Black version of the submarine in UFRJ Nautilus' logo.




Part of the Main-Hull that has the function of serving as an interface between the external part of the AUV and the internal part through subsea connectors. The belt itself also has the function of sealing by o'rings located in small cavities at its ends.


Structure used to join and charge all components that are not part of the Main-Hull, such as hydrophones, mechanical arm and battery attachment.

The robot's frame. A simple frame made with aluminum extrusions.
The gripper. A mechanical arm that oppens and closes its grippers.


Actuator used to perform object manipulation tests


Hydrophone arrangement used to capture the different ping signals at different frequencies. These signals, after being collected, serve for underwater geolocation.

An array of four hydrophones and a harness to attach it to the frame.
Kill switch render. A slide switch made of plastic with on and off indications.


Kill switch is the component responsible for the shutdown of the main system of the AUV. In case there is an emergency, it is possible to turn them off manually, using a switch and thus preserve the internal part of the robot.



FEA simulations were performed on the belt clamp to choose the fabrication material using Ansys software, analyzing the maximum deformations for each of the materials tested.

FEA simulation performed using Ansys software performed on the belt to confirm previously chosen material from its maximum total deformation.

Using the Ansys software, simulations were made with different width dimensions for the lateral arm and with the force applied in different places to analyze the total deformation and the equivalent stress in each case and thus, dimension the arm in the best way.

Simulations of the inner ring were made using the Ansys software in order to analyze the total deformation suffered with different dimensions and different materials to know if the inner ring would support the weight of the electronic components.





LUA's sensor management board, it is made as an embedded system using an ATMEGA 328P microcontroller. This system stores the data from the AUV sensors, the BAR 30 (external pressure sensor); the BMP180 (internal pressure sensor) and the Leak Sensor. On this PCB the three actuators are also driven: the torpedo, with a switching performed with a BJT NPN transistor; the marker launcher, with an L298N H-bridge; and the mechanical arm with the header outputs with the power supply and a PWM signal.


This PCB is responsible for controlling thrusters and is divided into three parts: power; control; protection. In the power part, there is an ATX24 connector responsible for receiving the necessary power to run the system. The control part is commanded by an ATMega 2560, where the code responsible for controlling the rotation of the thrusters is stored, which is electrically done by the ESC (Electronic speed control) after receiving a PWM signal. And in the protection part, current readings are taken so that the system can be shut down in case of overload.




PCB responsible for acquiring data from the measurements of the acoustic signals captured by the hydrophones. It is composed of the filter; butterworth topology and the signal amplification stages.


PCB responsible for the power distribution of the internal electronics. Here, the energy coming from the battery enters, and its regulation is done, with the decrease of continuous voltage to 12V, 5V and 3.3V. It is also responsible for the communication between all systems and the internal computer (Jetson TX2).





With the goal of creating a virtual environment similar to the real one where we test the robot, a replica of the pool in the Sports and Physical Education School (EEFD, in Portuguese) from our university was created in the 3D modeling software Blender, and then exported to Gazebo, where all the simulations with the AUV are done.


Besides the EEFD pool, a replica of the main pool where the competition takes place was also created. This simulation has many of the obstacles and tasks in the competition, such as the gate, buoys, the dropper and others. The files that launch the simulation also load the model of the robot and many of its resources and parameters. Different launch files were created so that each of the AUV’s functionalities could be tested individually, but there is a file that launches all of its features as well.



During part of the competition, the AUV is guided by two pingers. Basically the pingers are sound sources that produce an omnidirectional (in all directions) signal. Our robot possesses an array of hydrophones which can be used alongside the beamforming technique to identify the location of the sound sources, allowing it to navigate through the parts of the competition where it is guided by the pingers.


The neural network used in our robot is called Darknet, and built on top of it we use the object detection system YOLOv3-tiny. The neural network can be trained to identify different objects, such as buoys and the gate, for example, helping the robot orient itself. In a visual manner, the neural network acts by creating bounding boxes in the identified object, as can be seen in the image.

A 3D graph that shows the calculated trajectory the robot should take to go from a point to another.


PID (Proportional Integral Derivative) control is a technique for creating an appropriate response to a general error. This control methodology consists in the use of the information of proportional, integral and derivative error, in which each of them is associated with a different constant that defines the weight for each error. The constants can be tuned in many different ways. In our case, we use the “Ziegler-Nichols” method for this. Moreover, our error is nothing more than the subtraction of the goal position and current position of our robot.