Stories From the Field: Path Planning II



Welcome to the third entry in ZAPT's third blog series: Stories From the Field. Last time, we helped you understand how Nomad defines where to mow and where not to mow using global and local path planning to position itself. This week, we will dive deeper into Nomad's accuracy and ability to stay precisely on its path by looking at the difference between the planned and actual path driven.

As a refresher, the global path is a pre-determined sequence of coordinates that determine Nomad's most optimal mowing pattern in a field. This global path is determined once we have mapped the mowing site. During operations, the local path planner is the online software that keeps the mower on this path while looking at all the real-time data from sensors on the mower. These sensors include perception sensors such as LiDAR and stereo cameras and localization sensors such as GNSS and IMU. This local path planner is essential for Nomad to stay on the planned global path, avoid collisions and unknown objects and maintain safe operations.

Nomad drives in straight lines from waypoints the global path planner gives to drive the most efficient path possible every time. While following the planned path, Nomad will mow within a maximum offset of +/-5 cm from the defined mowing path with a maximum offset of less than 15% of the linear distance traversed. This assumes we have RTK GNSS. If we are under canopy and have LiDAR as our primary localization sensor this will increase to +/-15cm.

To test Nomad's capability to transit the planned path while maintaining swath control, we set up a demonstration that consists of more than five transit lines, at least 20 m per line, parallel to each other in a field with tree coverage.

Watch in the video below as Nomad "crab crawls" by independently turning all wheels 90° to efficiently move to the following parallel line without leaving the planned path. The alternative to crab crawling is Ackerman steering, where the wheels turn in a circle from one center point. However, this type of steering results in a large turning radius to get back to the path, so we have chosen to implement crab crawling on Nomad to maximize productivity.

Video: Nomad driving its planned path

In another view, watch the LiDAR sensor on Nomad scan with light as a pulsed laser to measure the distance of objects around Nomad to build a map of its surroundings. LiDAR's SLAM (simultaneous localization and mapping) algorithm then computes the map with colors indicating an object's intensity and distance from Nomad, with red as the closest and purple as the furthest. In areas with tree cover, LiDAR is necessary for accurate autonomous positioning to stay on the planned path as trees degrade or totally block GNSS signals that would otherwise transmit positioning signals from satellites in space. Additionally, we integrate LiDAR with GNSS and an Inertial navigation system (INS) to aid Nomads’ mowing accuracy and processing speed for quick decision making.

Video: LiDAR view of Nomad driving its planned path

In this video, each point represents the actual path driven by Nomad as it moves through the planned path. During this field demonstration, Nomad stayed true to the planned path with complete coverage of the mowing area, constantly building a map of its surrounds and verifying its position by matching each point driven against the base map of the site.

Next time, our Stories From the Field blog will discuss how Nomad uses LiDAR for mapping to build a base map to autonomously position itself to operate even if GNSS completely fails. Nomads’ integrated sensor suite is essential for accurate and high productivity under trees or alongside tall buildings – which often are a large part of any job site.


34 views0 comments

Recent Posts

See All