Revision history [back]

That hardware is more than enough. You don't need a dedicated graphics chip. RViz does have some problems on Intel integrated graphics chips, but those are just rendering problems of meshes (such as the robot model) and can be circumvented by using LIBGL_ALWAYS_SOFTWARE=1. If you don't want to go through that hassle, get an NVidia dedicated graphics card (but as I said, you don't really need to).

Other points to watch out for:

• The Kinect doesn't work on USB 3.0 ports, so make sure your laptop has at least one USB 2.0 port.
• You might not be able to write the point clouds to disk at full frame rate if your hard disk is too slow. One Kinect point cloud is around 10 MB, at 30 Hz that's 300 MB/s. Rosbag will buffer in memory, but once that's filled up, you'll lose some data. The best way to circumvent this is to record the depth and rgb images instead (around 2 MB/frame, or 60 MB/s) and use a script such as this or this to compute the point clouds for them offline. Or just throttle the frame rate. Anyway, I thought I'd mention it because an SSD might help there.
• The Kinect needs an external power supply (wall socket or a big battery and some soldering). If that's a problem, you could get an Asus Xtion Pro Live instead. Basically the same hardware, but USB powered. (And IIRC, it also works on USB 3.0).

That hardware is more than enough. You don't need a dedicated graphics chip. RViz does have some problems on Intel integrated graphics chips, but those are just rendering problems of meshes (such as the robot model) and can be circumvented by using LIBGL_ALWAYS_SOFTWARE=1. If you don't want to go through that hassle, get an NVidia dedicated graphics card (but as I said, you don't really need to).

Other points to watch out for:

• The Kinect doesn't work on USB 3.0 ports, so make sure your laptop has at least one USB 2.0 port.
• You might not be able to write the point clouds to disk at full frame rate if your hard disk is too slow. One Kinect point cloud is around 10 MB, at 30 Hz that's 300 MB/s. Rosbag will buffer in memory, but once that's filled up, you'll lose some data. The best way to circumvent this is to record the depth and rgb images instead (around 2 MB/frame, or 60 MB/s) and use a script such as this or this to compute the point clouds for them offline. Or just throttle the frame rate. Anyway, I thought I'd mention it because an SSD might help there.
• The Kinect needs an external power supply (wall socket or a big battery and some soldering). If that's a problem, you could get an Asus Xtion Pro Live instead. Basically the same hardware, but USB powered. (And IIRC, it also works on USB 3.0).powered.

That hardware is more than enough. You don't need a dedicated graphics chip. RViz does have some problems on Intel integrated graphics chips, but those are just rendering problems of meshes (such as the robot model) and can be circumvented by using LIBGL_ALWAYS_SOFTWARE=1. If you don't want to go through that hassle, get an NVidia dedicated graphics card (but card, but as I said, you don't really need to).to. (If you wanted to run a complex Gazebo simulation, or run KinFu, that's a different story; you should definitely get a good NVidia card for that).

Other points to watch out for:

• The Kinect doesn't work on USB 3.0 ports, so make sure your laptop has at least one USB 2.0 port.
• You might not be able to write the point clouds to disk at full frame rate if your hard disk is too slow. One Kinect point cloud is around 10 MB, at 30 Hz that's 300 MB/s. Rosbag will buffer in memory, but once that's filled up, you'll lose some data. The best way to circumvent this is to record the depth and rgb images instead (around 2 MB/frame, or 60 MB/s) and use a script such as this or this to compute the point clouds for them offline. Or just throttle the frame rate. Anyway, I thought I'd mention it because an SSD might help there.
• The Kinect needs an external power supply (wall socket or a big battery and some soldering). If that's a problem, you could get an Asus Xtion Pro Live instead. Basically the same hardware, but USB powered.

Regarding your question what kind of laptops I've successfully used to record Kinect bag files:

• 3.5 year old Dell Latitude E4300: Core 2 Duo, 4 GB RAM, dedicated NVidia (256 MB)
• 2.5 year old MacBook Pro (also running Ubuntu, of course): Core 2 Duo, integrated Intel chip set
• 0.5 year old Dell Latitude E6430: i7, 8 GB RAM, dedicated NVidia (1GB)

That hardware is more than enough. You don't need a dedicated graphics chip. RViz does have some problems on Intel integrated graphics chips, but those are just rendering problems of meshes (such as the robot model) and can be circumvented by using LIBGL_ALWAYS_SOFTWARE=1. If you don't want to go through that hassle, get an NVidia dedicated graphics card, but as I said, you don't really need to. (If you wanted to run a complex Gazebo simulation, or run KinFu, that's a different story; you should definitely get a good NVidia card for that).

Other points to watch out for:

• The Kinect doesn't work on USB 3.0 ports, so make sure your laptop has at least one USB 2.0 port.
• You might not be able to write the point clouds to disk at full frame rate if your hard disk is too slow. One Kinect point cloud is around 10 MB, at 30 Hz that's 300 MB/s. Rosbag will buffer in memory, but once that's filled up, you'll lose some data. The best way to circumvent this is to record the depth and rgb images instead (around 2 MB/frame, or 60 MB/s) and use a script such as this or this to compute the point clouds for them offline. Or just throttle the frame rate. Anyway, I thought I'd mention it because an SSD might help there.
• The Kinect needs an external power supply (wall socket or a big battery and some soldering). If that's a problem, you could get an Asus Xtion Pro Live instead. Basically the same hardware, but USB powered.

Regarding your question what kind of laptops I've successfully used to record Kinect bag files:files (works smooth on all of them):

• 3.5 year old Dell Latitude E4300: Core 2 Duo, 4 GB RAM, dedicated NVidia (256 MB)
• 2.5 year old MacBook Pro (also running Ubuntu, of course): Core 2 Duo, integrated Intel chip set
• 0.5 year old Dell Latitude E6430: i7, 8 GB RAM, dedicated NVidia (1GB)