Simplified Walking: A New Way to Generate Flexible Biped Patterns

Jinsu Liu, Xiaoping Chen, Manuela Veloso2009 · DOI

TL;DR

This paper introduces "Simplified Walking," a novel method for generating flexible biped robot motion patterns with significantly reduced computational cost. By treating walking as a sequence of discrete "simplified steps" based on the Linear Inverted Pendulum model and Zero Moment Point (ZMP) concepts, it enables real-time motion planning and smooth transitions between different gaits.

What Problem Does This Paper Address?

Generating flexible and dynamic walking patterns for biped robots is a significant challenge in robotics. Existing methods often rely on complex models and can be computationally expensive, limiting their ability to adapt quickly to changing environments or tasks in real-time. The core problem is to achieve smooth, stable, and adaptable bipedal locomotion without incurring prohibitive computational overhead, especially for tasks requiring rapid motion planning and replanning.

Methods

The paper proposes a "Simplified Walking" approach that breaks down complex bipedal motion into a series of discrete "simplified steps." Each simplified step is modeled using the 3D Linear Inverted Pendulum (3D-LIP) model, which approximates the robot's motion during a single-support phase. The key innovations include:

  • Simplified Step Concept: Instead of time steps, the fundamental unit of motion is a "walking step," simplifying the planning process.
  • ZMP Decision: The approach focuses on determining a single Zero Moment Point (ZMP) for each simplified step, rather than generating a continuous ZMP trajectory.
  • Periodic Motion Design: Techniques are presented to design repeatable bipedal motions (starting, walking, stopping steps) and connect them smoothly.
  • Double-Support Phase Generation: A method is described to convert parts of simplified steps into double-support phases, ensuring smooth transitions between feet and maintaining stability. This involves optimizing the CoM trajectory to minimize the difference between the desired and actual ZMP.

Datasets

Not discussed in the paper. The experiments were conducted on simulated and real Nao robots.

Key Insights and Ideas

  • Abstraction for Efficiency: By abstracting the complex dynamics of biped walking into discrete "simplified steps" and focusing on ZMP decisions rather than trajectories, the computational burden is drastically reduced. This is the core enabler for real-time planning.
  • Discrete Planning for Continuous Motion: The paper elegantly converts a continuous motion planning problem into a discrete one, making it more tractable for real-time applications.
  • Smooth Transitions are Key: The method explicitly addresses the challenge of connecting different walking motions (e.g., forward to sideways) and generating stable double-support phases, which are crucial for practical bipedal locomotion.
  • Flexibility through ZMP Control: The ability to modify landing gestures before the current swing foot lands highlights the enhanced flexibility offered by this approach.

Key Results

The paper demonstrates the effectiveness of the Simplified Walking approach through experiments on simulated and real Nao robots.

  • Successful Motion Connection: Figure 4(a) shows a smooth connection of two CoM trajectories, and Figure 4(b) illustrates the simulated robot successfully executing these connecting steps.
  • Real-time Planning: Figure 5 shows snapshots of online biped motion planning controlled by a joystick, demonstrating the system's ability to generate subsequent motions in real-time.
  • Practical Application: The authors report successfully using their simplified walking method in complicated tasks like robot soccer competitions, achieving first place at the 2008 RoboCup China-Open.

Limitations and Future Work

Limitations: * The paper primarily focuses on the ZMP and CoM dynamics, with the effects of inertia being ignored in the simplified model. * The description of the trunk movement is presented by a series of ZMP decisions, which might be a simplification for certain complex maneuvers.

Future Work: * The authors plan to focus on developing more "intellectualized" methods for designing periodic biped motions. * They intend to extend the dynamic model of simplified walking to accommodate alterable CoM heights, which would further enhance the robot's adaptability and expressiveness.

Summary updated

BibTeX
@misc{liu2009simplified,
  author = {Jinsu Liu and Xiaoping Chen and Manuela Veloso},
  title = {{Simplified Walking: A New Way to Generate Flexible Biped Patterns}},
  year = {2009},
  doi = {10.1142/9789814291279\_0071},
}

Loading related content...

Figure