Neural Control Strategies for Energy-Efficient Swimming

This work was published in Science Robotics in January 2026 as “Energy efficiency and neural control of continuous versus intermittent swimming in a fishlike robot” (Liu, Longchamp, Zunino, et al.).

Why Intermittent Swimming Matters

Many fish don’t swim continuously. Instead, they use bout-and-glide (also called burst-and-coast): short, active tail undulations (the bout) followed by a passive glide. This intermittent pattern appears across species and is often hypothesized to improve efficiency; however, the energetic advantage is still debated, partly because it is hard to control animal behavior precisely while measuring both kinematics and power.

This research study addresses that gap with a bio-inspired robotic system, enabling controlled experiments that disentangle the roles of:

• Neural control (how bouts start/stop and how gaits are modulated),
• Body mechanics & actuation (how thrust is produced),
• Hydrodynamics (how drag and momentum evolve across bout and glide).

The Core Idea: A Robot That Embodies a Neural Hypothesis

Zebrafish larvae are a powerful model for locomotion neuroscience, but most zebrafish-inspired robots have historically focused on kinematics alone, without integrated sensing and explicit neural mechanisms.

We developed ZBot, a zebrafish-inspired robotic platform designed to embody a neural model of bout-and-glide swimming:

• The robot provides a physical substrate with realistic body dynamics and fluid interaction,
• The controller implements a mechanistic hypothesis of how bouts are initiated, shaped, and terminated,
• Experiments can systematically vary gait parameters while recording both performance and energy consumption.

ZBot: A Physical Model of a Larval Fish

ZBot is designed to reproduce the qualitative structure of larval swimming while remaining experimentally practical:

• Articulated tail with multiple joints: The tail is composed of six hinge joints actuated by servomotors, enabling undulatory wave generation and modulation.
• Repeatable, measurable experiments: The platform supports repeated trials under matched conditions, enabling parameter sweeps that are infeasible in animal studies.
• Field-ready testing: Beyond lab-style tank experiments, the system can be deployed in real environments (e.g., stationary water outdoors) for realistic hydrodynamic conditions.

A Neural Model for Bout Initiation, Control, and Termination

A key scientific challenge is that intermittent swimming is not simply “continuous swimming with pauses”. It requires a controller that can:

1) Trigger a bout,
2) Shape the bout (frequency, amplitude, symmetry/asymmetry for turning), and
3) Stop the bout and transition into a passive glide.

Biological work suggests a division of labor:

• Spinal central pattern generators (CPGs) generate rhythmic motor patterns,
• Supraspinal gating mechanisms regulate when swimming starts/stops.

Inspired by this, we propose a neural controller that can generate a family of gaits by manipulating parameters such as:

• Tail-beating frequency and amplitude,
• Bout duration and glide timing,
• Left/right asymmetries to induce turning.

What We Measured: Kinematics and Energy

To move the debate on energetic efficiency from theory to evidence, we focus on two measurable outputs:

1) Behavior and kinematics

We compare robot trajectories and body kinematics against zebrafish-inspired targets:

• Forward bouts vs turning bouts,
• Evolution of head direction,
• Accumulated segment angles through the tail during each bout.

2) Power consumption and cost of transport

We quantify how much energy is required to travel a given distance under different strategies, comparing:

• Intermittent (bout-and-glide) swimming
  vs.
• Continuous tail-beating swimming

This allows a direct test of when intermittent swimming is beneficial.

Key Findings

Zebrafish-like intermittent gaits—generated by a mechanistic controller

By tuning neural-control parameters, ZBot produces diverse bout-and-glide patterns that qualitatively match zebrafish-like kinematics, including forward propulsion and controlled turning.

Intermittent swimming can reduce energetic cost at low speed

By recording both velocity and energy consumption, we show that bout-and-glide swimming achieves a lower cost of transport than continuous swimming at low velocities—supporting the idea that intermittency can be advantageous under specific operating regimes.

Why This Platform is Useful (Beyond This Paper)

ZBot turns a difficult biological question into an experimentally controllable one:

• It enables parameter sweeps (frequency, amplitude, duty cycle, bout timing) under matched conditions.
• It supports linking neural hypotheses to measurable outcomes: speed, turning, stability, and energy.
• It provides a bridge between neuroscience and robotics: testing how “neural design choices” shape embodied behavior.

Acknowledgements

Source of the cover image: Xiangxiao Liu, BioRob, EPFL
Complete author list: Xiangxiao Liu, François A. Longchamp, Luca Zunino, Louis Gevers, Lisa R. Schneider, Selina I. Bothner, André Guignard, Alessandro Crespi, Guillaume Bellegarda, Alexandre Bernardino, Eva A. Naumann, Auke J. Ijspeert.