From a mechanical guide dog to a self-learning exoskeleton and magnetically controlled bacteria, researchers at ETH Zurich are busy devising robots for medical applications.

Not just for play: robotic dogs may one day assist visually impaired people. Researchers are now equipping them with AI-based control software to make that possible. (Image: Marco Giordano / Center for Project-Based Learning D-ITET, ETH Zurich)

Robodog is an instant hit with the 20-strong party from Singapore. A woman beams with delight as she talks to the canine robot, while the man to her left bends down and pats it affectionately on the head. Robodog certainly looks adorable, capering around the group with a comical little dance. Yet this technology is designed for more than mere entertainment. This cute robot can be trained to assist those in need – as a guide dog for the blind, for example, or a helper for someone with paralysis.

Robodog moves and navigates autonomously. It can climb stairs and avoid bumping into obs­tacles, even at speed. A brief encounter with the Singapore group is enough to showcase its many capabilities. Developed by Michele Magno, head of the D-ITET Center for Project-Based Learning, and his research team at ETH Zurich, the robot was ori­ginally conceived as a commercial product. But, as Magno explains, it’s not a plug-and-play solution. “To use it for our purposes,” he says, “we first need to modify it.” For a start, the robot’s movements are still far too clumsy, and it makes a real racket when walking. “That’s a major drawback for a guide dog,” he adds, “because people with visual impairments depend on being able to hear ambient sounds.”

Smart sensors, cleverly used

The Energy-­Efficient Sensing and Systems team under Magno has therefore developed new AI-based control software for Robo­dog. Magno himself uses the term “physical AI” – one of the latest buzzwords to emerge from the field of artificial intelligence. The system makes smart use of data from built-in sensors that register parameters such as the robot’s movements and the spatial orientation of its mechanical parts. In the case of Robodog, this helps it learn how to move more smoothly and safely across different surfaces and terrains.

Magno and his team have also fitted depth cameras, which detect the distance to surrounding objects, along with other sensors that scan the environment. In addition, they are constantly trying out new sensors, checking to see if they give Robo­dog greater autonomy. Right now, they’re testing out a radar system. This emits radio waves and captures their reflection from nearby walls and objects, helping Robodog to map the surrounding area in more detail.

Even in the absence of external data, Robodog is astonishingly adept – a function of its ability to learn new tricks. Indeed, doctoral student Davide Plozza, the person primarily responsible for this continual refinement, recently guided Robodog to first prize in a robotics competition. Plozza’s protégé was the only contestant capable of mastering a tricky obstacle course ten times in a row without incident. Rival robots, by contrast, all got stuck or fell at one of the hurdles.

This technology might not be quite at the stage where it can assist the visually impaired, but it’s well on its way. If required, Robodog can also be fitted with a robotic arm, enabling it to help those with a physical impairment – fetching objects, for example, or opening doors. In this instance, Robo­dog is controlled via special glasses equipped with sensors that register the wearer’s line of sight. Here, a glance at the door handle suffices to issue the command.

Help with walking

Meanwhile, ETH professor Robert Riener and his team have developed a smart exoskeleton to support people with partial paralysis of the lower body. Known as the Myosuit, this self-learning machine is worn over everyday clothing and assists the wearer as they walk. “Most conventional exoskeletons are made mostly of metal, but our system is soft, flexible, lightweight and comfortable to wear,” explains Riener, who heads the Sensory-Motor Systems Lab at ETH Zurich.

Supported by a robotic exoskeleton, physiotherapy patients are able to complete more exercises, which helps them regain mobility more rapidly. (Image: Stefan Schneller / ETH Zurich)

The Myosuit combines functional textile elements worn around the waist and shoulders with lightweight plastic components hooked up to small motors. Power from these motors is transmitted via cables attached to the structural parts, helping the muscles and tendons in the legs to do their job. This in turn provides external support for hip and knee joints and relieves the effort of moving against gravity. The exoskeleton also adapts to each user’s abilities, says Riener: “Built-in sensors continuously measure the amount of force the wearer is able to exert when walking, and the robot learns to provide exactly the right amount of additional force.”

The Myosuit can also assist patients who are recovering from a stroke or injury to the spinal cord. It enables wearers to increase the intensity at which they repeat physiotherapy exercises, thereby ensuring a better and more rapid restoration of motor function. Equally important, the Myosuit is easy to put on and can be worn at home and outdoors, meaning that people can use it outside of a clinical setting. For example, it can enable people with lower limb impairment to climb stairs again, walk for longer and participate more fully in social life.

Patients with conditions such as muscular dystrophy, multiple sclerosis or heart failure could also benefit from the Myosuit, says Riener: “People with heart failure become tired very quickly, so they tend to move less and less and just get weaker and weaker.” As a recent joint study led by Riener and the Charité University Hospital in Berlin has shown, robot-assisted movement systems can help remedy this problem.

For some people, the Myosuit is already part of day-to-day life. Riener tells of a patient previously reliant on a wheelchair who could walk no more than 100 metres. But thanks to the Myosuit, he can now manage several kilometres. Indeed, as Riener delightedly explains, he was also able to complete 6 kilometres of the Zurich Marathon.

Bacteria as drug carriers

And then there are other robot-like applications in the medical field – minuscule ones, in fact – that are not really like machines at all. One striking example is the magnetically guided bacteria developed by Simone Schürle, head of the Medical Microsystems Lab at ETH Zurich, and her research group for use in cancer therapy. “In the search for new cancer drugs, the main reason clinical trials fail is that too little of the active ingredient reaches the tumour or its metastases,” Schürle explains. One possible solution is to use bacteria: “They can carry therapeutic agents directly to the tumour site, reducing the spread of drugs throughout the rest of the body, where they might cause serious side effects.”

Also robots of sorts: these bacteria are coated with magnetic nanoparticles, meaning they can be guided within the bloodstream using magnetic fields to a specific location in the body. (Image: Stephan Handschin / ScopeM, ETH Zurich)

To make this new drug-delivery method work, Schürle and her research group first had to find a way to steer the bacteria through the body. The team coated the outer cell wall of E. coli – a bac­terium naturally found in the human intestine – with about a thousand iron-oxide nanoparticles. “This allows us to guide the bacteria using magnetic fields applied from outside the body,” Schürle says. Once in the bloodstream, they are piloted to their target by a clever array of electric and permanent magnets. To ensure the bacteria actually deliver their payload in the tumour, they are given an extra push by the magnetic control system so that they rotate and pass through the blood vessel wall into the surrounding tissue.

That still left the question of how to get the bacteria to carry their payload of drugs. To achieve this, Schürle and her team chemically bonded tiny fat globules to the outside of the cell wall. Known as nanoliposomes, they transport the drug for uptake and release in cancer cells. In addition, the bacteria can be genetically programmed to produce and release a further active ingredient in the immediate vicinity of the tumour. This would enable two different agents to be delivered simultaneously to the same target.

Although her bacteria are not machines, Schürle, who cut her teeth in the field of robotics, still prefers to describe them as robots. “At the end of the day, their actions are controlled by us,” she says. Right now, she is researching a potential application for magnetically controlled micro-robots made of a hydrogel. Hydrogels comprise a network of polymers and can release active substances in a controlled manner – in this instance, substances that dissolve blood clots.

Experiments with mice and with human cell tissue models have shown that micro-robots can be used for the targeted transport of active ingredients. This may not be as cute as a canine robot, but it may end up being just as useful.