Cultivating Intelligence in Agricultural Robots | TC Sessions: Robotics + AI 2020

Foreign [Music] Forgot your blue shirt and uh so it'll Be okay though Um so thanks for joining us you were all Doing farming automation essentially Automating various aspects of the Farming process but this is Berkeley not Davis so we may have to explain a little Bit about agriculture plants and stuff Uh so maybe we could just really quick You could tell us what the problem is That you guys are solving with your Automation Yeah I think if we kind of look at the The current state of AG so if you look The way wheat and corn and soybeans are Grown they're they're very automated They have these giant machines that are Doing every step of the process Including Harvest and so you have these Like almost robotic Farms already but 200 billion dollars worth of fruits and Vegetables are still picked by hand Every year so there's this super Expensive manual activity and the labor That does this work is disappearing People figured out that these jobs are Basically really terrible they're not Willing to do them anymore and so we Need to automate Um so we have this really enormous need And robots are the right solution for it So that's like definitely a big

Challenge for us like we press it in a Slightly broader way saying that on the One head we need to drastically increase Food production over the next 20 30 Years population is going to continue to Grow and so we need to make food Production catch up with that but on the Other hand we have those tremendous System ability challenges so we need to Be able to grow much more food in the Next two or three decades but using Drastically less natural resources and So what we the the problems were Particularly interested in are typically Around how do we grow drastically more Food with with less water less land Overall less natural resources and Robots and AI can drastically help with That and yours is weeding correct so We're building a autonomous farming Platform to help farmers perform farming Action at the plant level so our goal is To help them increase efficiency and We're starting with reading processes Right which are today very not only very Labor intensive but also very chemical Intensive I didn't mean to minimize the uh the Quality of your effort with it's just Weeding right but that's yeah it is yeah And you're getting into more uh more Domains which we'll talk about later Yeah and Michael Pica Um so we're following a very like

Specific part of AG which is uh what you Do when you have to spray a crop and the Ground is either wet or the plants are Too high Um so usually you can drive a tractor Through a crop and spray it with this Thing that has this huge booms it's uh Kind of slow but really cheap so when These other conditions are true you Actually have to fly this enormous Airplane over the crop and spray it with Chemicals from the plane so crop dusting Obviously yeah Um and we're building basically Autonomous electric crop dusters that Are Um Well they're waste way safer to operate Because there's no person in them Um they're a whole lot more precise so They solve what's really a pretty Substantial issue which is like you want The chemical to go onto the field uh you Don't want it to drift away and like go Into a school right unfortunately that Happens kind of a lot and our plants Don't do that Uh so I think it's I think it's great That you guys are all these great early Stage companies taking on farming this Is like why why farming why now why Agriculture in with with robotics like Is did you guys all have the same dream One night and you're like I'm gonna I'm

Gonna wake up and start a make a startup Right now in the year 2017 or 18. like What made you decide to start these Companies now rather than a couple years Ago or wait until the technology matures Or the need is greater So for us the Urgency came from Um their system easy concerns like we as A society need to find ways to Drastically cut on not only CO2 Emissions but also on overall impact on Ecosystems and agriculture today is a Major contributor on on this impact so All motivation came from realizing that That's one of the biggest challenges That humanity is facing uh Let's find Let's try to find ways and we actually Didn't start by wanting to build robots Like we started by really focusing on The problem and we kind of discovered Robotics as a very powerful tool to Address those issues Another aspect of this is the Development of Technology over the last Decade has really finally made this Possible in a way that it wasn't before So we have neural networks getting a lot Better gpus and CPUs getting a lot more Powerful and a lot cheaper even the cost Of robot arms and cameras is coming down A lot and so what we're seeing is it's Now possible to solve these issues with Robots and it's economically viable

Um so just be clear chapstick is Building strawberry picking robots I Forgot to mentioned that before And um what we're seeing with this is is With the cost of the components where it Is now it actually makes a lot of Economic sense for us to be building These machines so even while running Harvest as a service we'll be able to Pay back our Capital expense in six to Seven months so that means because we Can pay off our Capital expense quickly We can run this as a service in a way That's really beneficial for the grower And in a way that allows us to scale our Business really really quickly And Michael was there like a sale on Batteries and you're yeah yeah electric Aircraft uh yep really cheap batteries Um yes so for us uh obviously same deal In terms of like Um most of the underlying Tech is just Becoming at least somewhat more Commoditized Um so you're you're able to build things Without having to develop like the whole Suite which is helpful but for us in Particular AG uh it's really it's that We want to build airplanes to move People someday Um but the hurdles to do that are just Like ridiculously high Regulatory hurdles are extremely high The technical hurdles are really really

High Um and so we wanted to find an industry Where we can use the same technology but But deploy it in a way that's both like Very economically viable and also super Safe and so aggs is very unusual space For Aviation where uh you're you're Performing you know like uh really Difficult Mission with the plane but if Something goes down terribly wrong Probably no one's going to get hurt Right except the plants yeah Um so Lewis and Sebastian you guys both Have pretty robust computer vision Systems in your robots because they need To be able to find the right ripe Strawberry and located exactly or decide Which plant is which and you know get in There to cut it or whatever but then you Also both have these manual tasks of Getting in there and touching the plant In one you know nicely or not so nicely Which which is harder the the Recognition or the interaction I would say for us Our team has split 50 50 between the two So in terms of our and power goes it's About equal the grasping is super hard Especially with strawberries you have to Be able to and with a lot of other Fruits and vegetables you have to be Able to pick the berry up you have to be Able to rotate it by 90 degrees so that You bend the stem to a 90 degree angle

And then you have to apply a great deal Of force to snap the stem of the berry So the amount of force you need is Equivalent to about 20 times the weight Of the berry so you have to have a Really a really firm grasp on the berry And you have to do all this while being Very delicate because the berry has a Soft skin so we had to develop a custom Gripper basically entirely from scratch To do this which was quite challenging Um so that manipulation side is Something that we focus on really deeply Um Yeah and I'm I'm sure Sebastian will Talk about the computer vision side of It too you can talk about you can talk About the cutting too so so the hardware Society is definitely extremely Difficult Um for sure I think for us it really Depends on like what's the goal if the Goal is to exactly automate a process That currently is already happening on The field for us that's not only the Case so if that's the goal then the Hardware component is sometimes one of The hardest thing to solve but where we See the biggest value in the long term Is actually in everything else all the Stuff that we are not capable of doing Today on the field which are enabled by AI so where the computer vision Technologies really comes into play for

Us is not only in automating the weeding Process which we're doing today But also starting to be able to do much More than that and starting to be able To for instance detect diseases or Insects yeah things are today are Impossible to even think about so Short term the hardware component is Definitely a very hard piece of the Puzzle to solve but longer term most of The value will come from the algorithms Right and it's a more multifarious like The problem is more complex in a way I'm Really interested of that in that idea Of the sort of Plant Doctor rolling Along you know reporting back on that is That how far out is that is that like Five years from now you're going to be Able to say oh this plant is at 60 of This and has this parasite no we're We're not very far away from that as a Matter of fact like their older tasks That we're automating today are designed To be Um crop specific so what we call Crop Specific is every single crop leads to a Particular decision that's optimal for That crop so for weeds for instance the First process we're automating is Killing weeds like we obviously Target Only the weeds not not the actual Vegetable crops you're trying to grow But when we start thinking about Disease Control and in set control all of the

Applications will we are developing are Crop specific so in some sense it's Already the case gotcha So Michael you have a different a Different challenge with basically Building a plane from scratch which Sounds pretty hard if I'm honest but you Did it so you know let's tell us a Little about that like what was you did You really just Start writing down like plain ideas I Mean I did that when I was a kid but Yeah yeah uh yes definitely writing down Some plain ideas Um I've been into airplanes since I was Like five uh sort of obsessively Building them so in many ways it's this Lifelong trajectory that's gotten me to Building these planes but the tech Challenges Um they're very numerous so I mentioned That there's kind of commoditized Technology that's true but in the Electric Aviation Space unfortunately It's not particularly commoditized yet Because no one's made much of a business Out of electric Aviation right so There's these core components like Subsystems as a plane that we've had to Develop ourselves Um so probably 90 of the plane is custom The batteries motor controllers Motors All the like core avionics is custom all The software that runs on the plane is

Custom Um but that's also just sort of part of Our culture where very much like Tech Heavy tech company And so why maybe you can explain why you Want to go with a winged plane rather Than a multi-roader because that seems Like multi-road has seen I've seen Solutions that like this is a great Option for spraying or whatever so why Go with wings yeah so um I think broadly Speaking the multi-rotor you can think Of multi-rotors as like a platform that Has been solved at this point Um you can use it it's fairly easy to Use it can complete a variety of Missions which is super cool and that's Why they're widely deployed but from an Aviation perspective a multi-rotor is Like fairly mediocre vehicle at best and Just in terms of metrics basically to Move like a pound of stuff a mile it Takes about four times as much vehicle To do that with a multi-rotor than a Fixed Wing plane Um so what we're trying to do as a Company is basically uh be the the first Company in the world to develop like a Very very short takeoff and Landing Fixed-wing aircraft and turn that into Something that's analogous to a platform That can just solve very complicated Missions in a really robust and reliable Way

Um yeah so it's it's it's better for the Mission but a whole lot more technically Difficult Speaking of technically difficult uh We'll continue with you with the Regulation side of things which I Imagine you guys have had to deal with Regulations in one way or another but Like with flight with flight over areas Where there are people even if it's Large fields and stuff I imagine it's Just been a nightmare but I understand You are the first to have a uh what is The exact uh approval that you got yeah Yeah I'm rightly proud of so the the Approval is is basically just we have Approval to operate a large autonomous Aircraft for commercial purposes in a Wide scale way so basically anywhere in New Zealand we can operate commercially There's no other company in the world That has that approval which is very Exciting Um In terms of do you have a question about What the challenge is behind that work Yeah so I mean like it's just like Getting to do running an autonomous big Plane over large distances I would I Would say no right off the I just say Yeah so the challenges are really Um one of the biggest one is that it's Uh it's just a whole new domain for Regulators

Um so we needed to find a regulator that Was creative enough and flexible enough Um regulation yeah to try and like wrap All of our heads around what you need to Do to make a safe product in this space Yeah um so we actually did it in New Zealand there's a really really cool Group of regulators there it's a small Group Um who we've been working with for about Two years now Um and yeah it was it was a really Extensive process it involved a ton of Load testing uh flight testing sort of Subsystem level testing Um but yeah it's it's pretty cool we're Basically sort of the pioneers of that Process And for you guys I imagine there's not Only just you know regulatory issues but The question of actually what get Getting out there and executing and Trying these things out in the real World versus in the lab with your lab Strawberries and your lab weeds you Gotta get out into the the real Farms so Please tell me direct Edition aspect of It is also quite important for us Um very similarly that's the current State of wig Edition uh that covers Robotics in farming is actually pretty Immature as you can imagine so you we Have to work very closely with Regulators

To Define that regulation Um over the next few years over the next Few years so the regulation question is Is a big one And then the getting machine on the Field lucky for us like The Regulators Were talking to are quite Um open-minded and so we are able to Prove that if we work safely on on under Certain conditions we can deploy and Studying our robots on on fields today Yeah and our experience has been uh the Economics of our service even with a Human driver are still really good so we Haven't needed to do the self-driving Piece so that makes the substantial Easier also at the same time farm Equipment is really dangerous there's Like literally these like thousands of Pound like metal things with like Exposed spinning blades are just being Dragged through the field this is like Par for the course in ag so as long as Our machine is not more dangerous than That it's okay Um our experience our experience has Been that the regulation and AG is is Surprisingly minimal in our case the one Thing we do think a lot about is food Safety so we've designed our machine to Be totally food safe then there's a Question of like how do you validate That and Um especially in produce the main Force

For food safety is the fact that certain Grocery stores like especially Costco Require that Farms are certified by These third-party food safety Auditors So these third-party Auditors are like The main Force for food safety Compliance in ag so what we basically What we've had to do is to work with These Auditors to make sure that that They're happy with kind of the the level Design and the cleaning procedures that We've put into this And there's there's a fair amount of Work involved with that but Um on the whole it's been pretty Straightforward Well glad to hear that at least uh so I Want to change gears a little bit here Obviously in any in any industry Automation brings up the question of Labor and the labor that's being Replaced by the automated aspect or if It is being replaced I know you Mentioned that there was a labor Shortage and of course also that this Labor is back-breaking and not Preferable for a lot of people but still Like a berry picking robot replaces a Person who picks berries and that's sort Of a fundamental thing that you have to Wrap your mind around and if you have Ethical qualms about it I'd love to hear About those yeah so this is something That we've thought about basically from

Day one I mean I think when I told my Friends I was going to start this Company every single one of them asked About this so it's something we thought About very early on Um the way we're going to deploy our Service is basically we're going to Deploy a machine that picks a lot of the Berries and then have humans pick the Rest so really our service will augment The existing Workforce so we go and Allow Growers to grow more berries with Kind of the same or shrinking Workforce Um but what we're seeing is that there's Such a severe labor shortage in ag the Strawberry guards are literally Literally leaving parts of their field Unharvested they're literally leaving Berries to rot on the vine or on the Plant just because there's not enough People available to do the work and this Is really because it is truly horrible Work to do you're literally walking Around for 10 or 12 hours a day picking Stuff up off the ground it's very Damaging to people's bodies and really People can only do it when they're young So this is like really work that we Don't want to be doing and so it makes a Lot of sense I think a lot of the Ethical issues you might see in Some other Industries we really don't Have as much here Yeah I think another premises of

Robotics in agriculture are around Making workers more productive that's Definitely true I think there are also a Lot of A lot of value that goes beyond again Like just automating tasks that today Exist there are the examples I mentioned Earlier around this is the detection but Overall around more data-driven Processes in farming all of those things Are Um it is progress that goes beyond just Automating tasks Um so one of the biggest challenges Around labor that we're seeing is more How do we train the existing Workforce Of the the workforce that is currently Um working in on farms how do we train That Workforce to be able to manage the Types of equipment that we're bringing To Market that's that's one of the Biggest challenges we see Around labor Michael do you have anything to say to The crop dusters yeah yeah uh well crop Dusters out there we want to work with You Um but uh Yeah our plane sort of ironically Actually uses more labor uh per acre I'm Just curious about that yeah than a Normal airplane and you might be Thinking okay this this is a terrible Idea then

Um luckily it's so it's it's both a very Different kind of Labor Um and also just the economics of crop Spraying are are pretty different than These other things where the airplane Itself is the primary cost component It's not any of the labor it's the the Planes consume a huge amount of fuel Really had maintenance costs Etc so we Get rid of that Um yeah so for us there's no qualms of We're taking jobs we're actually making Some more of them it's more questions of Kind of what you're alluding to like who Are these operators going to be Um you know we need someone right now Who can like SSH into a Linux laptop Which you don't usually find on a farm Yeah that's pretty rare Increasingly though yeah I suppose my uh My boss at the beginning of the day said The farmers are among the the biggest Early adopters of Technology out there Do you find that to be true Yeah definitely I mean I like Farmers Have needed to adopt Um technology ever since like the Tractors appear that like big like about A century ago so they're very technical People Um they used to repairing machines all The time Um and even adjusting or customizing Them for for a lot of them so farmers

Are really tech savvy in that sense when It comes to new technologies computers And and like data driven processes it's Not necessarily obvious but they most of Them are very open-minded and so yeah it Makes it not that challenging the other Thing we've seen is that these guards Are growing a commodity on very narrow Profit margins and so what that means is That if they don't run their business in Like a very optimal manner they won't be Able to make a profit and they'll go out Of business so all the Growers who are Still around today and who are still Growing are Growers have been very Aggressive about minimizing costs and Maximizing revenue and really running Their business in an efficient way And so that means that if you can Demonstrate that the service you're Providing will improve their business They actually adopt very quickly Um so they're they're very very good About adopting uh technology like that And when we see a problem as severe as The labor shortage they're facing Um it's easy to show how we can make Their business better yeah and I think For people who are offering a service It's an easier sell too Um so for crop spraying for example it's Like rather than calling uh you know X Crop spraying person you may instead Call Pica

Um so it's there's not a super high Barrier to entry initially that helps Significantly gotcha well we only have About a minute left but I wanted to ask You very quickly what what is the the Technical aspect of your Creations that You are most proud of just like like Awesome Wing design or like my you know The material we made the gripper out of Or whatever like what is was your Favorite part of your giant cool Machines you built I think it's been really incredible that We've built a durable robotic platform That's been operating the field every Day for a year yeah I think that's That's a part of it maybe it's not quite Obvious that that would be super hard But actually it's really incredibly Finished pull that off Yeah I would basically I'm about to say The same thing like the copy is answered Yeah like they're bringing the the level Of precision that you that we've already Know in other Industries like Manufacturing or Healthcare bringing That to the field comes with a lot of Challenges and being able to keep that Same level of precision but also match The level of reliability that you need Is probably one of the biggest Challenges gotcha so your answer is the Whole thing your answer is the whole Thing how about you mine was gonna be

The whole thing too but I don't have any Time so All right well we're out of time but Thank you for joining us uh I look Forward to seeing all your stuff out on The farm Thank you [Music]