Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Leader-follower formation enabled by pressure sensing in free-swimming undulatory robotic fish

Kundan Panta, Hankun Deng, Micah DeLattre, Bo Cheng

TL;DR

This work tackles enabling leader-follower formation in free-swimming undulatory robotic fish using flow-pressure sensing inspired by the lateral line. It integrates two bilateral pressure sensors on the follower, a Kuramoto-based gait generator, and an end-to-end imitation-learning controller (LSTM) trained with Behavior Cloning and DAgger to map pressure history and onboard state to steering. The follower achieves formation maintenance up to $200$ mm away at speeds around $155$ mm/s using less than an hour of training data. The study demonstrates the potential of fish-inspired robots to navigate fluid environments under flow feedback and provides a data-driven path toward robust, sensor-driven formation control in aquatic robots.

Abstract

Fish use their lateral lines to sense flows and pressure gradients, enabling them to detect nearby objects and organisms. Towards replicating this capability, we demonstrated successful leader-follower formation swimming using flow pressure sensing in our undulatory robotic fish ($μ$Bot/MUBot). The follower $μ$Bot is equipped at its head with bilateral pressure sensors to detect signals excited by both its own and the leader's movements. First, using experiments with static formations between an undulating leader and a stationary follower, we determined the formation that resulted in strong pressure variations measured by the follower. This formation was then selected as the desired formation in free swimming for obtaining an expert policy. Next, a long short-term memory neural network was used as the control policy that maps the pressure signals along with the robot motor commands and the Euler angles (measured by the onboard IMU) to the steering command. The policy was trained to imitate the expert policy using behavior cloning and Dataset Aggregation (DAgger). The results show that with merely two bilateral pressure sensors and less than one hour of training data, the follower effectively tracked the leader within distances of up to 200 mm (= 1 body length) while swimming at speeds of 155 mm/s (= 0.8 body lengths/s). This work highlights the potential of fish-inspired robots to effectively navigate fluid environments and achieve formation swimming through the use of flow pressure feedback.

Leader-follower formation enabled by pressure sensing in free-swimming undulatory robotic fish

TL;DR

This work tackles enabling leader-follower formation in free-swimming undulatory robotic fish using flow-pressure sensing inspired by the lateral line. It integrates two bilateral pressure sensors on the follower, a Kuramoto-based gait generator, and an end-to-end imitation-learning controller (LSTM) trained with Behavior Cloning and DAgger to map pressure history and onboard state to steering. The follower achieves formation maintenance up to mm away at speeds around mm/s using less than an hour of training data. The study demonstrates the potential of fish-inspired robots to navigate fluid environments under flow feedback and provides a data-driven path toward robust, sensor-driven formation control in aquatic robots.

Abstract

Fish use their lateral lines to sense flows and pressure gradients, enabling them to detect nearby objects and organisms. Towards replicating this capability, we demonstrated successful leader-follower formation swimming using flow pressure sensing in our undulatory robotic fish (Bot/MUBot). The follower Bot is equipped at its head with bilateral pressure sensors to detect signals excited by both its own and the leader's movements. First, using experiments with static formations between an undulating leader and a stationary follower, we determined the formation that resulted in strong pressure variations measured by the follower. This formation was then selected as the desired formation in free swimming for obtaining an expert policy. Next, a long short-term memory neural network was used as the control policy that maps the pressure signals along with the robot motor commands and the Euler angles (measured by the onboard IMU) to the steering command. The policy was trained to imitate the expert policy using behavior cloning and Dataset Aggregation (DAgger). The results show that with merely two bilateral pressure sensors and less than one hour of training data, the follower effectively tracked the leader within distances of up to 200 mm (= 1 body length) while swimming at speeds of 155 mm/s (= 0.8 body lengths/s). This work highlights the potential of fish-inspired robots to effectively navigate fluid environments and achieve formation swimming through the use of flow pressure feedback.

Paper Structure

This paper contains 16 sections, 2 equations, 10 figures.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Integration of Pressure Sensors
  4. Experimental Setup
  5. Generating Forward Swimming and Steering Gaits
  6. Random Path Generation for the Leader
  7. Imitation Learning
  8. Expert Policy
  9. Learner Policy
  10. Behavior Cloning
  11. DAgger
  12. Training the Learner
  13. Experimental Methodology
  14. Evaluating Leader-Following Performance
  15. Results
  16. ...and 1 more sections

Figures (10)

  • Figure 1: The follower µBot steers towards the leader, informed by its pressure sensors and IMU, which encode the hydrodynamic interactions between the two µBots.
  • Figure 2: (a) The follower µBot with a pressure sensor integrated on each side of its head module. (b) The modular PCBs onto which the pressure sensors are mounted, connected serially to the microcontroller and body modules via flat-flex cable (FFC) connectors. (c) The still water tank used for the leader-follower experiments.
  • Figure 3: Generation of the leader's random path and the expert follower's staggered path. The maximum allowable distance between the leader's tail and the follower's nose, $d$, during leader-following is capped at 200 mm.
  • Figure 4: Pressure signals at various lateral and longitudinal distances between a fixed follower µBot and a free-swimming leader µBot moving from rest. The distances refer to the initial distance between the leader's tail and the follower's nose. The leader was to the follower's right in the staggered formations.
  • Figure 5: The two imitation learning approaches: (a) Behavior Cloning (BC) and (b) DAtaset Aggregation (DAgger).
  • ...and 5 more figures