Can you tell me what the difference in bases is? It seems that nao's xyz-coordination is different from OpenNI's figure. Did you use orthonormal bases to extract the movements?
@roidroid yeah, the discrepancy between simulated and real robot is way to high. This is the biggest problem at this point. This means we can't trust any simulated movement, because when played back on the robot it'll look different.
- Very good, what you need to do now for improving the artificial intelligent system (humanoid robot) is to add memory to the sequence (while using the kinect controller). So the robot will be able to do the task not only in real time but when needed (learn) and then after getting "memory" it is important to generalize behaviors or tasks. Saddly this robot have some limitations on the hands which will make limitations on programing also; but to start is not bad. Then you can use other robots
- It will be ideal if you could build a mobile robot ("sofa") to be able to trap NAO by the back of the legs. It will allow him to move faster around which can be useful for many tasks or games... and you will be able to sell this platform to NAO users or for other humanoid robots... it is also interesting because it will allow to transport batteries, additional "minicomputers" or a screen; for example. The soccer strategy for improving them is wrong in some terms because it is a human game...
A little disappointed as I was waiting for your repairing in dynamics of lifting certain foot. Anyway, it has been a good job. Enjoy it.
JAweiwei 8 months ago
Thank you for your response, @comandertwo, but I figured it out the other day. Great video, great work!
squid329 8 months ago
Can you tell me what the difference in bases is? It seems that nao's xyz-coordination is different from OpenNI's figure. Did you use orthonormal bases to extract the movements?
jaapsneep1 8 months ago
@jaapsneep1 You only have to rotate the coordinate systems. The orientation of Nao's coordinate systems is well described in the documentation.
comandertwo 8 months ago
Comment removed
squid329 8 months ago
@squid329 Sorry, but we don't use OpenCV. Was is the problem you're currently having?
comandertwo 8 months ago
it is strange that the on-screen virtual NAO is lagging more than the real NAO.
Arn't the real NAO's movements based on the virtual one?
roidroid 9 months ago
@roidroid yeah, the discrepancy between simulated and real robot is way to high. This is the biggest problem at this point. This means we can't trust any simulated movement, because when played back on the robot it'll look different.
comandertwo 9 months ago
- Very good, what you need to do now for improving the artificial intelligent system (humanoid robot) is to add memory to the sequence (while using the kinect controller). So the robot will be able to do the task not only in real time but when needed (learn) and then after getting "memory" it is important to generalize behaviors or tasks. Saddly this robot have some limitations on the hands which will make limitations on programing also; but to start is not bad. Then you can use other robots
luisbeck007 11 months ago
Comment removed
luisbeck007 11 months ago
- It will be ideal if you could build a mobile robot ("sofa") to be able to trap NAO by the back of the legs. It will allow him to move faster around which can be useful for many tasks or games... and you will be able to sell this platform to NAO users or for other humanoid robots... it is also interesting because it will allow to transport batteries, additional "minicomputers" or a screen; for example. The soccer strategy for improving them is wrong in some terms because it is a human game...
luisbeck007 11 months ago
Comment removed
luisbeck007 11 months ago
Comment removed
luisbeck007 11 months ago