Robotics technology has created many never-before-seen applications over the last decade. But even now, people still easily expect more than what a robot is actually capable of. To shorten the cognition gap, robots need to be more intelligent, flexible and safe, which is one of the challenges I have been researching and working on for more than 10 years. This article discusses the limitations of existing technology and introduces the concept of adaptive robots, which I believe to be the right direction a robot should evolve towards.
The limits of existing robots
Industrial robots have been used to automate all kinds of production lines for more than 60 years. Traditional industrial robots are designed for fast and precise position control. They are perfect for tasks that can be fully described as a trajectory – for example, moving an object from point A to point B, cutting a circle on a metal part, painting a car shell, etc.
For tasks like these, robots are far more capable than humans in terms of speed and accuracy, thanks to their well-developed hardware and control system.
However, there are three major limitations with these robots:
- Safety — These robots are totally focused on finishing their programmed job, and are ignorant of the potential hazard they pose to people that may be around them. Usually, these industrial robots must be guarded by safety cages while they’re working, but accidents can still happen.
- Deployment — It typically requires robotic application engineers to program the sequence and trajectory of their desired motion in a specific language, so as to connect robots with the production line that they will be working on.
- Limited achievable tasks — By design, they can only complete a limited set of tasks that require only position control and a pre-defined trajectory. There are still many tasks that are too challenging for these robots, such as polishing on complex surfaces, assembling complicated parts with tight tolerance, or interactive tasks in open environments.
With the understanding of these issues, the idea of collaborative robots (also called “cobots”) was brought out in the late 1990s. The idea is certainly trending in recent years, with traditional industrial arm companies like KUKA, ABB, and Fanuc launching their own cobots, and collaborative robot companies like Universal Robots, Rethink, and Franka gaining in popularity. According to MarketsandMarkets research, the collaborative robot market is expected to grow from $710 million in 2018 to $12.3 billion by 2025, at a compound annual growth rate of 50.31% during the forecast period. To overcome the shortcomings of traditional industrial robots, collaborative robots are designed to achieve safety, simplify deployment and programming, and establish collaboration workflow with human beings.
Unfortunately, Rethink Robotics, the pioneering creator of collaborative robots, closed its doors last October (the brand lives on with Hahn Group), instigating a discussion across the industry about the value of cobot. Are they as useful as they are expected to be? Is there a truly scalable product market segment? Do manufacturers really want to pay for them?
To achieve collaboration, cobots typically need to sacrifice performance in specs, including payload, velocity limit, position accuracy, and so on. They still cannot be easily deployed by non-professionals, but rather rely on integrators with more advanced skills. The fact that they need collaboration with humans to get the difficult work done is viewed as undesirable by the manufacturing industry.
So how can we truly take the flexibility and intelligence of robots to the next level?
What defines an adaptive robot?
With these contexts, the next-generation robot must evolve beyond the concept of “collaborative” to tackle the problems at root.
- It should have intrinsic safety and performance without compromise;
- It should learn and accomplish new tasks just like an apprentice;
- It should have the capability to perform tasks that traditional robots are not capable of – there are increasing demands to automate such tasks due to labor shortages and harmful work environments.
A robotic arm that meets all these requirements must greatly adapt to complicated environments and intricate tasks. The next generation robot is therefore defined as “Adaptive Robot”.
“Adaptive” is the word that precisely describes the characteristics of the new generation robot. Traditional robots work without adaptability: parts must be at a fixed position and orientation; no disturbance is allowed while at work.
Having computer vision does add a bit flexibility, but requires high detection precision and well-designed lighting conditions, making deploying and optimizing vision systems on each production line painful and time-consuming. With collision detection implemented, the robot will immediately stop whatever job it is doing instead of adapting to the unmodeled interactions.
Traditional industrial robots are far from being truly intelligent due to the lack of adaptivity. We therefore propose three key requirements that a smarter Adaptive Robot must satisfy:
- High tolerance for position variation — Ability to perform tasks despite uncertainty in positioning (e.g., manufacture or mounting tolerance of workpiece, accumulated position error in production line);
- Great disturbance rejection — Ability to maintain performance, even amid significant changes in the environment (e.g. floating base, vibration, human interference);
- Transferrable intelligence — Ability to handle a wide variation of similar tasks and to support rapid redeployment to new tasks.
How does a robot become adaptive?
The concept of cobot is more like a compromise to the insufficiency of existing hardware and software technology. To achieve the goal of an adaptive robot, fundamental improvements in two key technologies are needed:
1) Force control technology with high accuracy and fast response;
2) Hierarchical intelligence based on vision and force sensing technology.
They are equally important and require bottom-up innovations in every aspect of a robotic system.
Whether you realize it or not, the vast majority of our daily activities rely on our force sensing and force control capabilities. Examples include mopping the floor, pushing a button, or inserting a plug into an electrical outlet.
Conversely, force control has long been absent in robotic arms, making them incapable of performing most force-guided tasks.
Research on robot force control has been going on for more than 30 years, yet this type of technology has still not been broadly applied. It is mainly because existing hardware and software technology of force control has not evolved enough for the cost and long-term reliability to meet the industry standard.
To develop force-control robots with great performance for targeted adaptability, we have to deal with several challenges:
First, existing torque sensor designs are not sufficient in terms of performance and cost. Both joint torque sensors and 6 degrees-of-freedom force/torque sensors need to be redesigned.
Second, each joint assembly should be specially optimized for less coupling effect and better dynamics for force control.
Third, low-level joint torque control should be executed with optimal electronics and well-designed algorithms to achieve ultimate performance.
Finally, an advanced whole-body force control software framework built upon all these improvements and optimizations can then exploit the full potential of a force control arm.
Overcoming this first barrier will get the robot ready with the ability to reply on sense of force when manipulate and interact with objects and the environment, just like how humans are doing in their daily life.
Force control is just the first step towards adaptability. A robot also needs to know how to utilize force control with the integration of other information. This leads us to the concept of hierarchical intelligence, which naturally applies to people’s daily activity.
As an example, when we wipe a window, we first recognize the window glass within the frame, then move our hand with a damp rag back and forth while applying some force perpendicular to the glass to increase rubbing friction. At the same time, we also watch carefully to make sure we have covered everything.
In this window-wiping process, some of our capabilities are used consciously, such as vision recognition and stain detection. Other capabilities, such as moving our arm and applying force, are more encoded in our subconscious.
The human brain, responsible for planning a trajectory based on visual and haptic information of the body, never has to think about the electric signals needed to actuate a specific muscle in order to complete a task.
We believe the intelligence system of a robotic arm should be similar: the AI responsible for recognition of the window and stain should not be directly involved in figuring each joint position or motor current. It should not even worry about the motion primitive of “moving back and forth while applying some force.”
The lowest layer of intelligence controls the basic motion of the arm and maintains stability, the mid-layer intelligence encodes a different sequence of motions, while the highest layer of intelligence takes care of perception, understanding, planning and other complicated cognitive tasks – this is what we call “hierarchical intelligence”.
In a hierarchical intelligence system, each layer of intelligence is relatively independent. Lower level intelligence cannot directly affect higher-level intelligence. The output of higher-level intelligence is executed by the lower level. Each lower level is instructed and tuned by its direct upper level.
In conclusion, an adaptive robot must have a hierarchical intelligent system, either for true intrinsic safety, for its work efficiency and effectiveness, or for good skill transferability. By combining well developed force control abilities and a hierarchical intelligent system, we are closer than ever to creating a truly adaptive robot.
What value will it bring to the industry?
1) Beyond conventional automation
As I mentioned above, restrained by position control, the automation level and task scope of traditional robots are quite limited. Developed in a new methodology that fuses fine force control and vision, the new adaptive robot can accomplish a larger set of tasks. Great tolerance of position error and disturbance rejection also lets it work reliably in a more open, complex and uncertain environment. Many existing automation challenges can therefore be overcome.
2) More flexibility to production
With transferable intelligence, it becomes much easier for enterprises to deploy and adapt to similar scenarios and similar workpieces, or to redeploy robots for new production lines. On the other hand, adaptability of the robot helps simplify the line, by reducing additional devices and customized mechanisms commonly required in traditional automation solutions. Both production flexibility and cost effectiveness will be improved significantly.
Adaptability unleashes the power of AI
Deep learning has been developing rapidly, enabling more sophisticated perception and decision-making ability of a machine than ever before. However, the nature of this methodology always leads to a trade-off between detailed accuracy (e.g. position accuracy of a detected object) and universality (robustness against variation and corner cases). It causes heavy constraints when AI is combined with traditional robots and automation lines, which are highly sensitive to position error. A robot with good adaptability can truly exploit the power of cutting-edge AI technology. Besides, the capability of fine force control provides feeling of touch and better dexterity, which facilitates more space for AI to prosper in the field of robotics.
A better world
People may get very worried when talking about robots taking over human work and leading to unemployment. However, it is worth mentioning that many jobs and the work environments are harmful to the workers physically or mentally, which is the reason why many industries are faced with labor shortages. It is our choice to facilitate the right automation. Ultimately, technology should always serve humans. When tedious and harmful tasks are automated, new opportunities will be created, which can lead us to a better world.