-
Notifications
You must be signed in to change notification settings - Fork 1.5k
Capstone Simulator Latency #210
Comments
Facing severe latency/lag when using the simulator for the capstone project. This happens specifically when the camera is turned on. Also, when the generated waypoints in the waypoint-updater node are set to the default 200 (camera off), the same latency appears (seems kind of OK if the waypoints are reduced to 50). |
I am having the same issue on two different machines |
Capstone project has now been moved to a workspace to help isolate any particular issues - please re-open if you see the same issues arising within the workspace. |
@mvirgo I have been using the workspace recently. BROWSER: Chrome Video at 640x480 with Graphics being Simple.
However, even with CAMERA OFF at 6:50 PM, car goes off track.
I tried ~ 2 hours later and now with same code base car stays in track with CAMERA OFF.
I have provided timelines if someone wants to analyze cluster health based on how many workspaces were active. CC @awbrown90 |
@MyCodeBits Can you post your internet connection speed as well as what browser you are using? |
Hi @mvirgo. Thanks for quick response. Earlier videos/tests were using Chrome Browser in last comment. Internet connection speedCreated another video of what the current state is at the workspace using Firefox on host (Mac OS) accessing Workspace at July 25th 3:00 PM PST
I would be happy if you want me to try any instrumented code for any reason. Further please note that there are other people in my team who are facing the same issue who are geographically at different locations and one person has commented in Slack #general also for this after i posted for the same issue. My understanding is that given the workspace is remote, if one has a decent enough internet connection, the connection speed in remote workspace only should be of consideration. Please comment if i am missing/over simplifying. CC @SIakovlev |
On the observation for Workspace getting struck: till the manual operation is present, the terminal runs at nominal pace. when the manual checkbox is unchecked, then the program runs at a slower pace. For each image topic received With automatic mode (manual checkbox unchecked, camera checked) With manual mode (manual checkbox checked, camera checked) 1532625320.6145500000 0.0969600677 The top rows are when Manual is unselected and Camera is selected. Bottom rows are when both Manual and Camera selected. So the camera processing is not creating any problem. When manual is selected, the delay between two topics received by our program is about 0.5 seconds whereas in the other case it is <0.1 seconds. Left side is the clock info, right column is the delta between previous to current clock. |
Couple items for further clarification and testing on our end - |
I did try with Safari and Chrome with not much changes. Performance was bad on both.
When I am testing my code, there is no detection code deployed. You can assume it to be walkthrough code getting tested, where car will move forward ignoring Traffic Lights. CC @SIakovlev |
I have the same problem. Since I see the issue is closed, is there a solution to the problem or not? |
I have the same problem as well. In workspace and on my laptop. |
Performance is still very bad. To the point of being unusable. Any way to get better performance out of the simulator when the camera is turned on? |
Same issue here....I am not doing any TL classification yet, and the simulator is still very slow when I turn the camera on. |
I am also facing the problem to the point that I cannot use the workspace to work on the final project. Any ideas/fixes? |
I solved it by switching to native Linux installation, I heard Docker also works pretty well, these are your 2 options, forget about the VM or the Workspace. |
@Elgeweily Thanks, I will do the same then! |
I have the same problems. Is there any other solution beside installing on native Linux? (I do not wanna install ros on my client ubuntu machine) |
I've never found one. At the end I did some optimizations and submitted on the blind |
@loewner try Docker then....this issue is not addressed properly by Udacity |
@Elgeweily Thanks, Docker worked for me as well. |
I have the same problem with the workspace, I tried the docker option locally but the car still goes off, although I don't have a GPU. My system info is: OS: Ubuntu 18.04 My question is: is a GPU required in order to be able to run the project? |
I have the same problem. |
Same here, lags with VM ROS and local simulator, but somehow at least for a part of the track usable if waypoint frequency is reduced to 10Hz. Workspace really bad, Docker only usable without Camera. I'm using Arch Linux locally, so in theory I could run native ROS, but unfortunately Opencv3 in ROS doesn't work. My machine is a Intel Core i7 Notebook with 16GiB RAM and an nvidia gtx 950m |
Any solution for the problem? |
Hopefully this article can help |
This remains an issue to this date, our whole team is stuck for now. Really disappointed that Udacity did not even care enough to answer. So what if it's 1 in a 100? Some HW combinations seems to trigger this, others which are slightly off, do not. I guess we will submit with "hacks" and hope Carla handles it fine. |
No description provided.
The text was updated successfully, but these errors were encountered: