{"id":3204,"date":"2024-04-18T11:00:10","date_gmt":"2024-04-18T09:00:10","guid":{"rendered":"https:\/\/factorylab.fr\/metramm-measuring-mechanical-interaction-between-humans-and-cobots\/"},"modified":"2025-09-10T11:52:54","modified_gmt":"2025-09-10T09:52:54","slug":"metramm-measuring-mechanical-interaction-between-humans-and-cobots","status":"publish","type":"post","link":"https:\/\/factorylab.fr\/en\/metramm-measuring-mechanical-interaction-between-humans-and-cobots\/","title":{"rendered":"METRAMM: Measuring mechanical interaction between humans and cobots"},"content":{"rendered":"\n

Project conducted by the FactoryLab<\/a> industrial consortium with cross expertise of CEA-List<\/a>, CETIM<\/a> and ENSAM-Paris<\/a> in collaboration with Isybot<\/a>, Naval Group<\/a>, Stellantis<\/a> and Safran<\/a>.<\/p>\n\n

\"\"<\/figure>\n\n

Manually guided machines, from manipulators to cobots<\/strong><\/h2>\n\n

Industrial manipulators are machines that allow operators to move loads without carrying them. They are different from manipulating robots, and are categorised as lifting equipment[1]<\/a>. When they incorporate functions derived from robotics such as virtual guides or software working space limitations, they are known as cobots or IADs (Intelligent Assist Devices). <\/p>\n\n

In industrial robotics, the manual guidance mode makes it easier to teach points by programming; it can also be used in load-carrying assistance applications. The term cobot has also become common for robots that offer such functions. As part of the subject of \u00abPhysical assistance to operators\u00bb<\/a>, the FactoryLab<\/a> consortium supports the development of innovative technologies that address industrial needs where the existing physical assistance resources are not satisfactory, such as the assembly of large parts in aviation (projet MANIPRES<\/a>), or work with tools on cluttered workstations (projet BAR10<\/a>). <\/p>\n\n

\n
\"\"
Cobomanip<\/strong>, Industrial manipulator with assistance functionsPhoto credit: CEA-List <\/em><\/em><\/figcaption><\/figure>\n\n\n\n
\"\"<\/a>
Bar-10<\/a><\/strong>, Tool manipulation assistance cobot<\/em>Photo credit: CEA-List <\/em><\/figcaption><\/figure>\n<\/div>\n\n

Assistance resources that are actually usable<\/strong><\/h2>\n\n

Industrial manipulators are frequently installed in factories, where they remain underused, particularly for medium-sized loads of about a dozen kg: when the load can be transported manually, operators may prefer to work without assistance, which they see as inconvenient or slow, even at the cost of their own health and safety. The issue is relevant in many operating positions of our user manufacturers, who find that loads that are handled repeatedly are sources of MSDs (musculoskeletal disorders) and difficulties to stay at the position. <\/p>\n\n

One of the factors of such underuse is mechanical: while these systems take over the load (weight of the object to carry), they can demand significant effort for setting into motion and controlling movement.<\/p>\n\n

To objectively identify that difficulty, ergonomics specialists can use a dynamometer to measure forces and recommend a level that must not be exceeded[2]<\/a>. These measurements reflect[3]<\/a> the difficulty to control movement only partly and can only be carried out after the event. <\/p>\n\n

When a working position is being designed, questions about the feasibility and sizing of the assistance resources arise: can assistance that compensates for gravity enable compliance with ergonomics limits? How can the performance of assistance resources be specified in order to stay within those limits? <\/p>\n\n

In the case of collaborative robots, robot manufacturers do not currently provide the performance characteristics of their manual guidance modes and comparisons remain subjective[4]<\/a>.<\/p>\n\n

Modelling and measurement of mechanical transparency<\/strong><\/h2>\n\n

An elementary yet instructive model<\/strong><\/h4>\n\n

The concept of transparency was introduced for remote operation with force feedback[5]<\/a>: the aim is to allow the operator to feel the object that is manipulated or the environment touched by the effector as if they were interacting directly. In the case of load-carrying assistance or manual robot guidance, that ideal of transparency would amount to not applying any effort, at the same time retaining the ability to control the movements of the load. <\/p>\n\n

The behaviour of the system to be characterised is an external mechanical behaviour, seen from its grasp by the operator. The concept of mechanical impedance[6]<\/a> is used to define how a system resists movement. Even though it is complex, such behaviour may be characterised locally for a given configuration, along a direction. <\/p>\n\n

For the systems of concern to us, a simple physical model most often allows a description of impedance. It is made up of an apparent mass and two viscous and static friction components. <\/p>\n\n

\"\"
Figure 1: Simple impedance model<\/figcaption><\/figure>\n\n

This model makes it possible to predict the forces in the direction of the required movement, providing the time profile of speed and acceleration are known. How can that profile be known in advance while specifying the assistance resource? In a natural movement of the arm, it has been shown[7]<\/a> that the path is one that minimises jerk (time derivative of acceleration). Based on that property, we build a typical path that corresponds to a movement where the length and duration are specified[8]<\/a>. That path does not reflect inter-personal variability, but the effect of the constraints of the task on the required speeds and accelerations. Based on the typical path, tables can be built to define the parameters of the impedance model that will allow compliance with ergonomic force limits. <\/p>\n\n

\"\"
Figure 2: Construction of a table to reflect model parameter requirements<\/figcaption><\/figure>\n\n

Measuring devices<\/strong><\/h4>\n\n

We propose the use of an external device that will stress the moving system and measure the reaction forces. Stress is applied in displacement along a direction. The measured force is the interaction force. <\/p>\n\n

\"\"
Figure 3: Principle of impedance measurement<\/figcaption><\/figure>\n\n

Three external measurement systems have been proposed. The first two apply rectilinear movement through a linear rail; in the third, the movement is not constrained: <\/p>\n\n

    \n
  1. The first one is motorised and makes it possible to impose excitation paths. The movement is measured by a motor encoder; <\/li>\n\n\n\n
  2. With the second device, movement is imposed manually along the rail and measured by a magnetic linear encoder; <\/li>\n\n\n\n
  3. The third one is mobile and the movement given by the operator is measured by an inertial unit.<\/li>\n<\/ol>\n\n
    \n
    \"\"<\/figure>\n\n\n\n
    \"\"<\/figure>\n\n\n\n
    \"\"<\/figure>\n<\/div>\n\n

    Figure 4: The three prototype measurement devices<\/em>Photo credits: CEA-List<\/em><\/p>\n\n

    We have used a six-axis force sensor to assess not just the force in the direction of the movement, but also lateral forces.<\/p>\n\n

    In the case of the motorised device, we uses paths of different types: speed stages that make it possible to separately identify the friction model, wobble sinus and minimum-jerk paths. In the case of manually activated devices, the recommendation is to combine tests with the paths naturally used by the operator and excitement tests[9]<\/a>. <\/p>\n\n

    The measured data were filtered[10]<\/a> and the simple model was identifiable by least squares. Indicators were calculated to define the quality of identification (identifiability of parameters and difference between measurement and model). In manual tests, tests were combined with paths of different types to obtain robust identification. <\/p>\n\n

    \"\"
    Figure 5: Correspondence between measured force and force calculated by the identified model<\/figcaption><\/figure>\n\n

    Practical test of measuring equipment<\/strong><\/h2>\n\n

    Identification of a physical model<\/strong><\/h4>\n\n

    A model was identified that connects force and the movement in the direction of movement. We were able to test identification in five different systems: <\/p>\n\n