What’s a Cobot?
A cobot or collaborative robot is a robot for direct physical interaction with a human operator, within a shared workspace. Cobots were invented by Northwestern University professors J. Edward Colgate and Michael Peshkin in 1996. Cobots were intitially called “programmable constraint machines”, highlighting their passive and safe method for allowing a computer to create a constraint surface for a human user (and optionally a payload) to follow.
Collaborative robots are complex machines meant to operate side by side with humans and assist the latter in the course of a working process reducing his amount of work. The key to taking full advantage of the contribution of Robotics in production processes, and thus increase the efficiency of the latter, it is making the interaction between human operators and robots as natural and safe as possible. In this way human operators may entrust all those processes that require a too complex automation, while the robots would take care of those tasks that require faster execution and high quality standards
To make this production strategy, we must identify and design methodologies of control and supervision that makes natural and harmless human interaction with robots. The approaches used to try to resolve human-robot cooperation issues are essentially two:
- passive safety;
- active safety.
All the methodologies designed to increase the passive safety in the interaction of humans with robots are essentially intended to modify the structure and functioning of the latter, so as to reduce the likelihood of accidents and their severity. Robots designed by following this policy are characterized by lightweight and flexibility so as to minimize the damage caused by a possible impact on a human being, and for this reason are covered with lightweight materials. In addition, many robots are externally coated with soft materials and have no sharp edges to avoid injures to the operator.
A further project methodology aims to reduce inertia and mass of the moving parts of a manipulator. This technique, often used in industry, is called Distributed Macro-Mini actuation (DM). It consists of a different distribution of the actuators of the robot. All the heavier parts, like engines, are placed near the base of the robot so then, to transmit the motion to joints, special cables and pulleys are used. Specifically, this technique consists in breaking down the actuation of robots between two actuators connected in parallel and arranged in different areas of the manipulator. The high frequency components of the actuation is provided by small motors with low inertia directly connected to the joints of the manipulator. The low-frequency components are generated from a set of heavier actuators placed at the base of the robot. This approach is shown in Figure 1.
Another technique aimed at improving intrinsic safety to robotic manipulators is to use so-called variable impedance actuators or VIA (Variable Impedance Actuator). These devices are capable of varying their mechanical impedance, i.e. the relationship between couples applied to a point and its motion, depending on the phase of work. The practical effect will be to have, during high speed machining processes, a modified impedance that reduces the stiffness and damping of the manipulator, so as to minimize the damage caused by a possible impact. At low speeds, the impedance change will cause an increase of stiffness and damping, in order to get a better response when accelerating links and a reduced swing during braking.
Active safety is the development of specific control strategies of the manipulator who, integrating some dedicated sensors, allows you to exploit the constant monitoring of what surrounds the robot to dynamically change the behavior of the manipulator in potentially risky situations: the approaching of an operator during the execution of a task or the changing of a tool. Research on active safety is particularly interesting for industry because, with the gradual development of increasingly sophisticated and cheap sensors and advanced control algorithms, this will exploit with greater efficiency and reliability the presence of robots in industries.
The exteroceptive sensors, i.e. measuring external quantity, normally used in algorithms for security are of two kinds.
- Sensors used to reconstruct the geometry of the environment surrounding the robot (time-of-flight cameras,laser scanner, proximity switches, etc.).
- Sensors that allow recognition of a collision with an external body (force sensors).
The use of such sensors are manifold and vary depending on the desired control strategy. Here we will focus on three of them:
- The supervision of the robotic cell;
- The use of kinematically redundant robots;
- The use of additional sensors for the development of active control strategies in real time.
The supervision of the robotic cell is to continually monitor the environment surrounding the robot by cameras (common surveillance cameras) arranged in such a way as to frame the robot itself and a suitably sized area around it. Next, thanks to sophisticated image processing algorithms and cognitive vision, the supervisory system will be able to identify any people who enter the scene and follow his path. Additional algorithms, based both on statistical classification of trajectories commonly followed by man in his walk and on walking patterns inspired by neuro sciences concepts, will be able to predict the motion of the person in his successive moves. This prediction will help to form an estimate of the intention and willingness of the framed person to interact with the robot in different ways.
A reliable and prompt enough estimate of the intention of the person can help activate, with an appropriate advance, an active security strategy more in keeping with the situation. However, these prediction techniques are not fully consolidated and reliable because the human being is dynamic and unpredictable and it is therefore hard to anticipate his intentions.
With regard to the use of kinematically redundant robots, this means using robots with additional degrees of freedom to those strictly necessary to perform a specific job in space. The presence of an additional degree of freedom over the traditional 6-axes robots, provides an opportunity to perform the same task in different ways increasing the dexterity of the manipulator arm. Thus, the robot’s programmed motion would suit during the production cycle to allow cooperation with human operators, without altering the assigned task.
Motion planning of the robotic arm is to study redundancy resolution strategies so as to make it as natural as possible the co-existence between humans and robots. In this case though, trajectory planning and control strategies, proven for traditional robots, get complicated considerably. The kinematic inversion problem is much more complicated, both conceptually and algorithmicly. Moreover, the adoption of kinematically redundant robot can pose significant problems from the viewpoint of predictability of the movement. In fact, depending on motion planning strategy, namely the choice of algorithm to translate a given task, programmed at cartesian level, into appropriate trajectories of individual joints, the robot’s behavior can be unpredictable.
Finally, active control strategies based on the use of exteroceptive sensors, are the use of innovative short range sensors that can detect the presence of any unexpected objects in the workspace of the robot arm, including parts of the body of any human operator present in the same workplace. Through these sensors, a dedicated security controller will get, in real time and with adequate sampling frequency, measurements of distance from obstacles that may be used in closed loop control models aimed at reducing the risk of collisions.
Additional safety strategies
Another active safety method is based on 3D recognition of obstacles or people and their approximation through elementary geometric shapes. In particular, man and the robot are approximated as a sum of spheres.
This is done to facilitate the speed of calculation of the distance between the two. Subsequently, based on a prediction of human movement, a special algorithm aims to calculate the ideal trajectory that the manipulator must adopt in order to achieve the final position, keeping a certain safe distance from the human operator.
An additional control methodology that is worth remembering is based on the concept of virtual impedance. This is to construct a virtual surface around the identified obstacle. The robot is then applied a force that opposes the motion, proportional to the depth of the crossing surface. Virtual force is then translated into antagonistic pairs applied to the joints of the manipulator, so to force the robot to move away from that area.