Sonar Based SLAMC (Simultaneous Localization, Mapping and Classification)

Contacts

Itamar Eliakim, Editor
Dr. Yossi Yovel, Advisor
Dr. Gabor Kosa, Advisor

Abstract

For the last 52 million years bats use echo-location for navigation, localization and classification purpose. Agrirobot, innovative approach to agricultural technology using bio-inspired SONAR, based on similar acoustic signals that bats transmit we can apply different methods for solving the well-known SLAM problem. Focusing at the on the robotic orientation aspects of the Agrirobot, developing a SONAR based method for robotic- mapping, obstacle avoidance and path planning in a greenhouse or an orchard. This will allow the yield-assessment robot to autonomously navigate in the greenhouse or orchard based on bio-SONAR only. Such an ability is an essential step on the way to developing a fully automatic yield assessment approach which will be far cheaper and more accurate than the all measures which are currently in use.

Thesis Outline

• Introduction
  

• Hardware & Software
  

• Passive Sonar Localization (BAT-GPS)
  

  • Basics of high frequency chirp signals
    
  • Transmit, receive signal using independent systems
    
  • Cross correlation between signals
    
  • TDOA – Time Difference Of Arrival Methods (Chan, Foy, Grid-Search)
    

• Active Sonar Navigation
  

  • Construct 2D map base on acoustic signals
    
  • Iterative Obstacle Inflation
    
  • Advantages of ultrasonic signals, use of one and two ears
    

• Path Planning and Classification
  

  • Classification of signals using Neural Network - TensorFlow
    
  • Path planning based ROS Navigation Stacks
    
  • Decision making - Cul-de-Sac
    

Hardware

• Komodo Robot Platform
• DJI Ronin Gimbal + Custom Control Board link:
  https://github.com/Itamare4/dji_ronin
• Speed of sound calibration
• Passive Localization Unit:
  
  • 4x Avisoft Vifa Speaker
    
  • 1x DAQ - Measurement Computing - USB-1608GX-2AO
    
  • 1x Avisoft Bioacoustics CM16/CMPA40-5V - Omni microphone
    
  • 1x Sony XM-GS4
    

* Front Sensing Unit:
* 1x Avisoft Vifa Speaker
* 1x DAQ - Measurement Computing - USB-1608GX-2AO
* 2x Avisoft Bioacoustics CM16/CMPA40-5V - Directional microphone
* 1x Sony XM-GS4
* 1x Thermal Camera - Flir Ax5
* 1x RGB Camera - uEye CP
* 1x Leica D410

Software

• Ubuntu 14.04 LTS
• ROS Indigo

Algorithm Architecture

Experiments

Experiments done in 3 different locations: Robotics LAB - Wolfson Building, Tel Aviv University, first experiments done indoor mapping of the lab environment, mapping dimensions - (6m x 4m), this set of experiments done based on one ear mapping, will explained below, at each point the robot collects chirp signals at 5 different angles, [-90°,-45°,0°,45°,90°], second set of experiments done at pteridophytes greenhouse at Botanical Garden, Tel Aviv University, mapping dimensions (5m x 12m), mapping based on two ears(ITD), at each point the robot collects chirp signals at 3 different angles, (0,-60, 60). third set done at the palm greenhouse at Botanical Garden, Tel Aviv University, mapping dimensions (40m x 5m), data collections similar to experiments done at the pteridophytes greenhouse.

Palm Greenhouse, Botanical Garden	Pteridophytes Greenhouse, Botanical Garden	Wolfson Building, Mechanical Engineering Faculty

Results

Sonar Localization, Mapping and Classification, with 3 iteration of IOI, iterative obstacle inflation.
link - https://github.com/Itamare4/ROS_smooth_map

Palm Greenhouse Top View, Ground Truth Based on Aerial Footage Using DJI Phantom 4

Videos

Photo Album

Photos from outdoor, indoor experiments are available on:
link:

Publications & Citations

▪ Eliakim I., Yovel Y., Kosa G., “Acoustic Self-Localization for Mobile Robots” IPIN (Indoor Positioning Indoor Navigation) 2016, Madrid, Spain
▪ Soon.

About

Itamar Eliakim
M.Sc Student Mechanical Engineering Faculty at Tel Aviv University, Israel
Email - [email protected]