Humans and animals manage to move efficiently without giving it much thought, thanks to the natural ways their bodies oscillate. A new tool has emerged that leverages this understanding to enhance the movement of robots for the first time.
When four-legged animals begin to walk and gradually accelerate, they eventually shift into a trot. This transition occurs because maintaining their initial gait would require more energy. This relationship was discovered over 40 years ago. Recently, Alin Albu-Schäffer, a professor at TUM in the Chair of Sensor-based Robotic Systems and Intelligent Assistance Systems, has effectively applied this principle to robot movement.
Experts refer to the energy-efficient movements seen in humans and animals as ‘intrinsic dynamics.’ For instance, when walking on a firm surface, they automatically adjust their muscle stiffness. These automatic adaptations are difficult to detect in both humans and complex robotic designs.
New tool isolates the most efficient movements
However, a new tool created by Prof. Albu-Schäffer’s team at TUM makes this identification possible. “For the first time, we’ve made these highly efficient, intrinsic movements quantifiable. This tool helps determine which actions of a system are the most efficient,” they explain.
A key subject in their research is BERT, a quadrupedal robot resembling a small dog. BERT was developed by Prof. Albu-Schäffer at the German Aerospace Centre (DLR), and the project investigates “efficient and versatile leg-based movement.” The study is supported by the EU via an ERC Advanced Grant.
The researchers pinpointed six distinct movement patterns for BERT, which Prof. Albu-Schäffer describes as incredibly easy and energy-free in a frictionless environment. Some of these movements resemble familiar gaits of four-legged creatures, such as walking, trotting, or hopping. “We have validated the idea that natural oscillation patterns can lead to efficient gaits,” clarifies Prof. Albu-Schäffer, who also participates in the Munich Institute of Robotics and Machine Intelligence (MIRMI).
Timing natural oscillations perfectly
To replicate these movements in a frictional environment, a computer-controlled regulator has been integrated to deliver impulses at optimal times. “It’s like a child on a swing receiving a push from a parent right as they reach the top,” explains Annika Schmidt from Prof. Albu-Schäffer’s research team. The difference lies in that “humans don’t require complex calculations to time their push; they just do it instinctively,” says Schmidt, a doctoral student who has been exploring how to instill the right rhythm in robots for years.
The effectiveness of the new method is demonstrated in a race among three BERT models. The BERT that has been programmed with this intrinsic movement technique tends to leap and dart more swiftly and energetically compared to its counterparts that use traditional movement styles.
