Agro Food Robotics Parcours at Wageningen University

Today I had the opportunity to participate in the Wageningen Agro Food Robotics Parcours, an open day for the labs of Wageningen university working on robotics related technologies for the agro sector. It was a very interesting and inspiring experience, and so I wanted to share a selection of the work I found the most fascinating.

1. Deep learning

Deep learning is everywhere, and it has become the tool of choice for machines to make sense of inputs, whether that is to make classifications or decisions in the operation of robots and drones. The applications presented were fascinating: being able to pick out fruit (and sort by ripeness), or identify weed/crop species by vision alone.

2. Precision agriculture and farming

The idea is simple: through technologies such as GPS, sensors, robotics and advanced data analysis, farmers can cater to specific plants or animals, rather than their entire field or flock. One example was the Akkerweb platform, where satellite, drone and bespoke sampling data are combined, to let farmers know which part of their field needs what treatment for a specific crop.

Another fascinating concept was the use technology to be able to recognize and analyze the pose and movement of cows, so that a farmer can be alerted if a cow appears sluggish or otherwise sick. The image below demonstrates the technology (with people!).

Tailoring treatment to individual plants can have great environmental impact. For example, this robot can identify an undesired plant in a mono-culture, and remove it physically, or (when this is not possible due to a bulbous route), spray it with exactly 1 drop of pesticide. 

3. Robotic harvesters

These have been a while for a while, but it is always interesting to see one up close. The one below is designed to work inside a greenhouse to collect sweet bell peppers.

An interesting extension was the idea of robotic gardening assistants. Autonomous lawn mowers are as ubiquitous as Roombas, but the next generation of robotic gardeners will automatically trim a hedge to your desired shape. Meet "trimbot"!

4. 3D imaging

3D imaging is not new, but done at such a speed and precision that someone can identify the species and measure the phenotype of a plant in more-or-less production speeds, is impressive. Coupling these techniques with spectral photography, now allows researchers to measure if a plant is photosynthesizing and if it is diseased.

5. Drones

Drones are valuable for providing terrain information. It turns out that vision systems are now advanced enough to be able to detect early weeds or disease at a millimeter resolution from a height. Another interesting development is the incorporation of Lidar systems (similar to those autonomous cars use to 3D map their environment), to fully map forests and coastal areas. The lasers penetrate the canopy, and provide a 3D image of a forested area, at a resolution fine enough to calculate biomass, trunk diameter etc.

I wonder what will be on display in 10 years time!