Improved Performance on Moving-Mass Hopping Robots with Parallel Elasticity

Eric R. Ambrose, Aaron D. AmesICRA 2020 · DOI

TL;DR

This paper introduces a novel hopping robot design featuring an additional spring in parallel with the actuator. This "double-spring" configuration significantly improves energy efficiency (2.5x) and reduces peak actuator force (40%) compared to traditional single-spring designs, enabling stable, open-loop controlled hopping up to 40cm.

What Problem Does This Paper Address?

Hopping robots, while impressive, face a fundamental challenge: efficiently transferring energy during the brief ground contact phase to achieve stable and energetic hops. Traditional designs often lose energy during these impacts, requiring powerful actuators and complex control systems to compensate. Previous work on moving-mass hopping robots demonstrated stable hopping but demanded substantial actuation force. This paper aims to improve both the energy efficiency and reduce the required actuation effort for these dynamic robots by exploring the impact of adding parallel elasticity to the system.

Methods

The core innovation lies in the robot's mechanical design: a "double-spring" configuration where one spring is in series with the actuator (traditional) and a second spring is placed in parallel with the actuator. This parallel spring acts to passively assist the mover's motion, reducing the burden on the actuator.

To analyze this new design, the authors employed a rigorous trajectory optimization method. This involved:

  • Hybrid Systems Modeling: The robot's dynamics were modeled as a hybrid system, accounting for both continuous motion and discrete events like ground contact and internal hardstops. This allowed for a more accurate representation of the complex hopping behavior.
  • Trajectory Optimization: A nonlinear programming formulation was used to find the optimal actuation inputs (current/force) to achieve desired hop heights. This optimization was performed for both the single-spring and the new double-spring models.
  • Experimental Validation: A custom-built, vertically-constrained hopping robot was used to experimentally validate the simulation results. The robot was programmed to follow the optimized trajectories.

Datasets

Not discussed in the paper. The authors generated their own optimization trajectories and used experimental data from their custom-built robot for validation.

Key Insights and Ideas

  • Parallel Elasticity as an Energy Saver: The parallel spring acts as a passive energy storage and return mechanism, significantly reducing the actuator's workload. Instead of the actuator solely driving the motion, it now assists the natural dynamics of the parallel spring, leading to more efficient energy transfer.
  • Open-Loop Stability is Achievable: A major finding is that the double-spring design inherently promotes stability. Unlike the single-spring model, which requires a closed-loop controller for stability, the double-spring robot can achieve stable hopping with simple open-loop playback of optimized trajectories. This is attributed to the parallel spring guiding the mover's motion, making the system more resilient to small deviations.
  • Reduced Peak Force, Increased Efficiency: The parallel spring's passive assistance allows for a substantial reduction in the peak force required from the actuator (40% less). This, in turn, leads to significantly improved mechanical and electrical efficiencies (2.5x and 3x respectively).
  • Internal Contact Management: The paper highlights the importance of managing internal contact between the mover and the body. While the parallel spring can lead to internal contact, the optimization and experimental adjustments (negative current pulse) were crucial for preventing detrimental energy loss during these events.

Key Results

  • Energy Efficiency: The double-spring model achieved 2.5x better energy efficiency compared to the single-spring model.
  • Actuator Effort: Peak actuator force was reduced by approximately 40% in the double-spring model.
  • Experimental Hop Height: The robot successfully achieved hop heights up to 40cm.
  • Stability: The double-spring model demonstrated inherent stability, allowing for open-loop control without the need for stabilizing controllers, a significant advantage over the single-spring model.

Limitations and Future Work

  • Limited to 1D: The current work focuses on a one-dimensional hopping robot. Extending this to 3D hopping would require additional actuators for balancing.
  • Lower Spring Constraint: The maximum hop height was limited by the physical length of the lower (serial) spring.
  • Experimental Deviations: While stable, experimental results showed some deviation from optimal trajectories between internal impact and take-off, attributed to factors like rail backlash and lateral spring deflection.

Future work suggested includes:

  • Developing an alternative elastic component for the lower spring to allow for larger deflections and increased energy storage.
  • Adding actuators to enable 3D hopping and balancing.
  • Further investigation into the dynamics of internal contact and its impact on stability.

Summary updated

BibTeX
@inproceedings{ambrose2020improved,
  author = {Eric R. Ambrose and Aaron D. Ames},
  title = {{Improved Performance on Moving-Mass Hopping Robots with Parallel Elasticity}},
  booktitle = {IEEE International Conference on Robotics and Automation},
  year = {2020},
  doi = {10.1109/icra40945.2020.9197070},
}
Figure