SerpApi.Org
Bioinspired multimodal robots that can switch between different forms of movement, such as flying, walking, swimming, or climbing, are rapidly advancing toward matching the versatility of animals.
The study highlights how researchers are developing robots capable of adapting to changing environments by combining multiple modes of locomotion in a single machine.
It also outlines the key engineering challenges, including limited onboard space and body transformations required for seamless movement, and proposes five performance metrics to evaluate these systems.
The framework aims to guide the development of more capable, efficient, and adaptable next-generation robots.
Nature inspires robotics
Bioinspired multimodal robots—machines capable of switching between multiple forms of movement such as walking, flying, swimming, climbing, or jumping—are rapidly advancing toward the agility and adaptability seen in nature, according to a new review.
The new study led by researchers at Beihang University, Dalian University of Technology, and EPFL examines the evolution of bioinspired multimodal robots, the engineering challenges limiting their development, and a new framework for evaluating their performance. Unlike conventional robots that use a single form of locomotion, multimodal robots combine two or more movement modes, such as walking, flying, or swimming, on one platform. The field has evolved from combining separate mechanisms to creating integrated, animal-inspired systems that use shared structures and intelligent control for greater adaptability and efficiency.
According to the review, the main goal of multimodal robotics is not merely to add more movement options but to improve overall performance in challenging environments. A robot that can both fly and walk, for example, can use flight for rapid long-distance travel before switching to walking for precise inspection on the ground. Similarly, amphibious robots can transition between land and water to perform search-and-rescue operations, environmental monitoring, or underwater exploration.
The study identifies several major engineering challenges that must be overcome to make these robots practical. One of the biggest obstacles is limited onboard space, as every additional movement mode requires actuators, sensors, batteries, and mechanical components that increase size and weight. Components useful in one mode can become dead weight in another, reducing efficiency.
Another challenge is body morphing—the ability to physically reconfigure the robot’s structure to support different types of movement. Designers must also balance stiffness and flexibility, integrate different actuation systems, and ensure that multiple locomotion modes work together rather than interfere with one another.
Smarter motion systems
To address the lack of standardized evaluation methods, the researchers propose five performance metrics for multimodal robots. These measure the number of movement modes a robot possesses, the additional cost of adding new capabilities, how many components can be shared across different modes, the time or energy required to switch between modes, and the overall performance gains achieved by combining multiple forms of locomotion.
The review also examines emerging design strategies that could improve future robots. Soft materials and flexible structures allow robots to deform and adapt to their surroundings, while structure repurposing enables the same components to perform multiple functions. Another promising approach is multirobot architecture, in which teams of simple robots cooperate to achieve multimodal capabilities collectively.
On the software side, the researchers note that conventional planning and control algorithms struggle with the dramatic changes in dynamics that occur when robots switch movement modes. They argue that advances in reinforcement learning, vision-language-action models, world models, and physically intelligent robot bodies will be essential for enabling seamless transitions and autonomous decision-making in complex environments.
Looking ahead, the authors envision multimodal robots that tightly integrate adaptive hardware with AI-driven perception, planning, and control. Achieving this will require advances in materials, high-performance actuators, sensors, and learning algorithms capable of handling unpredictable real-world conditions. If these challenges can be overcome, the researchers argue, future bioinspired multimodal robots could not only match the versatility of animals but, in some applications, surpass their natural counterparts.
