Electronics > Electronics

Mimicking Walking Robot

<< < (2/2)

gellieramos:
we actually found a thesis about gesture mimicking, the thesis is about to mimic the master using camera, the two stereoscopic camera which serves as the eyes of the robot, the problem we are going to settle is about the recommendations/further work of the thesis, which is about how to mirror the gesture of the master, for example, on normal state, the master moves his right hand then the robot moves his right hand also, then in mirror state, the master moves his right hand then the robot moves his left hand, our professor wants us to contribute on that existing thesis, so we focused on the recommendation part, we have no idea how to start, the simulation, the mechanical part. hope you can help us, some ideas how to mimic using camera, thank you sir.. for ur kindness.. :)

newInRobotics:
As we don't know anything about original thesis You are to write recommendations for, it's not possible to advice what to recommend as an improvement over existing one. If You post a link to the original thesis, someone might want to read it and post what they think can be improved, although I wouldn't count on it, as analysing thesis is Your job, as You'll be the one to get rewarded for the work You do, hence not many people might be willing to do the job for You.

As for "where to start", read the original thesis, make sure You fully understand it, think about requirements raised by Your professor, brainstorm and write down ideas that arise, evaluate feasibility of ideas written down (are skills that You have are enough to to complete it, do You have enough time to complete it, do You have/can You get required resources to complete it). After all that You should have only few/one feasible idea/s making final choice easier.

Implementation phase comes after You have fully decided upon what You want to achieve and requirements raised by Your professor. Going with simulation/mathematical model only might be enough to satisfy Your educator and is going to be way less expensive than building physical robot.

To implement motion capture using stereo-vision You should be able to understand how computer vision works and how to implement algorithms required to track human body and decode limb joint angles. There is vary good C++/Java/Python library for that called OpenCV.

To implement simulation, one has to be quite good in OpenGL, or DirectX, or XNA, or any other GPU language to create graphics (a human model/skeleton) to mimic whatever is detected by motion capture system. For starters I wouldn't bother with full blown human, something like skeleton (image attached below) would suffice. I must add that there are software packages for human simulation like the ones from Siemens or Dassault. You university might have development access to these, hence You would not need to reinvent the bicycle. I know, that these packages are capable of taking input from motion capture systems and applying it to virtual manikins.

Jack Kinect integráció

To implement actual robot one is required to understand and have know-how of all of the above plus mechatronics: materials, physics, actuation and control systems. Again, You university might be able to provide You with a humanoid robot, so You don't have to reinvent the bicycle and focus fully on human mimicking algorithms.

Must add, that if You think of creating whole thing (motion capture system and robot) from a scratch, then You should reconsider, as each of the systems (and even sub-parts of them) is a serious thesis on its own requiring a lot of resources (time, money, knowledge and skills).

gellieramos:
thank you so much sir for all the advice.. that will help us a lot.. God bless you always sir, more powers to this forum.. :)

Navigation

[0] Message Index

[*] Previous page

Go to full version