Search

Stories From the Field: Object Detection


 


 

Welcome to ZAPT's third blog series: Stories From the Field, featuring Nomad, ZAPT's fully autonomous robotic mower. This series provides insight into Nomad's detailed testing to demonstrate navigation capacity in object detection, global & local path planning, mapping, width transformation, and remote & manual control. All the capabilities a fully autonomous lawnmower should meet. This blog demonstrates Nomads' ability to detect objects through its panoramic "eyes," which are eight high-resolution, high frame rate stereo cameras and a 360° LiDAR.


Along with other visual sensors, the stereo cameras provide Nomad with the ability to detect objects like people, animals, trees, cars, and bikes while operating. Nomad uses an industry-standard Machine Learning perception system to identify objects and remove ambiguity about unknown objects reliably. Additionally, Nomad knows the relative speed and distance to the detected object to allow it to quickly determine if this object is static like a tree or dynamic and moving like a person or animal. Nomad uses imagery and its perception suite to maintain a fixed distance to drive around the object and return to the original planned path if the object is static. However, if the object is dynamic and within a specific mower range, Nomad follows strict security protocols that comply with the client's HSE requirements and will stop to maintain safety.


The protocol is that if a moving object comes within 5 meters of Nomad when operating:

  1. Nomad stops all movement, and the mowing deck immediately shuts down

  2. All users receive an alert

  3. Nomad continues once the object is no longer detected


But if the moving object continues to come within 3 meters of Nomad when operating:

  1. Nomad stops all movement, and the entire unit, including the engine, immediately shuts down

  2. All users receive a second alert

  3. Nomad continues once granted permission from a user after an all-clear has been determined.



In this video, look through the eyes of Nomad while it performs object detection to identify a person and safely stop all movement. Then, Nomad continues once the person is no longer detected within a specific range.



In the same object detection scene as above, watch Nomad's quick reaction to stop all movement once a person is detected.


3D object detection is necessary to identify an object and know the distance to the object. Stereo cameras work by turning 2D objects into 3D objects using simple geometry and one additional camera. On Nomad, the stereo cameras have a 110° horizontal field of view and a 70° vertical field of view to detect object depth to about 20 meters, ensuring safe, highly productive mowing.


Next time, our Stories From the Field blog will demonstrate Nomad's navigation capacity in global & local path planning.

 

28 views0 comments

Recent Posts

See All