RoboBoat Sprint TimeLine

MAY 2021

Boat:

Simulation:

  • Mandatory Navigation Channel completed

  • Speed Gate completed

  • Obstacle Channel completed

  • Obstacle Field completed

  • Acoustic docking completed in RViz

Mechanics - Electronics:

  • Diagram and plotting of routes in complete pcb.

Computer vision:

  • Fixed LiDAR perception system, now working on simulation with challenges.

  • Dock Corner beta detection system

  • PointRCNN

  • Projection and rigid body transformation matrix

  • Synched pointcloud and image capture​

  • YOLACT

  • Training / Test

Research:

  • MPC-based collision avoidance completed

April 2021

Boat:

Overall:

  • Scripts for the videos started development

Simulator:

  • Props started to be added to simulation world

Mechanics - Electronics:

  • Most PCB schematic completed

  • Component documentation and justification for their choice

Embedded Systems:

  • STM32 is going to be changed

  • Check for PCB

  • Advances in the FreeRTOS code (Engines, Depth sensor, sensor leak)

Computer vision:

  • Testing object detector & Sensor fusion with moving boat.

  • Tested PointRCNN

  • Labelled semantic segmentation dataset.

  • Trained Yolact

Research:

  • NMPC chosen as guidance

  • Implementation and simulations in ROS

March 2021


Dron:

The entire project was redefined. The Dji Phantom 4 Pro will no longer be used, a new one will be developed for RoboBoat 2022.


Boat:

RoboBoat 2021 goes online

Simulator:

  • VTEC simulator completed. It relies on the boat dynamic model for state simulation, a LiDAR sensor and a stereo camera

Software:

  • Overview reglas

  • Overview of components

  • Download the report

Mechanics - Electronics:

  • It was decided that the same PCBs would be used on the ship and submarine

  • Component decision

  • Design of the source, motor connections and microcontroller in the schematic

  • Component bases

Computer vision:

Late sensor fusion:

  • YOLO + Lidar Object Det.

  • Objects with YOLO color + Lidar Localization

  • Reviewing state-of-the-art methods and comparing models

YOLACT:

  • Understand instance segmentation

  • Prepare datasets to benchmark YOLACT

GANs dataset:

  • Introduction Course

Research:

  • MPC for speed and angle control from USV done

  • MPC for USV route tracking done

February 2021

  • Summary of previous efforts

  • General objectives defined

November 2020

Drone:

  • DJI SDK 4.13.1 Update.

  • Refactorization analysis for new app.


Boat:

Simulador:

  • Quite difficult to simulate correct thrust with methods proposed by simulator

  • Simulator discarded to opt to develop our own approach

Electronics:

  • PCB Modifications

  • FreeRTOS:

  • Task definition based on electronics bill of materials

  • Implementation of thrusters

  • Base skeleton of the submarine's engines (based on the Ship)

Computer vision:

  • Established point types

  • Functions Done:

  • Passthrough Filter

  • Radius Filter

  • CropBox Filter

  • VoxelGrid Downsampling

Mechanics:

  • Square battery base

  • Cylindrical battery base

  • Driver base

Research:

Collision Avoidance:

    • DRL:

      • Definition of Markov Decision Process (MDP)

      • Programming:

        • Environment

        • Rewards

    • MPC:

      • Definition of the optimal control problem (OCP)

        • Cost function

        • Constraints

      • Installation of acados

      • Use of acados from an independent directory

    • Overall:

      • Done:

        • New paper presented at OCEANS 2020

        • ICRA 2021 submission from collaboration with “Laboratorio de Robotica”

      • To do:

        • Define research projects for next semester

March 2021

Dron:

  • The entire project was redefined. The DJI Phantom 4 Pro will no longer be used, a new one will be developed for RoboBoat 2022.


Boat:

Software

  • Roboboat 2021 Rules were reviewed

  • Components review

  • Repository download

Mechanics/Electronics

  • What components to use were decided

  • Source design, motor and microcontroller connections established in the schema

  • Component bases were designed


October 2020

Drone:

  • Joystick Android App Implemented


Boat:

Electronics:

  • Second iteration PCB VTec

Mechanics:

  • Square battery base

  • Cylindrical battery base

  • Driver base

Software:

  • Xbees Topic connections established successfully

Hydrophones:

  • DOA Algorithm simulation

  • Separation between Hydrophones defined (1.8 - 3cm) because of frequency variation

Comms:

  • Achieve connection of Bacchus software topics between xbees

FreeRTOS:

  • Arduino to STM32 code translation completed

  • Creation of a new repository for current code

  • Feature/Pulse IN

  • Code Standard defined for Embedded Systems

Simulator:

  • Boat deployed in simulation

  • Difficulties in setting up the simulator, not all the parameters are known

  • Boat stability is not always achieved

  • There is a lot of uncertainty as to whether the simulation achieved is realistic

Computer vision:

  • BrainStorming: Current and future needs

  • Research: Point Cloud preprocessing strategies

  • Define point cloud preprocessing class template

Research:

Collision Avoidance:

  • Velocity obstacle programming completed

MPC related:

  • Literature review completed

  • Problem definition, scope still needed

September 2020

Drone:

  • New DJI SDK was tested with ROS. However, it will not be used at the moment.


Boat:

Electronics

  • VTEC PCB Second iteration

  • STM Integration: 3 - 3.5V Conversors

Mechanics

  • Squared and Cylinder Bases, as well as Driver Bases, were designed

Comms

  • Making sure Xbees are connected prior to starting coding

Simulator

  • UUV simulator chosen for the moment.

  • Get familiar with the simulator

  • Deploy the ship and submarine

Research:

  • Objectives defined:

  • Development of basic obstacle avoidance algorithm based on Velocity Obstacle (VO) for USV.

  • Development of an obstacle avoidance algorithm based on predictive control (MPC) for USV.

  • Development of an obstacle avoidance algorithm combining adaptive control (ASMC) and Deep Reinforcement Learning (DRL) for USV.

  • MPC-based controller development for USV.

  • MPC-based controller development for UUV.

  • Optimal tuner implementation in USV.


Collision Avoidance:

  • Done: Pipeline for the VO method.

  • Doing: VO implementation


MPC related:

  • Doing: Research related literature (state of the art).

August 2020

Dron:

  • General summary of progress in previous editions.

  • General description and establishment of work areas.

Boat:

  • Summary of previous efforts

  • General objectives defined

'20

  • Development of an adaptive sliding mode controller for speed and heading control

  • Implementation of a time-varying look-ahead distance line-of-sight guidance law for path-following control

  • Improvement of the USV-station communications

  • Reorganization of the overall software architecture of the system

  • Research in AI methods for guidance and control

  • Development of 3D vision-based guidance methods for RoboBoat task solutions

  • Development of a simulation environment for digital twin testing

  • Collision avoidance strategy based on velocity obstacle method