is ideally suited for predictive analytics, Artificial Intelligence (AI) training, decision making, Augmented Reality (AR) interaction, and teleoperation.
The potential applications for digital twins are broad and can include operations from part fabrication, assembly, testing, palletizing, robotic manipulation, and much more. When machining a part, a digital twin rendering of actual hardware motion can be used to visualize the machining operations from raw material to a finished part. When using a robot for additive or subtractive manufacturing the system’s repeatability can decrease for several reasons, so a digital twin is an important tool for comparative analysis.
If a manufacturing process is robotic, a digital twin rendering of the robot workcell using the actual motion produced by the hardware can be used to compare and validate the tasked motion sequences. This is important since many automotive workcells are multi-robot systems with robots working in close proximity. To emphasize the extent of the potential applications of digital twins, keep in mind that many unarticulated systems that utilize digital twins might also be manipulated or modified by robotic systems that may themselves utilize digital twins.
The magnitude of the impact and use of digital twin technology is staggering; however, this blog post is specifically focused on digital twins as they relate to robotics.
In the article Inside Toyota's Takaoka #2 Line: The Most Flexible Line in the World, Bertel Schmitt of The Drive points out that the Toyota Takaoka II assembly line does not stop production during retooling operations when a new model is introduced to the line.
The digital twin hype can subside quickly when application-specific context is included.
As manufacturing continues to become more flexible, the associated robotic control and digital twins will need to be able to accommodate changes rapidly.
Advanced real-time adaptive control software, like Actin, provides the ability to model and control any robotic system while automatically updating the virtual model based on telemetry from the robotic hardware in real-time. The captured information can then be used to render the digital twin for representation via AR. Overlaying the digital twin on the rendered environment using the Actin toolkit is accomplished with a simple “RenderVirtual” Application Programming Interface (API) call.
For additional clarity into the use of digital twins in robotics, let’s start with the core question of this post:
The digital twin hype can subside quickly when application-specific context is included. The definition of the term digital twin is not fully fixed in industry, yet it has become ubiquitous so rapidly that, without nuanced considerations of each specific use case, it can misdirect potential users from the fundamental robotics problem of control and instead concentrate on only observation and data collection when users can actually have both. The Actin toolkit provides simulation, control, and digital twins together as, in Actin, simulation and control are bilateral byproducts of each other and performance data can be captured from the digital twin.
That being said, there are many interdependent and typically separate components in the advanced robotic systems that utilize digital twin technology. When planning and managing robotics projects that include digital twins, attention to their synergy with the other elements of a comprehensive real-time adaptive robotics system can expedite deployment and increase value.
The field of robotics is still at a stage where it’s common for a company to find itself seemingly confronted with a make-or-buy scenario for robot simulation, control, or digital twin implementation due to a highly specialized problem/opportunity. However, most software teams with a valuable and/or imminent opportunity prefer the development of a solution using a robust off-the-shelf toolkit over committing company resources to control theory research and development.
..serious robotics opportunities require a robust, comprehensive, and extensible toolkit.
A digital twin is a useful tool for any simulation, control, or teleoperation solution, but it’s worth keeping in mind that a digital twin is actually a byproduct of control and/or external stimuli.
So, where does the digital twin component reside within a comprehensive robot control application like the Actin Software Development Kit (SDK) paradigm?
The following sections will help draw several distinctions among the key elements of a holistic robot control approach.
The statement: “Animation is not Simulation and Simulation is not Control.” is useful to help classify essential aspects of robot control, but there are additional categories that are often overlooked, even by the seasoned Actin user. Let’s start with animation.
In my experience, the most effective engineers are those who can best communicate technical information, and animation is a critical tool for conveying concepts, design intent, and instructions, especially at early project stages. High-value decisions can be heavily influenced by animation -- we believe what we see! However, as stated above, animation is not equivalent to simulation, and therefore does not provide accurate kinematics (geometry of motion for linked rigid bodies) or dynamics (physics of motion modified by forces).
Consider animation to be something like a sequence of digitally rendered images that can be used to communicate concepts effectively. With their highly realistic appearance, plus the ability to locate the viewer precisely where the animator intends, highlight important information, and to modify the direction and time scale of the sequence, animation's value is real, but certainly not without limits.
Simulation is a valuable tool for the creation and validation of motion with the ability to specify real-world properties or experiment with optional materials/properties/forces to optimize or determine requirements or viability, and then save and export performance data.
Among other analyses, kinematic simulations can include the impacts of link lengths, joint configurations, angular limits, and angular speeds for task and workspace analysis. Dynamic simulations can include mass, stress, strain, friction properties, and more. In the case of Actin, this information is provided, along with the capability of exporting the data in real-time.
The following video shows an example real-time digital twin of actual robotic hardware motion visualized as a transparent overlay on tasked motion.
Offline programming is not a new technology, but that doesn’t mean it isn’t valuable in specific fields of robotics use such as robotic metrology, additive or subtractive manufacturing, and more. Offline programming that includes online control goes far beyond creating and documenting a sequence of robot poses. In the context of real-time control Actin users refer to this as offline tasking.
In offline programming, there are often differences between the simulated work cell and the real-world environment. This is problematic and often requires multiple iterations between hardware and software to fine-tune physical hardware registration and waypoints.
Online programming is typically completed using a teach pendant. The pendant is a handheld control panel with a Human Machine Interface (HMI) which may contain information on joint angles and other properties and might include a Graphical User Interface (GUI) with a virtual representation of the system. The pendant is used to control the robot in joint or end effector space, which allows a user to configure and save robot poses for future playback.
The Actin paradigm shown in the graphic above is in an overarching control solution that is best described as a robotics toolkit.
Combining all of the above software components creates a comprehensive model-based control solution with optional simulation rendering and/or digital twin, online or offline tasking with online control via streamed joint commands.
Offline tasking with online control essentially means that Actin programs itself and streams joint angles in real-time based on task goals, environmental models, available Degrees of Freedom (DOFs), with variable degrees of constraint, robot performance limitations, and any other sensed data. To learn more about Actin, visit https://www.energid.com/actin.
The key takeaway from this post is that neither animation nor digital twins make a robot physically move. However, digital twins are useful tools as part of a comprehensive control solution like the Actin robotics toolkit. Used together, digital twins and Actin’s real-time adaptive motion control open new doors to decrease the time to market and increase the overall performance of advanced robotics applications including predictive analytics, AI training, decision making, AR interaction, teleoperation, and more.
To take the next step in applying hardware-agnostic real-time adaptive multi-robot control with a digital twin, contact an Energid Representative to discuss the Actin (SDK) — the most advanced robotics toolkit available.