Robotics: How can I start with my thesis on Bipedal Humanoid Robot? What basic I should learn first?

Not strictly Bipedal, but I suspect you'll find http://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-832-underactuated-robotics-spring-2009/index.htm to be relevant.

Your school's library (or their library website) can be a great resource for finding research articles about different types of robotic technologies. Additionally, I would highly recommend finding any professors at your school that do research in bipedal robotics and either e-mail them or drop by their office to ask about it. Many professors love when students come and talk to them because they are excited about learning more or getting into research. Someone who works in the field would be able to recommend good beginner resources, and they might have projects that you can get course credit to work on.

https://docs.google.com/viewer?a=v&q=cache:Mz7dGT93uI0J:www.stevens.edu/msrobotics/SMRDC2010/muecke06r.pdf+&hl=en&gl=us&pid=bl&srcid=ADGEEShLZF-5dTeGaJaf-IdpzhdEdasslkSWF5UkBdmdIk0yO8naLhO5DvgXSJiQHB0XA0LQNgZIUQWfKV_aGLQj7sgEaFSeV5vFNcuUVuw-pkScHyo8OuwSsdYDWEgSZBXIjsVskq3l&sig=AHIEtbQvjGetwqVSgiDQPv4VgTjwffI2UQ This is a good start to see previous examples of bipedal humanoid robots created in universities

http://techtomove.com has published some interisting articles about human legs movement. who can start from that: http://techtomove.com/2012/10/17/video-to-measure-walk-joint-angles/ http://techtomove.com/2012/11/14/legs-movement-chart/ http://techtomove.com/2012/10/12/design-software/

Some of the answers here have mentioned about starting with biology and by observing your own walking. I'm doing a thesis now on humanoid robots as well and while observing your own walking is definitely insightful at times, it's not as simple as that. In reality, human walking is not fully understood even today. The way our muscles work to execute walking is very different from using motors. Most humanoid robots work under a lot of assumptions that make their gait different from humans (ex: fixed height of CoM), but still manage to have a dynamic walking gait (while consuming a LOT more energy than humans). I would suggest you first read up on the human walking cycle, i.e. Single Support and Double Support phases. Next, look up mathematical models of human walking. A very common model is the 3D Linear Inverted Pendulum Mode. From there you could start reading on the Zero Moment Point criteria for dynamic stability. I personally found it quite tricky and challenging to actually understand the Zero Moment Point concept even after going through multiple books and papers, primarily because the ZMP is not really a textbook concept but was introduced by Vukobratovic nearly 40 years ago in a research paper and a lot of people seem to keep coming up with their own interpretation of the concept. I would recommend you start with Introduction to Humanoid Robotics by Kajita. An English version of the book has been available since 2013. Also do a Google search for Master/PhD thesises on biped walking. They usually go into much more depth and explanations than research papers. And for understanding ZMP, a good place to start along with the book by Kajita would be this paper called Zero Moment Point: Thirty Five Years of its Life. Biped Walking is still a very challenging research area. If you ever happen to start feeling bored or low on motivation, then watch the DARPA Robotics Challenge videos ;) All the Best!

I would start with the biology.  How do humans and animals walk?  The physics behind it.  The next step would be a simplified mathematical model that would be practical for use in a robot.  I would then look at other ways to simplify the design.  Counter weights for balance, etc.  Something that would require a lot fewer computations than the way that we walk, but that would achieve a good end result. A airplane uses the same physics as a big, but it doesn't need flapping wings.  Early on it became clear that idea was impractical.  We went with a fixed wing design instead.  It works just as well, and is far simpler to build.  The same might be true here.  Trying to move body parts into the correct position to keep balanced is likely to be overly complex and unnecessary.  Just come up with a simpler automatic balancing mechanism that will allow from simpler movements of the legs and body.