I see you added more details now. That's great. I also struggled with touch sensors for my robots -- custom force sensors were too complex to make, simple endstop switches took too much force to switch (my robot is light). Then I discovered the "detector switches", which are usually used to see if an SD card is inserted. They require very little force, have small travel, and come in SMD versions that are very easy to attach to the end of a foot.

As for position sensing on servos, you might be interested in this log: https://hackaday.io/project/6050-tote/log/18992-proprioception-on-the-cheap

Detector switches.... You're right, I had them on my back legs. They detect the touch really easily, but when the leg is moving forward and backward the floor touches the case of the sensor, not the little stick. Maybe I've choosen the wrong sensor, but they were the only ones that I had ;) 

Thank You for the link, I've already read it and now I'm quite excited about it. Today I'll try to do something with the force detection for sure. 

I've heard about Your projects many times, and I can tell that's You're quite good at walking robots. So I was wondering if You could tell me what's the best way to move a leg (or then move all 4 ofthem)? Do you also use loops, where You change the coordinates or is there another solution ? What do You think of using sensors such as accelerometer or gyro ( kalman filter if i'm right ? ) ? 

I still haven't found the best way. As for not-best-but-working ways, it's a long story -- we can have a chat at #Robotics Chat or you can look at the source code of my robots -- it's all on Bitbucket. And my robots really lack sensors -- I have a lot of plans for trying them, but somehow never get the time/motivation. I hope I will get to this eventually. I would love to see more people experimenting with this, trying different things -- maybe that would motivate me to play with the sensors more.

I tried to use an accelerometer as a general body position and movement sensor, but I failed at math and pattern matching to really get this to work. The accelerometer I used was also pretty noisy, it may have been part of the problem. I want to try it again with a better sensor (the Bosh one that does data fusion for you) at some point, but I'm not in a hurry.

Touch sensors (switches) have been much easier to work with and much more promising. You can try to attach a piece of wire to the plastic nub that sticks out of the switch, and bend it in a loop -- this way it should always touch the floor.

There are also those distance sensors (infrared, ultrasonic and light ToF), which allow you to program a lot of nice behaviors into the robot, such as obstacle avoidance and following moving objects -- almost as if they had vision. Vision is also a huge topic that I didn't even start with. There is a lot to explore.

Looks very nice so far. Have you seen the Cheetah Cub robot? It looks somewhat similar to what you have there: http://biorob.epfl.ch/cheetah

Oh yeah I've seen this one ;) I actually came across it while doing a quick resarch in the internet about this type of robots. However I decided to use metal rods instead of strings, don't know if it's the best solution, will see ;)