Zhejiang University Researchers Map Path to Bionic Motion for Legged Robots
- MM24 News Desk
- 3 days ago
- 2 min read
Zhejiang University researcher Jinyuan Liu and his team have conducted a comprehensive review that tackles the core challenges preventing legged robots from achieving true animal-like agility. By focusing on the single-legged robot (SLR) as a fundamental "limb unit," the research systematically analyzes design, modeling, and control strategies to bridge the significant gap between engineering and bionic motion.
Why can a mountain goat traverse a rocky cliff with ease while even the most advanced robot stumbles? The answer lies in the profound complexity of legged locomotion.
While legged robots promise unparalleled mobility on rough terrain, they are notoriously difficult to build and control. Their legs must be strong enough to support the body, powerful enough to propel it, and resilient enough to withstand the repeated shock of impact. Furthermore, controlling multiple joints in real-time with high precision presents a massive computational challenge.
"Compared to the complete multi-legged robots (MLRs), single-legged robots (SLRs) feature simpler configurations and typically admit a dynamic gait: hopping," explained the author Jinyang Liu, a researcher at Zhejiang University.
"Its hopping period can effectively characterize the behavior of multi-legged and single-legged structures, making it easier to tackle structural design and dynamic control problems." This focus on the SLR as a foundational building block allows researchers to isolate and solve core problems before scaling up to more complex multi-limbed systems, according to the team's report.
READ ALSO: https://www.modernmechanics24.com/post/hidden-sense-found-in-humans-shorebirds-sandpipers
On the hardware front, the review categorizes SLRs into two main camps. Telescoping designs act like a pogo stick, using a linear prismatic motion for jumping; they are mechanically simple and excellent for basic validation. In contrast, articulated designs mimic animal legs with multiple joints and are further classified by how they handle elasticity.
These include rigid legs (RALR), legs with Parallel Elastic Actuation (PEALR), and those with Series Elastic Actuation (SEALR). Each configuration involves a critical trade-off; for instance, a SEALR can absorb landing impacts and improve energy efficiency but adds mechanical complexity.
Modeling these dynamic systems is another major hurdle. The research delineates two principal approaches. The first uses Spring-Loaded Inverted Pendulum (SLIP) models, which simplify the robot to a mass on a spring to capture the essential dynamics of hopping. The second employs articulated reduced models, which are more complex but facilitate the design of sophisticated, task-level controllers.
When it comes to making these robots move, the choice of control strategy is paramount. The team systematically compared model-based methods, which are interpretable but rely on accurate system data, and model-free methods like Reinforcement Learning (RL), which are adaptable but can be "black boxes" that are difficult to train and transfer from simulation to reality. This Sim-to-Real transfer is one of the most significant bottlenecks, often causing performance to degrade due to unmodeled effects like sensor noise and actuation delays.
However, SLRs still face practical gaps, including model-reality mismatch under contact uncertainty and complex trade-offs between performance, weight, and reliability. "Therefore, future research should pursue tightly coupled advances in morphology, materials, and control," said Jinyuan Liu.
The path forward, he suggests, involves bio-inspired structures, lightweight fabrication techniques like topology optimization, and new materials like high-energy-density elastomers, all tightly integrated with intelligent control systems to finally achieve stable, efficient, and general-purpose legged robots in our complex world.
Comments