Research program:
Self-Actualization and Well-Being:
Wearable Hybrid Robotic Suits

We take many daily activities for granted – walking with a spouse or friend, playing with children, shopping in a store, using a bathroom on time and independently – until we have reduced mobility. Most importantly, reduced mobility leads to degenerative fitness and health problems. This program focuses on the challenges of improving mobility due to its potential for high societal impact.

While many parts of the world are experiencing rapid growth of younger populations, other, more developed regions are experiencing aging populations and population declines. There is both an economic and societal impact due to aging populations. The cost of social welfare programs is dependent on contributions from the current workforce, and as the population grows older, these programs must realize transformative approaches in cost savings to reduce spending in providing care, allowing people to live at home longer, and to keep people in the workforce longer. Aging populations can impact traditional family structures, which also may need to adapt to accommodate these trends via assistive or personal care technologies to support and encourage active aging populations within the home. In the workplace, aging populations may add additional considerations in terms of what kinds of tasks result in injury and disability such as repetitive tasks, heavy lifting, and/or prolonged sitting. New orthotic innovations that are adaptive and work transparently with the user throughout the day will be required. With regard to the economics, impact on family structures, and workplace needs, wearable, assistive robotic solutions can address these issues with cost and scalability that will allow more people to work later in life and live more independently as they age.

This program seeks to develop a novel design and development workflow for the creation of hybrid human/robot systems in which wearable systems can be designed quickly and deployed in such a way that the wearable device learns through a process of continual, mutual learning between human and machine, rather than through more traditional approaches involving extensive controls programming. This approach is dependent on (1) understanding what activities would be conducive for customized and tunable assistance, (2) design, development, and prototyping of active assistive devices using innovations in materials and manufacturing to address the deficiencies of products currently on the market, and (3) the integration of sensing and actuation with an AI-based approach towards mutual adaptive learning to improve wearer efficacy and address issues in scalability.

We propose a wearable hybrid robotic system that assists, enhances, and augments a person in their daily activities around the home and in the workplace in order to improve quality of life, increase productivity, and prolong independent living. Our approach focuses on three key research activities in order to achieve a wearable, hybrid system that works with its user to provide (a) alternate load pathways, (b) reconfigure itself for different activities, and (c) learn alongside its wearer to improve usability.

We first propose a biomechanics-based investigation of routine activities often seen in the workplace and in daily living to identify key opportunities for intervention.  This approach will consider factors that lead to slips, trips, and falls during walking and other transitions, identify key differences in the movement of elderly individuals vs younger counterparts when avoiding obstacles, and propose and investigate interventions that can improve balance and coordination.

Second, we propose a low-cost and highly customizable design approach in which arrays of passive devices may be engaged or disengaged to provide dynamic and low-power support to the user as they walk or more generally transition between different postures or gaits.  We envision using origami-inspired mechanisms fabricated via laminate processes to be key in this endeavor, by facilitating low-power and affordable rigidizing systems with distributed, embedded, multimodal sensing.  The design of this system will be directly informed by what we learn from our initial biomechanics studies.

Finally, we propose to pair this wearable device with a machine learning approach called “mutual adaptation” in order to learn about – and reciprocally guide and train – the robotic device for more effective use and symbiosis with the wearer.  This approach will utilize “predictive biomechanics” to engage the system when it senses that the user or environment is changing.  We will validate the ability of our system to improve stability and performance by evaluating its effectiveness via additional human motion studies.

Research project VI.1: Motion Studies, Benchmarking, and Evaluation

Understanding the activities that are conducive for customized and tunable assistance, and how the resulting designs from our efforts succeed to meet stated needs

The goal of this focus area is to apply experimental techniques in biomechanics to study the motion and performance of human activities both to identify daily activities in which a wearable robotic solution could feasibly help, and to benchmark and study the performance of resulting prototypes against a baseline in order to evaluate performance of our solution and compare against differing designs.

Research project VI.2: Development of Wearable, Adaptive Support Mechanisms

Design, development, and prototyping of active assistive devices using innovations in materials and manufacturing to address the deficiencies of products currently on the market.

The goal of this focus area is to develop low-cost and highly customizable wearable technology that can be tailored to joints in the human body. These devices must be able to engage or disengage quickly in order to provide support and assistance to the user; or to disengage, acting like a passive joint that is transparent to the wearer. One potential outcome of this research is the identification of lower-mass structures that are actively switchable between low and high impedance states to balance the tradeoffs between heavy, rigid exoskeletons and soft exo-suits.

Research VI.3: Machine learning and Artificial Intelligence towards mutual learning and control of Human/Robot Teams

The integration of sensing and actuation with an AI-based approach towards mutual adaptive learning to improve wearer efficacy and address issues in scalability.

The goal of this focus area is to develop and implement a human-safe strategy towards mutual learning and adaptation of human/robot teams, in which the robotic system is worn by an individual performing everyday tasks such as (1) determining how to reconfigure the suit based on human motion and feedback, and (2) training strategies for teaching a human how to use the suit.

The goal of this focus area is also to develop and implement novel machine learning algorithms that allow robots to adapt to the behavioral and biomechanical characteristics of a specific human user or interaction partner, with the ultimate goal to achieve human-robot symbiosis – a tight coupling between humans and robots in which each partner can ‘read’ and predict the other’s intent and act accordingly.

Research personnel

Principal investigators:

Researchers

Publications

  1. Papers in peer-reviewed journals
  1. G. Clark, J. Campbell, M. Drolet, H. Ben Amor, “Learning Predictive Models for Ergonomic Control of Prosthetic Devices”, International Conference on Robot Learning (CoRL 2020)
  2. G. Clark, J. Campbell, S. Mostafa Rezayat Sorkhabadi, W. Zhang, H. Ben Amor, “Predictive modeling of periodic behavior for human-robot symbiotic walking”, IEEE International Conference on Robotics and Automation (ICRA 2020)
  3. G. Clark, H. Ben Amor, “Learning Ergonomic Control in Human-Robot Symbiotic Walking”, Transactions on Robotics (TRO) – In Review
  4. M. Nevisipour, T. Sugar, H. Lee, “Trunk movement control is critical during obstacle avoidance with a cognitive task,” Journal of Applied Biomechanics – In Review
  5. M. Nevisipour, T. Sugar,  and H. Lee, “Trunk Control in Young Healthy Adults Requires Large Adaptations During and After Obstacle Avoidance with a Cognitive Task,” The 45th Annual Meeting of the American Society of Biomechanics (ASB 2021), August 2021, Atlanta (Virtual)

Conferences

  1. G. Clark, J. Campbell, M. Drolet, H. Ben Amor, “Learning Predictive Models for Ergonomic Control of Prosthetic Devices”, International Conference on Robot Learning (CoRL 2020)
  2. M. Nevisipour, T. Sugar,  and H. Lee, “Trunk Control in Young Healthy Adults Requires Large Adaptations During and After Obstacle Avoidance with a Cognitive Task,” The 45th Annual Meeting of the American Society of Biomechanics (ASB 2021), August 2021, Atlanta (Virtual)

Patents pending:

  1. G. Clark, X. Liu,, H. Ben Amor, “Systems and Methods for an Environment-Aware Predictive Modelling Framework for Human-Robot Symbiotic Walking”, U.S. Provisional Patent Application Serial No. 63/210,187

Events: