Product developers working on robotic lawnmowers are confronted with a wide range of questions: Which sensors best complement the positioning solution for their specific use case? How well do their sensor fusion algorithms leverage inertial sensor and wheel tick information? And how accurate is the resulting positioning performance in the vicinity of single-story houses, apartment buildings, trees, or fences?
OpenMower gives product developers a powerful tool to answer these types of questions, assess the performance of the high precision positioning solution, and evaluate their prototypes.
Comprising both hardware and software, OpenMower includes a ready-to-use mower base with a weather-proof housing for extended testing in all kinds of inclement weather. Once the system is set up, developers can simply run the mower over extended durations to harvest valuable performance data. In the absence of a “truth” positioning solution, KML and video recordings can be used to check cross track errors.
Tried and tested in the real world
We operated the OpenMower using the RTK+INS SimpleRTK2B board from Ardusimple, which features the u-blox ZED-F9R-03B, in a backyard in Germany. The test data collected shows three passes of an eight-by-eight-meter plot with an irregular boundary and a keep-out zone for a tree in the middle of the yard.
The results, captured in KML recordings and presented here, speak for themselves. The precise overlap of each pass is most apparent on the straight stretches. Meanwhile, variability seen in turns can largely be attributed to imperfect motor control. The accurate performance is enabled by the ZED-F9R module, which provides both the attitude and positioning information at a sufficiently high update rate to smoothly control the mower.
Supporting multiple modes of operation
OpenMower is designed to support a variety of modes of operation. As seen in the table below, the system can operate in a base/rover setup or using an augmentation service like PointPerfect, with or without inertial sensor measurements and a wheel tick input. The positioning solution can be powered by a u-blox ZED-F9P RTK GNSS receiver or a u-blox ZED-F9R RTK GNSS receiver. A variant using a u-blox XPLR-HPG-2 explorer kit is also available for designs focused on cellular technology.
|System config||GNSS Hardware||Connectivity||Corrections||Supported|
|Base/rover w/IMU and Wheel Tick||ZED-F9P||Wi-Fi||RTCM/NTRIP||Yes|
|Rover w/IMU and Wheel Tick||ZED-F9P||LTE||RTCM/NTRIP||Yes|
|Rover w/IMU and Wheel Tick||ZED-F9P||LTE||PointPerfect/MQTT||Yes|
Depending on the setup used, GNSS corrections can be delivered as RTCM/NTRIP or using PointPerfect/MQTT via Wi-Fi or 4G LTE wireless communication.
A look under the hood
A look under the hood reveals the key components that power the OpenMower demonstration and testing platform. The positioning solution consists of an Ardusimple RTK GNSS receiver based on a u-blox ZED-F9P or a u-blox ZED-F9R and draws on a 4G modem to receive GNSS correction data via NTRIP or MQTT.
Its main application processor, a Raspberry Pi 4 with gigabytes of storage, and its Raspberry Pico, tasked with controlling the main sensors and charging, are clearly an overkill as a practical reference design for a mower. Instead, the OpenMower is designed to serve as an exploration vehicle for many other robotics applications. These can include small robots that take soil samples on farms, entire fleets of robots that paint the lines on football fields or corporate logos on lawns, or even robots that look for buried treasure.
To learn more about OpenMower or the person behind it, reach out to Clemens Elflein at firstname.lastname@example.org.