Nerites Underwater Simulator

ROS 2 Foxy Simulation, Control, and Reinforcement Learning for Underwater Robotics

---

Motivation

Nerites Underwater Simulator is a complete ROS 2 Foxy port and extension of the original UUV Simulator framework:
https://github.com/uuvsimulator/uuv_simulator

The project focuses on:

• Full ROS 1 to ROS 2 Foxy migration of underwater simulation tools
• Support for realistic AUV and ROV vehicle models
• Reinforcement learning for underwater control
• Underwater manipulation with Oberon robotic arms
• Multi-vehicle, sensor-rich simulation environments

---

Included Vehicles

• RexROV -- Work-class ROV
• ECA A9 -- Torpedo-style AUV
• GRAALTech X300 -- Hybrid AUV/ROV
• Oberon4 and Oberon7 manipulators
• Gym-compatible RL framework with Soft Actor--Critic (SAC) training pipelines

---

Features

Underwater Worlds

Nerites ports and extends the original UUV Simulator environments to ROS 2:

• Ocean environments
• Pool and tank setups
• Hydrodynamic and buoyancy models
• Current, turbidity, and noise plugins
• Structured benchmarking environments

---

Supported Vehicle Configurations

---

Vehicle	Description
RexROV	Work-class ROV with full thruster configuration
ECA A9 AUV	Torpedo-style autonomous underwater vehicle
GRAALTech X300	Hybrid AUV/ROV with hovering and profiling capability
RexROV + Oberon4	ROV with 4-DOF manipulator
RexROV + Oberon7	ROV with 7-DOF serial arm
RexROV + Dual Arms	ROV with both Oberon4 and Oberon7

Overview of underwater vehicles: Top left: RexROV; Top middle: ECA A9 AUV; Top right: GRAALTech X300; Bottom left: RexROV + Oberon4; Bottom middle: RexROV + Oberon7; Bottom right: RexROV + Dual Arms (Oberon4 and Oberon7);

---

Manipulators: Oberon4 and Oberon7

Arm	DOF	Description
Oberon4	4	ROV-mounted arm with jaw gripper
Oberon7	7	Serial arm with wrist and parallel gripper
Dual-arm setup	11	Combined Oberon4 and Oberon7 on RexROV

Control and Kinematics Support

• gazebo_ros2_control integration
• joint_state_broadcaster
• Position and effort controllers
• Gripper control (jaw and parallel)
• Forward and inverse kinematics services

---

GRAALTech X300 Hybrid AUV/ROV

The GRAALTech X300 is a compact torpedo-shaped hybrid AUV/ROV designed for:

• Long-range autonomous navigation
• Precision hovering and station-keeping
• Environmental monitoring and inspection

In Nerites, X300 is fully modeled in ROS 2 with:

• Thruster allocation
• Hydrodynamic modeling
• Sensor integration
• A dedicated SAC RL training environment (x300_env.py)

Simulated Actuation

• Two rear main thrusters for surge
• Four vertical tunnel thrusters for heave
• Four horizontal tunnel thrusters for yaw and lateral motion

---

Controllers

RexROV Trajectory Controllers (Verified in ROS 2 Foxy)

Launch Gazebo world:

ros2 launch uuv_gazebo_worlds auv_underwater_world.launch.py

Launch RexROV:

ros2 launch uuv_descriptions upload_rexrov_default.launch.py

Available controllers:

ros2 launch uuv_trajectory_control rov_pid_controller.launch.py uuv_name:=rexrov
ros2 launch uuv_trajectory_control rov_mb_fl_controller.launch.py uuv_name:=rexrov
ros2 launch uuv_trajectory_control rov_mb_sm_controller.launch.py uuv_name:=rexrov
ros2 launch uuv_trajectory_control rov_nl_pid_controller.launch.py uuv_name:=rexrov model_name:=rexrov
ros2 launch uuv_trajectory_control rov_pd_grav_compensation_controller.launch.py uuv_name:=rexrov model_name:=rexrov
ros2 launch uuv_trajectory_control rov_sf_controller.launch.py uuv_name:=rexrov
ros2 launch uuv_trajectory_control rov_ua_pid_controller.launch.py uuv_name:=rexrov
ros2 launch uuv_trajectory_control rov_nmb_sm_controller.launch.py uuv_name:=rexrov model_name:=rexrov

Send a waypoint file:

ros2 launch uuv_control_utils send_waypoints_file.launch.py uuv_name:=rexrov

---

Disturbance and Thruster Tutorials

ros2 launch uuv_tutorial_disturbances tutorial_disturbance_manager.launch.py
ros2 launch uuv_tutorial_disturbances tutorial_body_wrench.launch.py
ros2 launch uuv_tutorial_disturbances tutorial_thruster_efficiency.launch.py
ros2 launch uuv_tutorial_disturbances tutorial_thruster_state.launch.py
ros2 launch uuv_tutorial_disturbances tutorial_timed_current.launch.py

---

Reinforcement Learning Environment

Base environment: - base_env.py

• reset() and step() API
• Continuous action spaces
• Rich observation spaces

Vehicle-specific environments:

• RexROV → rexrov_env.py
• ECA A9 → eca_a9_env.py
• X300 → x300_env.py

A custom Soft Actor--Critic (SAC) implementation is included for full-DOF control training.

---

Training SAC Controllers

1. Select target vehicle: rexrov, eca_a9, or x300

2. Set configuration path in:

SETTINGS_PATH = "settings/settings_rexrov.txt"
SETTINGS_PATH = "settings/settings_eca_a9.txt"
SETTINGS_PATH = "settings/settings_x300.txt"

3. Configure:

• Episode length
• Steps per episode
• Observation and action scaling
• SAC hyperparameters

4. Run training:

python3 custom_algorithms/my_sac_train.py

---

Requirements

• ROS 2 Foxy
• Gazebo Classic 11
• Python 3.8 or newer
• Colcon build tools
• kdl_parser_py

---

Installation

git clone <your-repo-url> ~/nerites_sim_ws
cd ~/nerites_sim_ws/src
rosdep update
rosdep install --from-paths . --ignore-src -y
cd ~/nerites_sim_ws
colcon build
source install/setup.bash

---

Contributing

Contributions are welcome in:

• New vehicle models
• Sonar and sensor simulation
• Manipulation and dual-arm coordination
• Reinforcement learning benchmarks
• Documentation and tutorials

---

Citation

To be announced.

---

Parent Repositories

• ROS2 X300 single Vehicle Implementation: (https://bitbucket.org/isme_robotics/auv_core_ros2/src/main/)
• UUV Simulator: (https://github.com/uuvsimulator/uuv_simulator)
• Plankton Simulator: (https://github.com/Liquid-ai/Plankton)
• Eca A9 vehicle: (https://github.com/GSO-soslab/eca_a9_plankton)
• UUV manipulators: (https://github.com/uuvsimulator/uuv_manipulators)