Welcome to another entry in ZAPT's third blog series: Stories From the Field. This week we will help you understand how Nomad defines where to mow and where not to mow using global and local path planning.
But first, what are global and local path planning?
Essentially, the global path is a pre-determined sequence of coordinates that determine Nomad's most optimal mowing pattern in a field. Our software automatically generates this global path within bounding data from processed satellite imagery that defines the job site's outer and inner bounding polygons. These polygons define the mowing areas for Nomad. The primary area to be mowed is an "outer bounding polygon" (OBP), and obstructions within the OBP, such as flower beds or ponds, are defined as "inner obstruction polygons" (IOP), which are areas to avoid. Suppose no satellite data (or low-resolution data) is available for the global path. In that case, we can always collect this data during our first mow on-site using Nomad under radio control.
Image: Satellite data imagery for global path planning
The local path planner is what runs in real-time on the mower. It takes the path defined by the global path planner and the real-time sensor data from Nomad, and this local path planner drives the mower. In a perfect world, the local path planner would use our localization data to follow the global path and cut grass. But because Nomad has to drive around unknown static obstructions (utility poles, pipeline markets, bench seats), the local planner brings in our perception sensors and manages all of these changes from the original global path. Nomad uses both stereo cameras and LiDAR for perception. The stereo cameras are Nomad's panoramic "eyes" that are eight high-resolution, high-frame-rate cameras. LiDAR is a high-resolution survey-grade mapping system that uses light as a pulsed laser to measure ranges to what Nomad can see to build an accurately scaled map around Nomad.
Image: An accurately scaled map from LiDAR
With global and local path planning together, Nomad precisely knows where it is on the earth's surface and the pre-determined mowing coordinates for a job site. Then, as Nomad mows, the imagery available from LiDAR and stereo cameras provides constant feedback for autonomous positioning with real-time obstacle avoidance.
Above is an example of an artificial "potential field" showing obstacle avoidance with Nomad to simulate how Nomad thinks while moving throughout its defined mowing path. Think of the repulsive forces shown above as trees. An artificial potential field is a technique used in path planning to let Nomad navigate to the end of a mow line while avoiding obstacles shown as repulsive forces. In a cluttered space like above, the sum of the attractive and repulsive forces dictates the steering direction of Nomad.
This video demonstrates a time series of the mow line attractor as Nomad moves through the artificial potential field.
Next time, our Stories From the Field blog will dive deeper into local path planning to demonstrate Nomad's trueness to the path by discussing the path planned, the actual path driven, and Nomads' accuracy while driving that path.