Rethinking energy-driven modeling, control, and design for soft robots
Philipp Mitterbach defended his PhD thesis at the Department of Mechanical Engineering on 23 June.
In today鈥檚 technological landscape, machines are expected to do exactly what we want. This is achieved by giving them precise instructions, supported by accurate sensing, carefully modeled behavior, and tightly controlled environments. Such an approach has been highly successful in industrial settings, where reliability depends on reducing uncertainty as much as possible. However, as technology moves away from factory floors into everyday environments, these assumptions begin to break down. Real-world conditions are not fixed or predictable, creating a growing tension between how machines are designed and how they are expected to function in practice. In his PhD research, Philipp Mitterbach addresses this challenge by exploring how soft robotic systems can be modeled, controlled, and designed to not rely solely on explicit instructions.
Current systems are built on the premise that behavior must be explicitly defined and enforced. As systems become more complex, more interconnected, and embedded in everyday environments, this approach becomes harder to sustain. An alternative is to design systems that achieve desired behavior by leveraging their physical properties, natural dynamics, and interaction with the environment. Philipp Mitterbach focuses on soft robotics to make this shift tangible.
Potential applications for soft robotics
Soft robotic systems are built from highly deformable materials. Because of this, they cannot be controlled with the same precision as rigid machines. However, their deformability allows them to absorb, store, and release energy, enabling environmental forces such as gravity and contact to support motion. This opens up new possibilities for how machines operate in real-world contexts. For example, soft systems could move across delicate or uneven terrain without damaging crops or compacting soil in agriculture. In environmental monitoring, they could operate in fragile ecosystems such as wetlands or forests while minimizing disturbance. In infrastructure inspection, they could navigate confined or irregular spaces by conforming to their surroundings. In applications involving human interaction, their ability to distribute forces enables safer and more natural physical assistance.
Principles for rigid machines
However, this potential is not automatically realized. In practice, many soft robotic systems are still designed and controlled using principles developed for rigid machines, forcing them to follow predefined trajectories and maintain stable configurations through continuous actuation. This requires constant energy input to sustain motion and stability. As a result, their physical properties are not fully exploited, limiting their efficiency, adaptability, and ability to operate robustly in real-world environments.
Incorporating physical dynamics
This reveals a gap between the promise of soft robotics and its current state. Simply making a system soft does not guarantee efficient or robust behavior under real-world conditions. The challenge lies not only in the material but also in how the system is modeled, controlled, and designed. Addressing this gap requires rethinking how systems are developed. Instead of treating physical dynamics as effects to be corrected, they can be incorporated into system behavior, reducing the need for continuous control and improving efficiency and adaptability in complex environments. This involves organizing how energy is generated, stored, and exchanged with the environment, and designing both hardware and software to support this process.
Harnessing gravity, elasticity, and environmental forces
In his research, Philipp Mitterbach developed principles that align modeling, control, and design with this perspective. He demonstrates how soft robotic systems can move beyond merely adapting to their environment toward actively using the forces present within it. Such approaches use control strategies and design principles that harness gravity, elasticity, and environmental forces rather than counteracting them. In this way, physical dynamics become an integral part of system function, enabling more effective and reliable behavior. Ultimately, this reflects a broader technological shift toward systems whose functionality emerges not only from explicit instruction, but also from their physical interaction with the world around them.
Title of PhD thesis: Supervisors: Dr. Sasha Pogromskiy, Dr. Irene Kuling and Dr. Simon Eugster.