|Increase quality and reduce cost with our robot mowers||We make small, light, low cost robots that can map your soil moisture, perform soil sampling and pest surveys, aerate, apply nutrients and chemicals only where they are needed|
|Our robots can multispectral scan your turf every day, to detect wilt and disease before it is visible to the human eye||Our robots can see in 3D|
This makes them safer and lets them mow when gps reception is lost.
|Peter James loves growing things using technology. Here's a video on the hydronpics systems he did for University of DC (Starts at minute mark 14:10)||
Robert James recently won a Genius Award from NYC MTA.|
Bob helped the city of Columbus, Ohio win the $50 million Smart City competion.
T Currently, Bob is designing autonomous transport systems for many large airports and transit systems. Bob is the "father" of connected vehical approach to automonus vehicles. Bob was a research professor at Virgina Tech's autonomous vehicles group in the 1990's.
Our robots are seeing things.
Below is a vision system to map tree positions and girth on a local tree farm.
|Machine vision allows us make a precise 3D map of your site. The map is a referemce system that provides additional positioning ie RTK GPS and compares "empty site" 3D model to realtime 3D scans to detect any nearby obstacle.|
|From aerial photos we can capture perimeters to direct the mowers where to mow and with which mower (ie. frairways, greens, approaches). Mounting cameras on mowers assists other positioning systems for a sharper cut line. This filter finds bunkers.||Aerial 3D mapping doesn't provide the resolution and accuracy needed for autonomous mowing. Envision your course as shown below with the detail of the house and tree above.|
Real Time Kinemetic (RTK) GPS provides accuracy and precision down to 1 cm. Ublox gave us a pair of C94 M8P single frequency RTK transcievers a few months back for the project. The pair with active antennas cost $399. We will include these modules with open field mowing sytems.
The base/rover pair were used for the following data collection.
The low cost Ublox seems good enough for most tasks. Cutting the collars of greens precisely would require more expnsive dual frequency RTK systems. An alternative to more expensive GPS equipment is deep learning vision. We believe we can acheive better precision by combining RTK GPS, vision, inertial and vision data. Those sweethearts at Ublox just sent us two of their, just released, F9 dual frequency RTK transcievers. Not only are they more accurate then the single frequency C94s, but these dual frequncy units recover a fix from lost sync in a few seconds. Single frequency RTK takes up to several minutes and requires known location of the rover. Each module cost $250. With antennas, enclosures, NRE board design will bring this to $1K a pair or more depending on number of units we can get be orders.
Below is some test data from Blue Mash golf course. Green data points are "fixed" data, yellow points are "float" reception which not as precisise as fixed. The rover reciever was mounted on the roof of a golf cart.
Our GPS was offset a little south and west of google maps. We've fixed this since the mapping. Each site might have different offsets.
We traveled counter clockwise in the mapping below. Notice the "fix" was lost along the tree line and never recovered. The advantage of a dual frequency RTK GPS receiver is they can recover fixed mode in a few seconds. Running in float mode is good enough to return to the shop or to a known point to re-acquire "fixed" mode satelite lock. Float mode is also probably good enough to mow roughs, survey turf health, and material applications.
Since there are many different vehicles on golf courses and many courses lease their equipment, it would be ideal to have a robot that could operate all these vehicles without having to modify any equipment.
Below is a Nasa humaniod robot Peter has programed. Peter got the robot to walk to the doorway, push the button to open the door and after a few tries made it past the finish line without failing. His fastest time was 10.72 seconds. Nasa's valker robot cost $5 million. We are working on a low cost version that could operate all your turf equipment without the need to modify your equipment.
|We are applying genetic algorithms and deep learning to train the robots . Below is an example of using genetic algorithm to get robots to walk better.|
|In fact this humanoid "toy" robot is already driving a car.||We re working on applying genetic algorithms to evolve millions/billions of machine vision and AI methids to every golf course's aerial photos. In order to automate the extraction of mowing aera permiters. This will automate the task of programing the mowers and other equipment for each golf course.|
|Here is software that detects and tracks moving objects. This is only one of many algorithms we are implimenting that will keep the public and workers safe.||Below is a different software routine that was written to find parking spaces. We are modifying it to detect out-of-place objects and things the mower is getting too close to.|
|We are developing a reel mower version. This is designed to operate in swarms. It uses less energy and provides a better cut than rotary mowers.||One of our developer wrote some software to fill any polygon area with parallel mowing pattern.|