Author Topic: Ideas for a master thesis  (Read 1957 times)

0 Members and 1 Guest are viewing this topic.

Offline RobotomanTopic starter

  • Beginner
  • *
  • Posts: 3
  • Helpful? 0
Ideas for a master thesis
« on: April 13, 2013, 12:51:20 PM »
Dear robotics society.

I'm looking for a good master thesis. I'm a student in Artificial Intelligence, so the research must be approached from a higher level (Machine Learning/AI). The craziest idea about object recognition/detection, learning from demonstration/observation or reinforcement learning/reinforcement learning is welcome.

Thanks for reading this.

Offline RobotomanTopic starter

  • Beginner
  • *
  • Posts: 3
  • Helpful? 0
Re: Ideas for a master thesis
« Reply #1 on: April 17, 2013, 02:50:39 PM »
***bump***
Anyone?

Offline jwatte

  • Supreme Robot
  • *****
  • Posts: 1,345
  • Helpful? 82
Re: Ideas for a master thesis
« Reply #2 on: April 17, 2013, 07:03:44 PM »
Build a rolling robot that can navigate through a crowd from point A to point B without bumping into people.

The input could be a known floor plan for the area, and the starting position and ending positions. The robot then does localization relative to the map, and detects people as dynamic obstacles and avoids them or waits until they pass.

This is not a robustly solved problem with good implementations, although there's good resource you can use to guide you. If you can contribute a robust implementation of this kind of behavior to some existing project (ROS or whatever) then that'd be of real value to the world.
« Last Edit: April 17, 2013, 07:05:40 PM by jwatte »

Offline RobotomanTopic starter

  • Beginner
  • *
  • Posts: 3
  • Helpful? 0
Re: Ideas for a master thesis
« Reply #3 on: May 02, 2013, 05:30:11 PM »
Build a rolling robot that can navigate through a crowd from point A to point B without bumping into people.

The input could be a known floor plan for the area, and the starting position and ending positions. The robot then does localization relative to the map, and detects people as dynamic obstacles and avoids them or waits until they pass.

This is not a robustly solved problem with good implementations, although there's good resource you can use to guide you. If you can contribute a robust implementation of this kind of behavior to some existing project (ROS or whatever) then that'd be of real value to the world.

I like that idea, more ideas like these are welcome!
And it doesn't have to be practical. It may be focussed more on research itself too. Remember that I am a computer science student (AI), not a mechanical engineer.

Offline jwatte

  • Supreme Robot
  • *****
  • Posts: 1,345
  • Helpful? 82
Re: Ideas for a master thesis
« Reply #4 on: May 02, 2013, 08:49:56 PM »
The mechanical part of rolling through a shopping mall is a well solved mechanical engineering problem. You can buy an electric wheelchair, or a radio-controlled car, or some simple rolling rover chassis, to get you going without needing to do any mechanical engineering yourself.

 


Get Your Ad Here