Differences

This shows you the differences between two versions of the page.

--- ires_journal9 [2018/08/05 20:39] – [Korean Cultural Insight] yuhanghe
+++ ires_journal9 [2018/08/06 02:27] (current) – [Project Status] yuhanghe
@@ Line 26: / Line 26: @@
 ==== What I Learned about Myself ====
-Staying in Korea constantly prompt me to be reminiscent of my childhood in China. Many aspect of Korean culture, from food to tradition, appears to me as quite similar to Chinese culture. The similarity reminded me how much I missed my birthplace. I was too concentrated on my daily life in American before to think back to my childhood and birthplace. Even though it has been 8 years since I left my birthplace, I still hold many fond memory of my old home.
+The traveling last week reminded how much fun traveling is. Unfortunately, I did not find many opportunities to travel during this internship at Hubo Lab. If there is one thing I regret about this trip, it would be that I should plan more traveling trips. I learned that I really like learning and experiencing new things.
 ==== Project Status ====
-Last week, I implement the inverse kinematic function and created PODO AL that will transform mocap data to joint data and execute the motion. This week, I focused on filtering the data and implementing joint angle, velocity, and acceleration limit on the output joint data.
+After implementing the joint angle and velocity limits last week, I focused on enabling the upper body imitation this week. The previous tests and programs only focused on right arm only. So this week, I modified both the motion capture algorithm and PODO AL programs to capture motions from both arms and imitating these motions on Hubo 2.
-The implementation of joint limit turn out to be more of a challenge then I anticipated. The simplest method to apply joint angle constraint is to limit the angles once they exceed maximum or minimum. The simplest method to limit joint angle velocity and acceleration is to apply a time warp to slow down motion within velocity and acceleration limit. However, both method are not suitable for the project because this project aim at preserving as much original characteristic of motion as possible.
+[[https://youtu.be/SEt20aL732g|Youtube Link]]
-In order to accomplish this, I decided to implement the filters developed by [[https://ieeexplore.ieee.org/document/1014737/|Pollard et al]]. Here are the result of implementation.
+{{:yuhang:ires2018:demo1.mp4|}}
-Joint angle limit:\\
+As seen from the video, the imitation is not perfect. In the beginning, the left shoulder yaw motor rotated to an undesirable position. That is because the left arm was outside of mocap area in the beginning. Another major difference is the speed of motion. The joint angle data was calculated at 0.010 ms. The real time thread on Hubo 2's PC runs at 0.010 ms interval. Some delay between actual execution of motion accumulated overtime and resulted in the delay.
-{{ :yuhang:ires2018:screenshot_from_2018-07-30_10-50-18.png?700 |}}
-The blue lines are the original joint angles and the green lines are joint angles after applying joint angle limit and filtering. The algorithm is not perfect, however, it was able to limit the joint angles while preserving some characteristics of original motion.
-Joint velocity limit:\\
-{{ :yuhang:ires2018:screenshot_from_2018-07-30_10-20-20.png?700 |}}
-Again, the blue lines are the original joint angles and green lines are joint angles after applying joint velocity limit and filtering. One problem I noticed for this algorithm is that, for relatively small amount of velocity exceed the limit, the algorithm worked very well. However, once the velocity and acceleration exceeds the limit, the filter will result in a flat angle.
-Another problem I noticed is that the first set of data I collected using mocap system are inaccurate and inconsistent. I need to collect more data from mocap system this week and test the filtering.