Underactuated Biomimetic Autonomous Underwater Vehicle for Ecosystem Monitoring

TL;DR

This work addresses the need for energy-efficient, maneuverable underwater sensors for ecosystem monitoring by proposing an underactuated, biomimetic fish-like AUV powered by a single motor to drive a cable-linked tail. The approach couples a novel mechanical design—comprising a cable-driven oscillating tail with a passive tail segment and bevel-gear transmission—with reinforcement-learning based navigation in the FishGym simulator, using a reduced action space to improve learnability. Key contributions include the mechanical implementation details and a learning framework for fish-like swimming in simulation, along with planned validations of thrust and navigation performance. The proposed system aims to enable robust monitoring in both marine and freshwater environments with simplified actuation and improved energy efficiency.

Abstract

In this paper, we present an underactuated biomimetic underwater robot that is suitable for ecosystem monitoring in both marine and freshwater environments. We present an updated mechanical design for a fish-like robot and propose minimal actuation behaviors learned using reinforcement learning techniques. We present our preliminary mechanical design of the tail oscillation mechanism and illustrate the swimming behaviors on FishGym simulator, where the reinforcement learning techniques will be tested on

Figures (2)

• Figure 1: (a) Design of the tail oscillation mechanism and its initial prototype is presented in (b)
• Figure 2: FishGym simulation illustrating various poses of the fish.