I’m setting up a scene with 3 Nova Carters running ROS2 (I need the Nav2 Stack to send waypoints). I’m trying now to get it as close to real-time as possible for demonstration and data generation purposes. However, the best I’ve been able to get is a Real-Time-Factor of about 0.14 (a single robot gets an RTF of 0.19).
The RTF started at about .08 until I disabled nearly every sensor’s action graph/node but the right front hawk camera, the 3D lidar, /odom, and /cmd_vel. These are about the bare minimum for running a nav2 stack and getting the robots to move around autonomously. I also am skipping every other frame for publishing the other sensors.
Some other data points:
- No robot:
0.9 RTF - Nova Carter Sensors:
0.88 RTF - Single Nova Carter ROS2:
0.19 RTF - 3 Nova Carter ROS2:
0.14 RTF
If I disable rendering any cameras or lidars, Ican get up to about 0.5 RTF. But then the simulation isn’t very valuable and I can’t run nav2.
I’m using the AWS Nvidia AMI with a g5.8xlarge computer. Here’s the output of nvidia-smi:
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.183.01 Driver Version: 535.183.01 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA A10G Off | 00000000:00:1E.0 Off | 0 |
| 0% 42C P0 99W / 300W | 6457MiB / 23028MiB | 52% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 412649 C+G /isaac-sim/kit/kit 6415MiB |
+---------------------------------------------------------------------------------------+
Although I did try upgrading it to a computer with a handful of very powerful GPUs and didn’t really move the needle much.
I’m thinking there are other settings or factors I can tweak here? Probably something buried in the settings, like a render rate or physics step I can play with?