StateSpace Control and Python Bindings for the Amphibious AQUA robot

1
State-Space Control and Python Bindings for the
Amphibious AQUA robot

• by
• Gabriel Charette
• supervised by Gregory Dudek

2
Outline

• Problem description, goal, motivation
• Control system infrastructure
• State-space control, a new concept
• Experiments
• Future work
• Conclusion

3
Problem Description

• Complex Application Programming Interface(API)
  needed to control the robot
• Hard to add new features to the API without prior
  knowledge
• API depends on a lot of external libraries

4
Goal

• Wrap the current API
• Provide access to all of the robot's control
  functions (roll, pitch, yaw, ...)
• Provide access to all of the robot's outputs
  (inertial measurement units (IMUs), cameras, ...)

5
Motivation

• Provide ability to write scripts controlling the
  robot at a high abstraction level
• Provide ability to add to the current system
  without modifying it
• Provide a user-friendly system to control the
  robot and receive feedback from it.

6
Related Work

• SWIG, C to Python Wrapper, www.swig.org
• Mobile Robotics Lab
• A Vision-based Control and Interaction Framework
  for a Legged Underwater Robot
• Junaed Sattar and Gregory Dudek
• Proceedings of the Sixth Canadian Conference on
  Robot Vision (CRV), May 2009, Kelowna, BC,
  Canada, pages 329-336
• Aqua An Amphibious Autonomous Robot
• Gregory Dudek, Philippe Giguere, Chris Prahacs,
  Shane Saunderson, Junaed Sattar, and Luz-Abril
  Torres-Mendez, McGill University, Michael Jenkin,
  Andrew German, Andrew Hogue, Arlene Ripsman, and
  Jim Zacher, York University, Evangelos Milios,
  Hui Liu , and Pifu Zhang, Dalhousie University,
  Martin Buehler, Boston Dynamics, Christina
  Georgiades, MathWorks.
• A Visual Language for Robot Control and
  Programming A Human-Interface Study
• Gregory Dudek, Junaed Sattar, Anqie Xu

7
Choosing a language why Python?

• Very modern high-level programming language
• No need for compilation (easy to quick fix small
  bugs in the field)
• Highly-portable
• Different operating systems
• Two operating systems on board

8
Control System Infrastructure
System Independent Interface
Any Offboard Script Python
Any Script Python
PathGenerator Python
robocontrol module Python module
PathGenerator GUI Python
RoboControl Server Python
robomodule visionmodule C wrapper Python
module
robochat_module Human-Robot Control Interface C
wrapper Python module
VisionSandBox C
RoboDevel C
Robot Machine
Offboard
Onboard
9
Path Generator
10
State-Space Programming

• A new programming paradigm!
• Code executed based on state and position in
  space rather than following the regular flow
• Global vs local script

11
State-Space Control

• Allows for important tests to be made before
  continuing with path
• Allows to load a new path on a given state-space
  event
• Allows for manipulation of control components
  such as speed depending on situation faced at a
  particular node

12
Python Robot Control Live

• Successfully controlled AQUA using python ONLY!
• Robot controlled from python prompt using
  robomodule directly
• Robot controlled from internet using TCP
  communication with robocontrol_server through the
  robocontrol python module

13
State-Space Control - Experiment

• On end of path, take left or right depending on
  power status
• Proof of concept
• Tested live in water

14
Live Test - Video
15
Future Work

• Add more functionality to the different modules
  used by the robocontrol server
• Immediately benefits the top layer
• Provide GPS to the robot
• Add precision to the path generator
• Allow for state-space control anywhere
• Implement this control system in other robots
• Easy as top layer is only python dependent!

16
Conclusion

• Built a new user-friendly API to control AQUA
• Highly flexible and portable system
• New programming paradigm state-space control
• All validated both in laboratory and on field!

17
Some code comparison!
In VisionSandbox or RoboDevel??
State-Space Control Path Code

• It would take at least 100 lines of code to do
  this in VisionSandbox
• It would be incredibly painful to do this in
  RoboDevel

• global i
• takePicture("Pictures/picd.jpg" (i))
• ii1
• -!!-
• -2.82,3.28,0
• -!!-
• 2.72,3.06,0
• -!!-
• -2.76,-2.66,0
• -!!-
• 2.94,-3.12,0
• -!!-
• 0.04,-0.04,0
• -!!-
• Note This is generated by GUI too!!