As populations age and the demand for care continues to grow, researchers are exploring how robots can help seniors maintain their independence while supporting overburdened healthcare systems. At the Munich Institute of Robotics and Machine Intelligence (MIRMI) at the Technical University of Munich (TUM), the geriatronics team located in Garmisch-Partenkirchen is tackling this challenge through GARMI, a mobile manipulation platform designed specifically for real-world elderly care applications.
Since 2026, GARMI has been based on the omnidirectional mobile Clearpath Ridgeback platform and combines mobility, manipulation, telepresence, and human-centered design into a single research platform that helps researchers bring assistive robotics from the lab into everyday life.
Designing Robots Around Real Human Needs
The team’s work is rooted in a simple observation: people are living longer, but the number of caregivers is not keeping pace. Most older adults would prefer to remain in their own homes for as long as possible, yet many daily tasks become increasingly difficult with age.
Rather than developing technology for technology’s sake, the team works closely with seniors, clinicians, and care providers to identify where robotics can make a meaningful difference.
GARMI serves as the centerpiece of this effort. The humanoid service robot is designed to support activities of daily living, healthcare applications such as telemedicine and tele-rehabilitation, and immersive telepresence experiences that allow family members or healthcare professionals to interact remotely through the robot. The platform operates within a fully equipped pilot apartment and is developed in partnership with care provider Caritas, allowing researchers to evaluate new robotic capabilities in realistic residential environments.
“We develop field-test assistive, telepresence and safety skills under authentic everyday conditions. GARMI is our central experimental platform for exploring how robots can support aging societies.”
GARMI at the Clearpath Robotics by Rockwell Automation booth at ICRA 2026.
Why a Mobile Manipulator Instead of a Legged Humanoid?
While public attention often gravitates toward walking humanoid robots, the research team took a different approach. For environments such as homes, clinics, and care facilities, they determined that a wheeled mobile manipulator offered a more practical solution than a bipedal robot.
“These spaces are largely flat and structured indoor environments. That’s exactly where an omnidirectional wheeled base excels.”
The decision was driven by a combination of safety, reliability, and practicality. A wheeled mobile manipulator provides a stable platform for physical interaction, supports the payload requirements of a full upper-body system, consumes less energy by eliminating the need for constant balancing, and offers the reliability needed for long-term field studies. The team also wanted to pursue a platform that could one day be affordable enough for broader deployment, a goal that would be significantly more difficult with a legged system.
Because GARMI is intended to operate around older adults, many of whom may be physically vulnerable, safety remains a central design consideration. The robot’s whole-body compliance allows it to respond safely to unexpected contact, while its stable mobile base provides a predictable and reassuring presence during interaction.
Why Ridgeback?
Earlier versions of GARMI used an in-house differential-drive mobile base. While effective during early development, it became a limiting factor as the platform moved into increasingly realistic care environments.
“The non-holonomic constraints of the previous base became a genuine limitation for precise positioning and maneuvering.”
For the latest generation, researchers selected the Clearpath Ridgeback because its omnidirectional mobility, payload capacity, and mature ROS ecosystem aligned closely with the project’s requirements. The ability to move freely in any direction allows GARMI to position itself precisely in tight domestic and clinical environments, making it easier to perform manipulation and assistance tasks.
At the same time, the platform comfortably supports the robot’s lift-column torso, dual Franka arms, perception hardware, and communication systems while maintaining a stable, low center of mass. Equally important for a research group working on complex assistive robotics challenges, the established ROS software support allowed the team to integrate the platform quickly and focus their efforts on higher-level research problems rather than maintaining custom mobile-base infrastructure.
Bringing GARMI to Life
One of GARMI’s most distinctive features is its appearance. Rather than creating a machine that resembles the intimidating humanoids often seen in pop culture, the team deliberately pursued a smaller, approachable design shaped by feedback from seniors and healthcare professionals.
The resulting robot stands approximately 165 centimeters tall and weighs around 105 kilograms. Rounded contours, a compact upper body, forward-facing arms, and a large interactive head display were all chosen to create a friendly and approachable impression. The researchers describe the design philosophy as one that seeks to “under-promise and over-deliver,” an approach that tested well with users throughout the development process.
Integrating the complete system, however, was far from simple. The latest iteration required researchers to bring together the Ridgeback base, lift-column torso, dual-arm manipulation system, perception sensors, head mechanism, and an entirely new ROS 2 software stack into a single field-ready platform. According to the team, this convergence of hardware and software components proved to be one of the most challenging stages of development and required substantially more commissioning effort than initially anticipated.
Like any research platform, GARMI reflects a series of engineering trade-offs. The team intentionally prioritized a compact, approachable form factor over a larger and more imposing design, while navigation currently requires the robot’s arms to remain folded within the footprint of the base until a full 3D body-aware navigation system is completed.
Building a Research Platform for Real-World Validation
Today, GARMI operates on a ROS 2-based software stack that combines Clearpath’s mobile platform software with MoveIt 2, libfranka, and the team’s custom autonomy framework. Researchers developed a reproducible environment that integrates the mobile base, robotic arms, perception systems, and autonomy software into a single platform that can be deployed consistently across experiments.
For navigation, the robot uses an RTAB-Map-based pipeline that combines data from onboard 3D cameras and 2D LiDAR sensors to build maps and localize within the team’s pilot apartment. Nav2 is then used to plan collision-free paths and move the platform between target locations. Beyond serving as GARMI’s current navigation solution, this software stack has become the benchmark against which future autonomy approaches will be evaluated, giving researchers a reliable foundation upon which to build more advanced capabilities.
A research team member posing with GARMI at ICRA 2026.
Looking Ahead
With much of the platform development now complete, the team’s focus is shifting toward real-world validation. Upcoming work includes field trials with seniors and caregivers through their partnership with Caritas, where the researchers will evaluate GARMI’s assistive capabilities in authentic care environments once ethics approval is in place.
At the same time, the team continues to expand the platform’s technical capabilities. Current research efforts include developing navigation systems that account for GARMI’s full three-dimensional geometry, integrating semantic mapping and navigation-based manipulation, and exploring large-language-model-driven interaction through the robot’s communication interface. Researchers are also working on immersive telepresence experiences, remote caregiving applications enabled by next-generation wireless networks, and digital twins that can support safety validation and future certification efforts. In parallel, new healthcare applications such as robotic frailty assessments and cognitive screening are being explored.
For the team at TUM, however, the long-term goal remains unchanged: translating robotics research into practical tools that can support aging populations. As assistive robotics continues to evolve, GARMI demonstrates what can be achieved when researchers design systems around real users, real environments, and real-world challenges.
Learn More
Interested in building your own mobile manipulation platform? Explore the Ridgeback omnidirectional mobile robot and see why research teams around the world use it as the foundation for applications ranging from autonomous navigation to advanced human-robot interaction.