Part of the UZ Leuven Neurophysiology Research Group Celia Bougou Santiago Iregui Sofie De Schrijver and Peter Janssen

Opposite: Part of the UZ Leuven Neurophysiology Research Group. From left to right: Celia Bougou, Santiago Iregui, Sofie De Schrijver and Peter Janssen.

Mr. Iregui is a PhD candidate performing research on constraint-based shared control for assistive robotics. Among others, he is in charge of the development of the robotics-related part of the multidisciplinary KU Leuven C1 project “Neurophysiological investigations of the human and nonhuman primate brain: from perception to action”.

What is the background of the project?

Every year between 250 000 and 500 000 people suffer a  spinal cord injury (World Health Organization, 2013). In order to improve their quality of life, robotic solutions can be used to aid them in performing daily tasks.

"These solutions should not be fully autonomous, since people with nervous system lesions have the need of recovering autonomy and they want to feel in control as much as possible. Nevertheless, it is very difficult for them to perform tasks when they have full autonomy of the robot."

What was the specific challenge regarding these robotic solutions?

There is a wide range of mobility limitations that can be caused by lesions like spinal cord injuries. Thus, the assistive strategy should be easily adaptable to fit the needs of the individual person. The person’s remaining mobility should be maximally exploited through a suitable human-machine interface (HMI), without losing his/her autonomy.

"Also, the robot needs to be teleoperated in an unstructured and unpredictable environment. Therefore, the robot should be equipped with intelligent algorithms that handle multi-sensor information to perceive the environment and seamlessly guide the operator to reach their intended goals."

Diagram showing assistive strategy for object grasping object detection novel invasive BMI computer's & operator's anatomy

What was the situation you were encountering that required you to seek out a company like Kinova?

Kinova is a company specialized, in particular, in assistive robots. Other robots in the market provide very big and robust robots that are suitable for industrial applications, but not for interacting with people with disabilities. If I’m not wrong, we found out about the great features of the Gen3 robot in IROS 2018.

And what approach did you choose to perform?

Our constraint-based task specification and control framework eTaSL enables an assisted teleoperation of the Kinova Gen3 robot arm by means of a shared autonomy approach. We use reactive control to deal with the unpredictability of the environment, such as avoiding dynamic obstacles and interacting with humans.

"To enable seamless assistance, an intent estimation algorithm combines the partial input of the subject with information from a commercial off-the-shelve 3D vision system (Pick-it)."

To comply with the needs of people with different levels of mobility, multiple human-machine interfaces (HMI) are being tested within the project. For example, in collaboration with UZ Leuven (university hospital), a novel invasive brain-machine interface for tetraplegic people is being developed. On the other hand, a 3DConnexion SpaceMouse is being explored to aid paraplegic people.  


This video above was presented in the ICRA 2020 Workshop on Shared Autonomy: Learning and Control. You can find an extended abstract on the website of the workshop.

What are the key objectives of the project?

To develop robotic assistive strategies that:

  • Accelerate the learning curve, reduce the required effort (fatigue) and improve the dexterity of the patients when teleoperating the robot arm.
  • Offer flexible, easily adaptable, and personalizable assistance.
  • Guide people to perform daily tasks in unstructured and unpredictable environments.
  • Estimate the intention of the operator and deal with it reactively.

What are the results so far?

As preliminary results of this ongoing project, we developed a tube-shaped Reactive Virtual Guidance Fixture (RVGF) that adapt its shape at two levels:

Globally, to adapt the target position of the RVGF towards goals determined by online intent estimation, while preserving the geometric shape learned from demonstrations; locally, to perform collision avoidance of dynamic obstacles by deforming locally (i.e. a section of the RVGF), while overcoming the limitations of the learned information.

We tested our approach by using a 3DConnexion SpaceMouse to teleoperate the robot. In a later phase of the project, we plan to perform experiments with a visuomotor Brain-Machine Interface.

The video above was presented at the IROS 2019 Workshop called "Legacy Disruptors in Applied Telerobotics: Improving the Machine, the Interface and the Human".

What was the impact of Gen3 on your project?

The use of the Kinova Gen3 arm has had a very positive impact on our research, allowing safe and enjoyable interaction with humans. Also, the lightweight feature of the robot has been very useful for us since we need to reallocate our setup when performing experiments at UZ Leuven. 

Who would you recommend Gen3 for?

Anyone that is either performing research in assistive robotics or integrating assistive solutions and bringing them to the market.

“I think that the compact design of the Kinova Gen3 robot, in combination with the multiple capabilities offered by the Kinova Kortex, has boosted our research in the field of assistive robotics.

Hopefully, soon we’ll be able to see robots with these great features assisting more and more people in their daily tasks. As researchers and industry representatives, it is our responsibility to make this happen."


For more information about this specific project:


Orocos real-time driver for the Gen3 robot


Assistive Robotics page of KU Leuven Robotics:


YouTube Channel of the group:


Our constraint-based framework for reactive robot control:


Learn more about the Kinova Gen3 robotic arm

About KU Leuven's Robotics Research Group

The KU Leuven Robotics Research Group has pioneered robotics research in Europe since the mid-1970s. The group was among the first to develop active force feedback for assembly operations. Already in 1980 it developed learning insertion algorithms based on stochastic automata. It has covered virtually all aspects of sensor-based robotics, from the high-level task specification and planning down to low-level sensor-based control, and applied the research results in a variety of industrial applications. In the last decade, the group shifted its attention towards service robots (behaviour-based mobile manipulation, shared control, learning control), medical robotics (natural interfaces, haptic bilateral control), industrial robot assistants, and active sensing (vision, force, etc).

KU Leuven has created several spin-off companies that are active in robotics-related activities, have initiated several free and open-source software projects in robotics, such as Orocos, eTaSL and KDL, and has participated in multiple robotics and control-oriented EU projects. The group is a university core lab of Flanders Make, the innovation network and strategic research center for the manufacturing industry in Flanders, Belgium.