Weekly Report 02 (04.15. ~ 04.19.)

Overview

• Ni9elf-colab
• End-to-End Learning of Driving Models
• Refined Project Idea
• Plans for Week 03

Ni9elf-colab

• Read through the Ni9elf-colab Documentation
  • Summary in /docs/notes/Ni9elf-colab.md
• Relevance for the project:
  • Real time capable Object Detection from RGB-D data useful for object/floor type detection
  • Issues with small objects may not be applicable for mobile robot identifying objects in an enclosed space

End-to-End Learning of Driving Models

• Read through parts of End-to-End Learning of Driving Models with Surround-View Cameras and Route Planners
  • Summary in /docs/notes/EndToEndLearningOfDrivingModels.md
• Relevance for the project:
  • Datasets probably irrelevant
  • Driving Model Learning potentially useful
  • Cited sources may contain useful information

Refined Project Idea

• Object/Floor Detection and Path Planning for Mobile Robots
  • Floor Type & (Canny) Edge Detection for Environment Mapping
  • Floor Type & Object Detection to find target waypoints for the mobile robot
  • Mobile Robot Driving Model
    • Task Planning and Prioritisation Module
    • Route Planning Module setting waypoints in the virtual environment map and adjusting route for obstacles
    • low-level navigation routine driving to the next/highest priority waypoint

Plans for Week 03

• Read remaining sources
  • End-to-End Learning of Driving Models with Surround-View Cameras and Route Planners
  • End-to-End Navigation with Branch Turning Support using Convolutional Neural Network
  • Taskonomy: Disentangling Task Transfer Learning
  • LayoutNet: Reconstructing the 3d Room Layout from a Single RGB Image
• Create first draft of mission statement