so what's the big deal about mechanization?

Why is it so important to mechanize ‘best practices’ (see previous post)? (Caveat: ‘conventional management’ is something of a moving target. The entire industry is in slow upheaval and transforming itself incrementally, in response to diverse pressures. At some point, it may even become an ally in the pursuit of ‘best practices’, and short of that it's quite likely to adopt new technologies as they prove themselves feasible.)

Anything new must compete effectively with what already exists. In the developed world, the cultivation of commodity crops, like wheat, feed corn, and soya, is already highly mechanized. Every operation from preparing the seed bed (or planting through stubble) to harvest is performed by machine. Consequently, the prices the market will pay for such commodities is predicated on mechanical production. Any competing production regime must manage to operate profitably despite this, or no farmer will use it.

In fact, this is such a compelling argument that it simply doesn't make sense to start out by attempting to compete with conventional agriculture in the context in which it is at its most efficient, as defined in its own terms. Instead, the objective should be to bring the prices of other crops, the production of which are not so easily mechanized, more in line with the prices of commodity crops, to increase the proportion of people's diets they comprise, while managing their production according to ‘best practices’, gradually removing land from conventional management in the process. Once the technology is mature and economies of scale are in place, then maybe it will be time to include bulk commodities; a development which could be accelerated by the advent of perennial varieties of cereal crops.

But being able to compete in the market just gets you into the game; the real value in the mechanization of ‘best practices’ stems from its necessarily robotic nature, meaning that it would unavoidably be built around the acquisition, transformation, and utilization of information. Properly handled, that information could form the basis for an understanding of plant husbandry comparable to that of a master gardener, but more detailed, and vastly more reproducible (what one such machine knows others can learn almost instantly).

It is this sophistication, founded partly in programming and partly in machine learning, that can ultimately make mechanized horticulture uniquely successful, given that there will never be enough master gardeners to go around, at producing what people need while protecting biological diversity and rebuilding the fertility of the soil.

making ‘best practices’ meaningful

We've heard a fair bit about best practices lately, which is very encouraging, but I seriously doubt everyone means the same thing by that phrase.

Chances are many with a financial stake in agriculture use ‘best practices’ to mean the best that is achievable without excluding whatever it is they're selling, whether that be tractors and other heavy equipment, or herbicides and pesticides, or GMO seed, or equipment for pumping water from aquifers, and so forth.

So constrained, ‘best practices’ becomes a hollow expression, meaning something like the best we can do without changing anything we're doing. To restore meaning to the phrase, to get to the core of what ‘best practices’ should mean, we must push aside all such considerations. As internally consistent and coherent as current practice may be, it can play no part in determining what ‘best practices’ really means.

You can think of this as getting back to basics, as a mental exercise. We need food (rather a lot of it considering how many people are expected to be sharing the planet in another decade or two), and we get most of our food from the land. But the land has needs of its own, and if we ignore those we undermine its capacity to provide for our needs over the long term. How can we continue to get food from the land without diminishing its capacity to produce it? Or, better yet, how can we get food from the land while increasing its fertility and turning it into a carbon sink, at the same time?

There are other issues as well. Other species also depend on the land, for food and habitat, and it would be a much poorer world without them, but we cannot afford to take too great a hit on production to accommodate them, and we also need to improve the safety and nutritional characteristics of food in general, so our evolving question becomes how can we get non-toxic, nutritious food from the land, on a scale comparable to current practices, while, at the same time, increasing its fertility and making what biodiversity remains more resilient?

Let's provisionally define ‘best practices’ as being any combination of practices which achieve all of these goals, while also eventually eliminating the reliance of agricultural on non-renewable resources of any sort.

Notice that I haven't included the need to minimize labor in this list of goals. That has been at least a side-effect, if not actually a goal, of the transformation of agriculture brought about by petroleum and the internal combustion engine, having already resulted in a dramatic reduction in the percentage of the population directly engaged in the production of food, perhaps too large a reduction given that finding something useful for people to do is becoming a problem. Also, with the development of robotic technologies, operations formerly requiring the eyes, brains, and hands of a human being will increasingly fall within the reach of automation. So minimizing labor (whether human or some robotic surrogate) only figures in as it effects the profitability of production operations, not as a goal in itself. If labor allows you to apply methods that achieve those best-practice goals, it's bringing something valuable to the table.

It is my belief that ‘best practices’ will turn out to mean polycultures incorporating both annuals and perennials, requiring detailed manipulations, and eschewing the use of routine tillage or any use of heavy equipment that compacts the soil, as well as most use of inorganic fertilizers, pesticides, and herbicides.

This is a huge change from current common practice, but a much-needed change. Happily, that need for change is coming to a head as robotics is maturing to the point that it is reasonable to think it could enable ‘best practices’ as I've used it here, by supplying the means to mechanize them, and that a concerted effort to develop the necessary technologies could make that happen in a timely manner.

SparkFun soil moisture sensor

Granted, soil moisture sensors aren't new. What's new is being able to purchase one for $4.95, in a form that lends itself to experimentation.

a virtuous circle to advance a much-needed research agenda

In case it isn't clear, the vision, represented by this blog, for the application of (appropriate) robotics to making the best practices of horticulture scalable is, at this point, primarily a proposal for a research agenda, one that is much needed but which thus far has not attracted resources on a scale to match the task at hand.

In the absence of more enlightened distribution of what resources exist for research, I believe the most helpful turn of events would be the development of a virtuous circle that makes use of what is already available to create the beginnings of a market, which, as it demonstrates growth potential, can help drive research in the direction of this agenda.

I hope to contribute to the development of that virtuous circle, and invite all who read this to think about how they might also contribute to it.

Addendum (17July2016): If luck is the convergence of preparation and opportunity, my task in this blog is to prepare the minds of as many as possible to recognize the opportunity presented by the application of robotics to making the best practices of horticulture scalable.

FRE2015 video recap

This playlist is assembled from four sources, and includes preparation for the the event as well as many videos of the Field Robot Event itself. (Click anywhere on the text to go to the playlist on YouTube.)

Robots could make GM less appealing

With the advent of techniques that make gene splicing relatively quick, simple, and cheap (as compared with the more arduous methods of the past), it's probable that the genie is out of the bottle and no effort of collective will ever be sufficient to get it back in. But that doesn't mean we have to accept genetically modified organisms (GMOs) in our food as inevitable.

I'm going to skip the whole GMO argument, at least for now, and cut straight to the chase.

Genetic modification in the laboratory is attractive because it can produce dramatic results for far less effort than the more tedious approach of breeding plants and animals in the field and on the farm. At least with regard to plants, robots can (in principle) change that, by automating the most tedious aspects of plant breeding, such as exerting some control over which plant receives pollen from which and labelling the seeds of each plant and keeping them separate from all the rest, as well as creating a record of the growth and vigor of each plant.

Taking over such tasks and thereby reducing the amount of work involved would make folding plant breeding into crop production possible, either as an experimental plot in every field or as a layer in the handling of crops in general. This, combined with techniques for sifting through large amounts of data, would vastly increase the number of plants involved in breeding programs, making the discovery of useful genetic recombinations or mutations far more likely.

By making conventional, phenotype-driven plant breeding both easier and more effective, robots could help shift the balance as compared with gene splicing in the laboratory, making it relatively less attractive.

Drought & desertification: Robots can help

A NYTimes article published April 2nd, Mapping the Spread of Drought Across the U.S., leads off with an animated map supplied by the National Drought Mitigation Center, which shows the spread of drought conditions across the contiguous 48 states since late fall, 2014.

From that article: “Droughts appear to be intensifying over much of the West and Southwest as a result of global warming. Over the past decade, droughts in some regions have rivaled the epic dry spells of the 1930s and 1950s. About 37 percent of the contiguous United States was in at least a moderate drought as of March 31, 2015.”

There are two major ways in which robots can help with the effects of climate change, whether permanent or cyclical, upon food production.

Most immediately, robots can operate indoor production facilities using artificial light to produce high value, quickly maturing crops requiring moist environments. To operate most efficiently, that artificial light would be predominantly red and blue, since green light is mostly reflected away by plants, which is why they appear green to us. This might prove a stressful environment for human workers, but robots won't care.

The other way in which robots can help is in dry fields under the hot sun. This can be as simple as reflective umbrellas, nets, or horizontal shutters that shade the ground from the mid-day sun, but uncover it again in the late afternoon to allow cooling radiation into the night sky. Robots could also maintain drip-irrigation systems or make daily rounds to inject water into the soil near root crowns.

In principle, they could also perform planting, weeding, pest control, pruning, harvesting, and deal with plant materials left behind after harvest, and do it all working a mixture of annuals between and around standing perennials, although much of the technology needed for such a scenario remains to be developed.

On the other hand, given that level of utility, much becomes possible that currently is not. The weight of machinery can be kept entirely off of productive soil, rendering it more capable of holding water. Mulch can be applied at any time. When expected precipitation fails to materialize, plants can be pruned to reduce their leaf area and the amount of water they require. Windbreaks can be installed surrounding relatively small patches of land, in a manner not conducive to working them using tractors and conventional implements, but affording much better protection from drying winds as well as providing a secondary crop of woody fiber and habitat for wildlife. If planted in low berms, those windbreaks would also help to keep what moisture there is in the fields and eliminate water erosion.

The benefits of such technology aren't limited to coping with drought, of course, but given that drought is likely to be a widespread, persistent problem, it can help to keep marginal land, which might otherwise turn to desert, in sustainable production, and perhaps even help to reclaim some land that has already been lost to desertification, beginning with the construction of windbreak fences (like snow fences) to accumulate wind-blown dust that will become the berms into which living windbreaks can be planted.

relief and gratitude

When I began this blog, I was the only person I knew about with this vision. Over the intervening years, I have become aware of others with similar, compatible visions.

This has been a huge relief and both validating and liberating, to know that others have come to similar conclusions. My thanks to all of you!

IEEE RAS Summer School on Agricultural Robotics (SSAR 2015)

Early last December, Frank Tobe published the article Agricultural technology making news on Robohub, in which he mentioned the IEEE RAS sponsored Summer School on Agricultural Robotics, to be held in February, 2015. Well, February has come and gone, and the inaugural IEEE RAS Summer School on Agricultural Robotics (SSAR 2015) took place as scheduled.

Organizers for this event included Dr. Robert Fitch (convener), Senior Research Fellow at the Australian Centre for Field Robotics (ACFR), Professor Salah Sukkarieh, ACFR Director of Research and Innovation, Marcel Bergerman of Carnegie Mellon's Field Robotics Center (FRC), Professor EJ (Eldert) van Henten of Wangenigen UR, Professor John Billingsley of the University of South Queensland, John Reid, Director, Product Technology and Innovation at John Deere's Moline Technology Innovation Center (MTIC), and Professor Mingcong Deng of the Graduate School of Engineering, Tokyo University of Agriculture and Technology.

Invited speakers included Andrew Bate, Founding Director and CEO of SWARM FARM Robotic Agriculture, Marcel Bergerman (mentioned above), Professor Simon Blackmore, Head of Engineering at Harper-Adams University, Bruce Finney, Executive Director of Australia's Cotton Research and Development Corporation (CRDC), David Johnson of ACFR, Anthony Kachenko, Research & Development Team Leader & Portfolio Manager at Horticulture Innovation Australia, Associate Professor Kendra Kerrisk of the University of Sydney, Juan Nieto, ACFR Research Fellow, Timo Oksanen, University lecturer in the Helsinki University Department of Agricultural Sciences, Professor Tristan Perez of Queensland University of Technology, Rohan Rainbow of Crop Protection Australia, Andrew Robson, a Research Fellow with the Precision Agriculture Research Group (PARG) at the University of New England, Daniel Schmoldt, National Program Leader in the Division of Agricultural Systems, National Institute of Food and Agriculture, Professor Salah Sukkarieh (mentioned above), Professor EJ (Eldert) van Henten (mentioned above), Brett Whelan, University of Sydney Faculty of Agriculture and Environment, and Qin Zhang, Director of the Center for Precision & Automated Agriculture at Washington State University.

Of these, Robert Fitch,Salah Sukkarieh, and Kendra Kirrisk have all been interviewed by Robots Podcast, and Andrew Bate was mentioned prominently in an interview with Peter Corke of Queensland University of Technology. Kendra Kirrisk was also included in 25 women in robotics you need to know about (2014) and in Robotic cornucopia: Robohub focuses on the state-of-the-art and the future of agricultural robotics, and Simon Blackmore, Salah Sukkarieh, and EJ (Eldert) van Henten were all mentioned in a review of an article by James Mitchell Crow. Simon Blackmore has also been interviewed numerous times by BBC Radio4's Farming Today.

The Australian Broadcasting Corporation published several articles about the event on their website: one highlighting Robert Fitch and Andrew Bate, another highlighting EJ van Henten, and a third highlighting Simon Blackmore.

This is an auspicious beginning for what should become an important annual event.

Crop-neutrality

The US government currently favors production of certain crops, including corn (maize) and soy beans. A proposal, authored by Tamar Haspel and published yesterday in The Washington Post (Unearthed: A rallying cry for a crop program that could change everything), would change that by shifting subsidies from support for particular crops to crop-neutral support.

While this isn't specifically about robotics, it would have the effect of making more money available for equipment to produce crops other than the handful that have traditionally been subsidized, and, increasingly over time, that will mean robotic equipment, as the value added by sensors, processing, and flexible behavior will become too compelling to forego.

The costs and benefits of a 'moonshot' project

I've recently come to the conclusion that what I've been proposing in this blog is essentially equivalent to President Kennedy's proposal that the United States should mount a space exploration effort sufficient to send an astronaut on a round trip to the Moon within the decade, at a time when many people still believed such an enterprise to be impossible, despite that, for experts, it had clearly become a matter of doing the engineering, and only the timeline was in doubt, not the objective.

Similarly, while we don't yet have the technology to accomplish the economic robotic performance of horticultural best practices on the scale of agriculture, it's quite clear that no fundamental obstacle stands in the way of developing such technology. We have only to apply ourselves, as roboticists and as citizens supporting their efforts (private, corporate, academic, and governmental citizens), to the project.

I've already addressed many benefits we might expect such a project to produce, directly, as a result of success in the goal of creating the technology necessary to accomplish scaling up those best practices. What I haven't yet more than mentioned in passing, are the spinoffs that can be expected, even if we can't predict what most of them might be in advance.

It's widely understood that the Apollo program produced spinoffs that, taken together, amounted to a huge contribution to the economy, perhaps even more than offsetting the cost of the program itself, not least being the concentration of engineering expertise in US universities and US-based corporations.

Similarly, a determined effort to develop the necessary technologies to support, for example, polyculture incorporating perennials, can be expected to produce numerous spinoffs along the way, not the least of which would be a generation of engineers versed in the many technologies which are collectively referred to as robotics, with the confidence to apply those technologies to the tough problem of cleaning up the environmental damage humanity has done over the past few centuries.

While measured in millions of dollars, perhaps even hundreds of millions to a few billion per year, the cost of underwriting the R&D to accomplish all this would be minor compared with the cost of the Apollo program (in inflation-adjusted dollars), in part because of the far more modest scale of the resources required. Compared to the cost of recent military campaigns it would be paltry. Most importantly, in comparison with the costs of failing to do so, of leaving the future to fend for itself, it would be inconsequential. We cannot afford not to do it!

Agricultural robotics related projects funded under Horizon 2020

While many projects funded under its rules are still running, the EU's Seventh Framework Programme for Research (FP7), as a basis for new funding, has run its course, and has been replaced by Horizon 2020, The EU Framework Programme for Research and Innovation, which began in 2014 and runs through 2020.

Included under H2020's Digital Agenda, Robotics is only a small part of the overall funding framework. Nevertheless it includes “over 100 collaborative projects.”

The first H2020 call for proposals was issued last year, and on January 13th the list of robotics projects funded as a result was announced. Frank Tobe of The Robot Report has already covered this announcement and placed it in context.

Agriculture is one of four “priority domains” for robotics funding under H2020, and, of these newly funded (or refunded) projects, two – Flourish and SWEEPER – are explicitly related to agriculture.

Flourish, which has been funded at just over €3.5 Million for 42 months, will be managed by Cyrill Stachniss of the University of Bonn, Germany. It is described both in the H2020 document and on the university's website as follows: “To feed a growing world population with the given amount of available farm land, we must develop new methods of sustainable farming that increase yield while reducing reliance on herbicides and pesticides. Precision agricultural techniques seek to address this challenge by monitoring key indicators of crop health and targeting treatment only to plants that need it. This is a time consuming and expensive activity and while there has been great progress on autonomous farm robots, most systems have been developed to solve only specialized tasks. This lack of flexibility poses a high risk of no return on investment for farmers. The goal of the Flourish project is to bridge the gap between the current and desired capabilities of agricultural robots by developing an adaptable robotic solution for precision farming. By combining the aerial survey capabilities of a small autonomous multi-copter Unmanned Aerial Vehicle (UAV) with a multi-purpose agricultural Unmanned Ground Vehicle, the system will be able to survey a field from the air, perform targeted intervention on the ground, and provide detailed information for decision support, all with minimal user intervention. The system can be adapted to a wide range of crops by choosing different sensors and ground treatment packages. This development requires improvements in technological abilities for safe accurate navigation within farms, coordinated multi-robot mission planning that enables large field survey even with short UAV flight times, multispectral three-dimensional mapping with high temporal and spatial resolution, ground intervention tools and techniques, data analysis tools for crop monitoring and weed detection, and user interface design to support agricultural decision making. As these aspects are addressed in Flourish, the project will unlock new prospects for commercial agricultural robotics in the near future.”

SWEEPER, which has been funded at just over €4 Million for 36 months, will be managed by Jan Bontsema of Wageningen UR, Netherlands. It is described in the H2020 document as follows: “In modern greenhouses there is a high demand to automate labour. The availability of a skilled workforce that accepts repetitive tasks in the harsh climate conditions of a greenhouse is decreasing rapidly. The resulting increase in labour costs and reduced capacity puts major pressure on the competitiveness of the European greenhouse sector. Present robotization of this labour has entered a high level of technological readiness. However, a gap remains which halts the transition from science to economic and societal impact; the so called ‘Technological Innovation Gap’. In the EU-FP7-project CROPS, extensive research has been performed on agricultural robotics. One of the applications was a sweet pepper harvesting robot. It was shown that such a robot is economically and technically viable. The proven hardware and software modules (TRL: 6) developed in CROPS will be used as the groundwork. The successful CROPS software modules based on the Robotic-Operating-System (ROS) will be maintained and expanded in SWEEPER. Also the gripper end-effector will be retained. This patent pending module is able to grasp the sweet pepper without the need of an accurate measurement of the position and orientation of the fruit. In several experiments, it turned out that different growers use different cropping systems ranging in crop density. In SWEEPER, the cropping system itself will be optimized to facilitate robotic harvesting. In CROPS it was concluded that instead of a 9DOF, a 4DOF robot arm is sufficient, greatly reducing costs. To improve the level of robotic cognitive abilities, plant models will be applied to approximate location of sweet peppers. This “model-based vision” will increase and speed up fruit detection. Based on the insights of CROPS, sensors will be placed onto the gripper only. Also a LightField sensor will be introduced, which is able to record both colour and 3D information simultaneously.” The “CROPS” referred to above is Clever Robots for Crops, a program of the 7th Framework, which preceded H2020.