In autonomous ground systems, developing a clear model of the surroundings is crucial for operating in any environment. Three-dimensional light detection and ranging (LIDAR) sensors, such as the Velodyne HDL-64E S2, are powerful tools for robotic perception. However, these sensors generate large data sets exceeding one million points per second that can be difficult to use on space, power, and processing constrained platforms. We report on GPU assisted processing within a Robotic Operating System (ROS) environment capable of achieving greater than an order of magnitude reduction in point cloud ground segmentation processing time using a gradient field algorithm with only a small increase in power consumption.