ABC Robotics Initiative – Agricultural, Biological & Cognitive - ABC Current Research Development Projects

Show All

Hide All

A Bimanual Robotic Simulator and Data-driven Personalized Training for Optimizing Epidural Analgesia Skill

Ilana Nisky, Biomedical Engineering
Yair Binyamin, Soroka Hospital, Faculty of Health Sciences

The anesthesiologist's case experience is the main cause of the most common complications of epidural analgesia. To enhance patient safety and create a controlled learning environment, we will combine robotics, sensorimotor control theories, and advanced data analysis to improve the skill of novice anesthesiologists. We will develop a bimanual robotic simulator for optimizing training by presenting realistic force feedback to trainees, and allowing for an extensive practice before ever touching the first patient. We will combine 3D printing for rendering rigid anatomically-accurate structures and two robotic devices for rendering modifiable soft tissue resistance and for kinematic data recording. Sensorimotor learning theories will help to optimally harness explicit and implicit learning, and advanced data analysis will help to quantify the progress of trainees. Our aims are: (1) Develop a bimanual simulator for epidural analgesia (2) Develop quantitative metrics of skilled analgesia and an optimal epidural analgesia training protocol. (3) Examine the efficiency of training with the simulator compared to the standard training protocol in place currently.

Adaptive Feedback for a Personalized Robotic Trainer

Simona Bar Haim, Physical Therapy
Yisrael Parmet, Industrial Engineering and Management
Yael Edan, Industrial Engineering and Management

As the world's population is aging, the global shortage of caregivers and trainers is increasing significantly. Integrating social assistive robots into older adults' lives can improve their quality of life and fulfill this shortage. However, developing such robots, that one would accept for long-term use, requires several developments. The robots should be personalized and adaptive to the changing capability and needs of different users. Human-robot interaction, and specifically feedback, have proven to take a meaningful part in technology acceptance among users.

This Ph.D. research aims to develop an adaptive personalized robotic trainer. The training is focused on upper body exercises. The research is divided into four stages, the first stage is the development of the robotic trainer system and its evaluation to detect personalization preferences depending on age groups. The second stage is focused on adding an emotion recognition ability to a robotic trainer and evaluating the adaption of the robot to the changing emotions and comparing personalization features regarding cultural differences. The research's third stage is focused on motion recognition of the users: developing classification algorithms to detect specific performance problems during the training sessions and adapting the training program to try and improve failure motions (e.g., detect left arm problems and provide focused training for left arm). The fourth and final stage is to improve the adaption by detecting the user's motor learning stage; This research aims to examine if it is possible to identify one's motor learning stage and how this can be identified. For this purpose, the motion pattern of users will be analyzed using advanced feature extraction and classification algorithms and feedback will be implemented accordingly.

In each stage, the robotic trainer will be evaluated and assessed by technology acceptance measures that will include both subjective and objective measures. The proposed research is expected to contribute to several aspects of human-robot interaction research by system and algorithm developments, social assistive robots design guidelines, and proving the importance of developing adaptive and personalized robotic systems.

Aquaculture Monitoring Robot

Amir Sagi, Life Sciences
David Zarrouk, Mechanical Engineering
Yael Edan, Industrial Engineering and Management

Fish products account for about 16% of the human diet worldwide, as of 2017. The average annual increase in global food fish consumption from 1961 to 2017 was 3.1% - nearly twice the growth rate of world population (1.6%) and more than the growth rate of all other animal protein foods (meat, dairy, milk, etc.) of 2.1%. Within aquaculture, the crustacean sector is the fastest growing at 69$ billion annually (FAO, 2018). In the process of growing and producing crustaceans, the counting action is a significant component. Accurate counting is essential for feeding, marketing and sales. Manual counting is time-consuming and inaccurate (Ibrahin et al., 2017). This proposal aims to develop a robot to monitor crustaceans in industrial aquaculture systems for improved management operations focusing on the counting operation due to its importance for automatic and precise aquaculture (Ibrahin et al., 2017) and significant reduction in expenditures.

Automatic electrical wiring process planning for robotic manufacturing systems

Sigal Berman, Industrial Engineering & Management

Connecting multiple electrical wires is a major manufacturing challenge encountered in multiple products and applications. Due to the complexity and variability of the design, most electrical wiring tasks are currently done by human operators. Recent advances in robotic manipulation of deformable objects facilitate changing this situation. However, robotic wiring process planning is complex. Automatically determining optimal wiring order and wiring paths suitable for robotic production processes are important for reducing manufacturing setup times, increasing robustness, and for reducing wiring errors. These advantages are critical for facilitating widespread implementation of robotic wiring in small batch-size facilities. The research will advance the state of the art in design for robotic wiring, through integration of formal mathematical methods with automatic data extraction from production documents.

Bio-hybrid Artificial Muscle for Soft Intelligent Robots

Aslan Miriyev, Mechanical Engineering

Mirko Kovač, Materials and Technology Center of Robotics, Empa, Switzerland

Soft actuation is deemed one of the central and longstanding gaps to achieve soft collaborative ro-bots. In recent years, the trend in soft actuation gradually shifted from pneumatic devices to func-tional materials. However, soft actuators still suffer from relatively low durability, which is reflected in a small number of actuation cycles. In recent years, we began working on understanding the effect of chemical triggering on muscle fiber contraction for further implementation in soft bio-hybrid ro-botics. We demonstrated the feasibility of using chemical stimuli to control muscle contraction in vitro. However, the chemical-based control of muscle fibers required a minute-scale duration for actuation and de-actuation, rendering the actuation speed and number of cycles challenging to utilize in the context of soft collaborative robots. To address this issue, here we propose to combine chem-ical stimulus with optogenetics, a technique that utilizes light to induce muscle fiber contraction, resulting in a sub-second response, and implement it in a bio-hybrid artificial muscle. We can po-tentially control a muscle fiber contraction with adequate response speeds by employing such a hy-brid approach. The proposed research activities include (i) the development of bio-fiber-based 3D-bioconstruct capable of hybrid chemical/optogenetics-triggered actuation, (ii) integrating the con-trollable 3D-bioconstruct in a synthetic robotic environment, and (iii) showcasing a new type of bio-hybrid artificial muscle in a representative robotic device and subsequent performance characteriza-tion (actual robotic implementation). The demonstrators will be capable of various motion primitives via introducing structural kinematics by actuator design and time-modulated stimulus supply. We suggest that the proposed approach may redefine human-robot interaction and lead to a paradigm shift in robotics.

Building Construction Robots with Frame-based Locomotion

Shabtai Isaac, Civil and Environmental Engineering

Thomas Linner, Chair of Building Realization and Robotics, Technical University of Munich

Buildings are among the most complex products created by mankind, in terms of the sheer number of types of components they contain, and the complex dependencies between those components. On the other hand, buildings contain many features that are similar to each other, and their construction involves numerous repetitive activities. Yet so far, the use of robotics in the building industry to improve its efficiency has been relatively limited. The development of robots for construction activities requires, among other things, a solution to the question of how such robots will move in a highly dynamic and spatially complex environment. Currently, a building is constructed by teams of workers who pass through it and around it and carry out a series of repetitive successive actions at different locations. At each location, the team adds a certain component to the product, or processes the existing product. Efforts to replace those teams of human workers with robots in a cost-effective way have been hampered by the size and technical complexity of buildings, which make it impossible to implement the assembly line concept at a scale that exceeds the offsite manufacture of a relatively small building components. Consequently, construction robotics research has often focused on the effort to replace workers with robots that have locomotion capabilities (Bock & Linner 2016). Yet, this is challenging due to the spatial complexity of buildings, in terms of the rooms and corridors of infinitely varying shapes that they typically contain, situated at different levels and continuously evolving while the building is being constructed. This makes the development of a robot able to move through such a spatially complex landscape extremely difficult and expensive, in particular since such a robot also needs to be heavy enough so that it is capable of executing activities involving significant loads.

Developing an Understandable Robot: Combing learning and natural language processing

Yael Edan, Industrial Engineering and Management
Suna Bensch, Computer Science UMEA
Thomas Hellstrom, Computer Science UMEA

This work aims to develop novel techniques for understandable robots, by a framework that combines learning and natural language processing. We intend to develop a method to enable robots to learn the right utterances during an interaction in a collaborative task thereby leading to create a human mental model. Furthermore, the robot will explain to the user its doings. We premise the explanation will be based on three questions i.e., what needs to be explained, when should it be explained and why an action is being taken by the robot. With these three questions, we will focus on developing a model for the explanations which includes clarity, a pattern in which explanations are being communicated to the user, and justification for a particular option. These three questions will provide a basis for defining different levels of understanding (corresponding to levels of automation). This study would also employ the combination of verbal and non-verbal communication which would increase the modalities of the communicating agents.

Not knowing the prior information about the action of the robot in human-robot collaboration has negative impact on user perception. By creating a human mental model, we will increase the understandability of the robot's actions leading to improved human-robot collaboration.

Specifically, we propose a framework in which the mental model of the human contained in the robot's state of mind (Mr) would act as a state. A reward function would be derived from the human response. The robot would generate explanation for its task according to the current state i.e., the human's mental model. If there is a mismatch between the current state of the human's mental model and the user's response, the framework (implemented in the robotic system) would move toward the next state and generate another explanation with another complexity. This way the framework would incorporate a reinforcement learning algorithm to generate the optimum natural language utterance for better understanding in human-robot collaboration.

Evaluation will be conducted through a series of user studies performed on different robotic platforms and tasks. The key performance indicators will include satisfaction, trust, curiosity, fluency, the 'goodness' of the explanation and additional measures which will be assessed through objective and subjective measures. We aim to define and evaluate the quality of understanding.

This proposed work paves the way to developing robot understandability for nonprofessional users and bystanders, by combining learning and natural language processing. This is expected to increase trust and acceptability of robots that collaborate with humans, which is especially important for these populations.

Hellström, T., S. Bensch. 2018. Understandable Robots - What, Why, and How, Paladyn - Journal of Behavioural Robotics 9 (1): 110-123.

Singh, A.K., N. Baranwal, K.-F. Richter, T. Hellström, S. Bensch. 2019. Verbal explanations by collaborating robot teams. Paladyn. Journal of Behavioural Robotics.

Development and piloting of an Elliptical Perturbation Robotic System that provides unexpected perturbations during in place walking (the EPES robotic system)

Amir Shapiro, Mechanical Engineering
Itshak Melzer, Physical Therapy

Gait and balance impairments may increase the risk of falls, the leading cause of accidental death in older adults. Overall, fall-related injuries constitute a serious public health problem associated with high costs for society as well as human suffering. The scope of this problem will continue to expand as the number of elderly individuals is projected to increase dramatically over the next 25 years. Consequently, there is a general need for developing cost-effective interventions that can prevent the occurrence of falls, especially in the elderly population. The Elliptical Perturbation Robotic System project proposal (EPES) are motivated by the fact that older adults can improve their balance by an unexpected perturbation training (PBBT) method. PBBT was found to be a very effective and promising training approach. Recent data suggests that the PBBT reduces the rates of falls by 46% among older adults. The main innovation of the EPES project is providing Unexpected perturbations during elliptical training (i.e., walking in place) that triggers automatic reflex-like balance responses that act to recover equilibrium. Such balance responses are not under volitional control and cannot be trained and improved through voluntary exercises.

The traditional PBBT methods evoke reactive balance responses during treadmill walking or standing, i.e., stepping responses. However, many older adults suffer from Osteoarthrosis of the lower limb joints, which may reduce their ability to benefit from this effective PBBT method. The PBBT during walking or standing may expose the trainee to high-ground reactions impact forces when reactive stepping is triggered. The basic idea of this project proposal is to design and build a perturbation training system that provides unexpected perturbations while walking in place on an elliptical system to perform perturbation training during walking with less impact ground reaction forces.

Improved Methods for Indoor Navigation in Crowded Environments

Ronen Brafman, Computer Science

Or Tslil, Elbit
Tal Feiner, Elbit

Elias Goldsztein, BGU

Our work seeks to improve existing navigation algorithms by learning to optimize their performance as a function of the current state of the environment. Specifically, we will learn to automatically tune the parameters of the well-known and widely DWA local navigation algorithm. DWA is well known and well understand, but it has a number of hyper-parameters. Their choice affects the quality of the local navigator. We seek to learn environment features and mappings from environment features to good parameter choices. By building on a well understood algorithm we ensure explainability and predictable behavior. By automatically learning environment features and computing optimal parameter values for DWA we enable fast adaptation to diverse environments.

Keep Healthy and Active Ageing: A comprehensive approach linking early risk detection of disability, fall prevention and intervention in older adults

Itzik Melzer, Physical Therapy

Amir Shapiro, Mechanical Engineering

Our project will cope with risk of frailty, disability, fall and correlated cognitive impairment in older adults that may gain huge advantages by timely detection of changes in lifestyle, gait and personal behavior in daily life to trigger suitable clinical assessment & interventions program. To accomplish this objective, Keep Healthy & Active Ageing will develop an advanced model of long-term activity monitoring at older adults' home i.e., indoor monitoring, using Lidar system. In case of risk detection, a devoted clinical examination of the balance function will be performed using the TEMI robot as a IPAs Tele-robotic approach at their homes. In addition, to the robotic assessments of balance function, Keep Healthy & Active Ageing will manage the administration of adequate, personalized and timely Tele-robotic-interventions performed by the TEMI robot, for an active and healthy living- mainly targeting balance function to implement an effective prevention program addressing the main risks of the older life such as frailty, falls, physical and cognitive decline.

Pest Control Using UV-B Irradiation

Guy Shani, Software & Information Systems Engineering
Vered Tzin, The French Associates Institute for Agriculture and Biotechnology of Drylands
Shai Arogeti, Mechanical Engineering

Pest control is a major concern in modern agriculture [Goggin, 2007]. It is in
the best interest of farmers, both from the economic as well as the environ-
mental perspective to avoid using chemical pesticides to eliminate pests. Thus,
many researchers focus on developing alternative methods to control the pest
population.
This research will have the following goals:
1. Develop algorithms for the automatic identification of aphids and mites on leaves using RGB images in commercial greenhouse conditions.
2. Develop algorithms for moving a UVB emitter towards and infected leaf using an RGBD sensor.
3. Constructing an autonomous cart carrying a robotic arm that is able to drive through the greenhouse, stop at a specific plant, and irradiate a
specific leaf.

4. Demonstrate the ability of the system in reducing pest damage in commercial greenhouse conditions.

Robots for Parkinson's Disease

Shelly Levy-Tzedek, Physical Therapy
Amir Shapiro, Mechanical Engineering
Yair Zlotnik, Soroka Hospital

Parkinson's disease (PD) is a progressive neurodegenerative disease, often characterized by tremor, slowness of movement and rigidity. The degeneration of neurons in the substantia nigra creates a shortage in the neurotransmitter dopamine, resulting in movement impairments that characterize the disease. These impairments often develop to be highly debilitating, preventing participation in many activities of daily living. For decades, PD has been considered a primarily motor disorder, though at recent times evidence is mounting that cognitive processes are also affected by the disease.

In the US alone, nearly 1.5 million people are thought to suffer from PD, with annual treatment costs approaching $25 billion. Aging of the society will likely lead to a larger prevalence of the disease in the population. At this time, there is no cure for PD.

In this project, we propose to use socially assistive robots as a non-medicinal approach to alleviating some of the symptoms of PD. We will use the participatory-design approach, including all major stakeholders in the process.

Social Robots Domestication in Later Life

Galit Nimrod, Communications
Yael Edan, Industrial Engineering and Management

Social robots have great potential to promote autonomy and wellbeing in later life. As most devices were only recently launched, an in-depth understanding of what makes older adults' interaction with such devices successful is necessary. Domestication theory (Silverstone et al., 1992) is one of the most widely used theories in multidisciplinary studies that explore the integration of technologies in everyday life. The present study is the first to explore the domestication of mobile social robots by older individuals.

The study was designed to answer the following questions:

What uses, outcomes, and constraints characterize older adults' domestication of mobile social robots?
Do the uses, outcomes, and constraints change during the domestication process?
To what extent does the domestication of mobile social robots in later life differ from the domestication of stationary Intelligent Personal Assistants (IPAs) and why?

Study participants will include 20 community-dwelling women ages 75 years and over. Each participant will be given a TEMI robot for three months. Study participants' uses of the social robot, the benefits they will derive from usage, and the constraints they will face will be measured using various methods. These will include pre and post research period questionnaires and in-depth interviews, observations on the participants' use, and weekly surveys.

This study replicates our research on stationary IPAs (Nimrod & Edan, 2021). Therefore, it will enable comparing the domestication of an IPA with that of a social robot, which differs by the fact that it is mobile. Accordingly, the study is expected to yield critical theoretical contributions as well as practical implications for robotics R&D to promote wellbeing in later life.

The TOuch BAlance Trainner (TOBAT): Facilitating Frequent and Effective Home- Based Balance Training in Community-Dwelling Older Adults with the Help of Touch Feedback

Shirley Handelzalts, Physical Therapy

Ilana Nisky, Biomedical Engineering

In this collaborative clinician-engineer applied research project we aim to enhance balance performance and self-efficacy in community-dwelling healthy older adults by integrating an innovative, wearable haptic device providing augmented haptic feedback into off-the-shelf home-based balance training platform. Sensorimotor learning theories will help to optimize performance through attention and motivation. Specific aims: (1) Developing a wearable haptic device – the Haptic Bracelet – and its integration into an off-the-shelf balance training system (2) Evaluating user experience, feasibility, and short-term effectiveness in the lab and at home (3) Testing the user-experience and effectiveness of a home-based balance training program performed with and without haptic augmented feedback.

Towards Cognitive bio-inspired robots – iRSTAR learning 3D motion trajectories autonomously

David Zarrouk, Mechanical Engineering
Sigal Berman, Industrial Engineering and Management

Background and Objectives

The RSTAR (Rising Sprawl Tuned Autonomous Robot), is an innovative crawling robot capable of reconfiguring its shape and moving the location of its center of mass. These features give the RSTAR inherent robustness and enhanced ability to overcome obstacles and crawl on different terrains for a variety of applications. Executing trajectories that utilize the robot's capabilities is difficult, especially when complex maneuvers are required. We have applied reinforcement learning using Unity game engine to determine optimal strategies to overcome typical off-road obstacles. In the current research, we further advance the development of intelligent RSTAR (iRSTAR) capable of learning autonomously. The innovations will be in both the learning methods and in the robotic structural and perception capabilities. We will aim at learning in both known and previously unknown environments. Additional innovations are in learning based on sensor fusion and later on regarding the interaction between the learning method, intelligence, and the mechanism's design.

Novelty

The long-term aim of the research is to create cognitive bio-inspired robots capable of reasoning, adaptation, and autonomous operation in dynamic environments. In parallel to the creation process, we aim to advance the in-depth understanding of the relationship between mechanisms and the intelligence they afford and require. The current effort aims to develop the iRSTAR capable of learning 3D motion in unknown terrains autonomously. The research will advance the state of the art in two layers: the single robot layer and the higher conceptual layer. At the single robot layer, as each robot type has a unique design, equipping a robot such as the RSTAR with cognitive abilities, and therefore, enhancing its range of capabilities is an important contribution. The current effort will contribute mainly to this layer. The capabilities of the iRSTAR will be considerably enhanced by the addition of motion learning capabilities. At the conceptual layer, the range of robot designs facilitates a unique opportunity to assess the relationship between mechanism design and its intelligence, leading to a wider understanding of the interactions between them.

Research Plan

Dynamic learning using a continuous mapping in simulation
Dynamic learning using a continuous mapping in hardware
Sensor fusion and high level-level control