Periodic Reporting for period 3 - Ergo-Lean (Rethinking Human Ergonomics in Lean Manufacturing and Service Industry: Towards Adaptive Robots with Anticipatory Behaviors)
Berichtszeitraum: 2022-11-01 bis 2024-04-30
Occupational ergonomics is facing a new complex challenge caused by the adaptation of industrial processes to the demands of the high-mix, low-volume production. In such processes, humans operate in, and interact with dynamically changing environments. The underlying physical interactions can cause variations of human states, and make a traditionally identified ergonomic pose of a human non-efficient and unproductive, or vice versa. This challenge has contributed to the growth of musculoskeletal disorders in manufacturing and service industries undergoing a lean transformation, and calls for new thinking on occupational ergonomics.
Ergo-Lean studies human ergonomics during complex human-robot-environment interactions, and investigate methods to anticipate the effect of worker actions in the short, middle and long term. It explores the potential of collaborative robotics technology to deliver a set of anticipatory behaviours that contribute to the improvement of human psycho-physical states during interaction. Ergo-Lean will create radically new HRC systems where the robot and human directly interact, forming a dyad which optimally solves manufacturing problems in the environment, with the robot flexibly also contributing to ergonomic improvement of workplace conditions.
Ergo-Lean research is articulated along five Scientific Objectives (SO): SO1) Understand and formulate human ergonomics and delivered effort in dynamic human-environment interactions; SO2) Investigate ways of applying the HRC technology to the mitigation of physical occupational risks; SO3) Study the influence of Ergo-Assistant Interfaces on ergonomic coordination of motor redundancy; SO4) Study whether shared authority models can simulate the behaviour of the ergonomic HRC systems; SO5) Challenge and demonstrate the improved flexibility and acceptability of Ergo-Lean HRC systems that enable adaptability to human states and to the variability of the task.
Ergo-Lean studies human ergonomics during complex human-robot-environment interactions, and investigate methods to anticipate the effect of worker actions in the short, middle and long term. It explores the potential of collaborative robotics technology to deliver a set of anticipatory behaviours that contribute to the improvement of human psycho-physical states during interaction. Ergo-Lean will create radically new HRC systems where the robot and human directly interact, forming a dyad which optimally solves manufacturing problems in the environment, with the robot flexibly also contributing to ergonomic improvement of workplace conditions.
Ergo-Lean research is articulated along five Scientific Objectives (SO): SO1) Understand and formulate human ergonomics and delivered effort in dynamic human-environment interactions; SO2) Investigate ways of applying the HRC technology to the mitigation of physical occupational risks; SO3) Study the influence of Ergo-Assistant Interfaces on ergonomic coordination of motor redundancy; SO4) Study whether shared authority models can simulate the behaviour of the ergonomic HRC systems; SO5) Challenge and demonstrate the improved flexibility and acceptability of Ergo-Lean HRC systems that enable adaptability to human states and to the variability of the task.
In summary, the main scientific achievements of the project so far are:
i. A set of online and personalisable tools for the monitoring of the physical and cognitive load of the human in the workplace and for teleoperation systems (Figure 1). This tool can be used for ergonomics validation of industrial workplaces.
ii. An ergonomics-aware framework for task planning and role allocation (Figure 2) in multi-agent hybrid working environments with multi-objective and ergonomics-optimized control of Cobots and Telerobots. These frameworks are essential to make collaborative robots to deliver human-centric and high-performance behaviors for the sake of human ergonomics and productivity.
iii. SUPER-MAN framework, as supernumerary robot bodies for human robot conjoined actions, as a new and alternative technique to wearable assistive systems. The system has minimal wearability constraints and a much larger payload in comparison to the exoskeletons (Figure 3).
iv. Scalable and ergonomic teleoperation systems for multi-robot robot tele-impedance control. It has a wide application potential in industry and healthcare domains.
v. Two open-access datasets, one on human kinodynamic measurements for ergonomics validation and another, an action recognition dataset for industrial human-robot interaction. We also created Open-VICO: An Open-Source Gazebo Toolkit for Multi-Camera-based Skeleton Tracking in Human-Robot Collaboration. Researchers will certainly benefit from these datasets (see Figures 4).
The significant results achieved by the Ergo-Lean project in the first period have additionally led to several scientific talks (see https://ergolean.eu/presentations) in important events, and were covered in important national and international news (https://ergolean.eu/press-and-news). The Ergo-Lean team organised several scientific events (e.g. workshops in top conferences such as ICRA, IROS, and Humanoids).
Ergo-Lean PI, Arash Ajoudani, received the prestigious IEEE Robotics and Automation Society Early Career Award 2021, which was highlighted by the ERC website: https://erc.europa.eu/prizes and https://erc.europa.eu/content/arash-ajoudani-has-won-ieee-ras-early-career-award-2021 as well as on LinkedIn. Two of the Ergo-Lean PhD students have been nominated for best PhD thesis in robotics (one European level and one national).
i. A set of online and personalisable tools for the monitoring of the physical and cognitive load of the human in the workplace and for teleoperation systems (Figure 1). This tool can be used for ergonomics validation of industrial workplaces.
ii. An ergonomics-aware framework for task planning and role allocation (Figure 2) in multi-agent hybrid working environments with multi-objective and ergonomics-optimized control of Cobots and Telerobots. These frameworks are essential to make collaborative robots to deliver human-centric and high-performance behaviors for the sake of human ergonomics and productivity.
iii. SUPER-MAN framework, as supernumerary robot bodies for human robot conjoined actions, as a new and alternative technique to wearable assistive systems. The system has minimal wearability constraints and a much larger payload in comparison to the exoskeletons (Figure 3).
iv. Scalable and ergonomic teleoperation systems for multi-robot robot tele-impedance control. It has a wide application potential in industry and healthcare domains.
v. Two open-access datasets, one on human kinodynamic measurements for ergonomics validation and another, an action recognition dataset for industrial human-robot interaction. We also created Open-VICO: An Open-Source Gazebo Toolkit for Multi-Camera-based Skeleton Tracking in Human-Robot Collaboration. Researchers will certainly benefit from these datasets (see Figures 4).
The significant results achieved by the Ergo-Lean project in the first period have additionally led to several scientific talks (see https://ergolean.eu/presentations) in important events, and were covered in important national and international news (https://ergolean.eu/press-and-news). The Ergo-Lean team organised several scientific events (e.g. workshops in top conferences such as ICRA, IROS, and Humanoids).
Ergo-Lean PI, Arash Ajoudani, received the prestigious IEEE Robotics and Automation Society Early Career Award 2021, which was highlighted by the ERC website: https://erc.europa.eu/prizes and https://erc.europa.eu/content/arash-ajoudani-has-won-ieee-ras-early-career-award-2021 as well as on LinkedIn. Two of the Ergo-Lean PhD students have been nominated for best PhD thesis in robotics (one European level and one national).
A huge step forward compared to the state-of-the-art methods is represented by the set of online tools developed within the project for the monitoring of the physical and cognitive load of the human in the workplace and for teleoperation systems.
Concerning the physical workload, a great majority of the tools currently adopted in industrial environments relies on the so-called ``pen-and-paper'' observational techniques. Most of them analyses a particular aspect or specific activity (e.g. NIOSH: carrying and lifting), however, such a high specificity severely limits their application field. Alternatively, many studies have been conducted by researchers wherein direct measurements collected on the subjects are integrated with models of the human body. Most of the proposed approaches adopt detailed biomechanical models of the human musculoskeletal structure to estimate dynamic states such as joint reactions by using inverse dynamics and then optimization techniques to compute the muscle tensions. Nevertheless, all the models underlying these techniques require the estimation of a large number of parameters, or otherwise, they can be obtained by means of anthropometric standards thus the achieved estimated quantities are not subject-specific.
Concerning instead the mental workload, state-of-the-art methods in cognitive science work offline and/or involve bulky equipment hardly deployable in industrial settings. Generally, cognitive load measurements belong to three main categories: physiological measures, subjective rating scales and performance-based measures.
In the brand-new industrial background, the careful assessment of human ergonomics should be performed via fit-for-industries sensing technologies that require short preparation time as well as advanced yet rapid probing techniques that can provide instant data on human physical and mental load. In view of this, we have conceived a set of indexes to estimate the human psychophysical status that exploits robotics-inspired methods featuring minimum computational cost and thus online capabilities. To address workers’ individual demands, subject-specific and fast re-identifiable models are developed, which only require measurements from easy-to-use and non-intrusive sensor devices (e.g. cameras) to meet the requirements of real factories.
Concerning the physical workload, a great majority of the tools currently adopted in industrial environments relies on the so-called ``pen-and-paper'' observational techniques. Most of them analyses a particular aspect or specific activity (e.g. NIOSH: carrying and lifting), however, such a high specificity severely limits their application field. Alternatively, many studies have been conducted by researchers wherein direct measurements collected on the subjects are integrated with models of the human body. Most of the proposed approaches adopt detailed biomechanical models of the human musculoskeletal structure to estimate dynamic states such as joint reactions by using inverse dynamics and then optimization techniques to compute the muscle tensions. Nevertheless, all the models underlying these techniques require the estimation of a large number of parameters, or otherwise, they can be obtained by means of anthropometric standards thus the achieved estimated quantities are not subject-specific.
Concerning instead the mental workload, state-of-the-art methods in cognitive science work offline and/or involve bulky equipment hardly deployable in industrial settings. Generally, cognitive load measurements belong to three main categories: physiological measures, subjective rating scales and performance-based measures.
In the brand-new industrial background, the careful assessment of human ergonomics should be performed via fit-for-industries sensing technologies that require short preparation time as well as advanced yet rapid probing techniques that can provide instant data on human physical and mental load. In view of this, we have conceived a set of indexes to estimate the human psychophysical status that exploits robotics-inspired methods featuring minimum computational cost and thus online capabilities. To address workers’ individual demands, subject-specific and fast re-identifiable models are developed, which only require measurements from easy-to-use and non-intrusive sensor devices (e.g. cameras) to meet the requirements of real factories.