Harvesting technology; how can we build better agricultural robots?
How can the manufacturing industry help accelerate the development and availability of cutting-edge farming robotics?
Harvesting robots are increasingly seen as the answer to shocking levels of labour shortages and escalating consumer prices in markets such as the soft fruit sector. But automating the harvesting of delicate crops can be notoriously slow and expensive.
5% of soft fruit grown in the UK rots in our fields every year because there are not enough workers to harvest it. According to the BBC, this is costing the farming industry a staggering £36 million per annum.
Can robots harvest crops?
When it comes to harvesting soft fruit, (where the bulk of many farmer’s costs lie) machines have consistently failed to impress. It’s proved very hard for a harvesting robot to match the speed, dexterity, accuracy and judgment skills which humans have evolved over millennia to feed themselves.
Robots failed to deliver in speed and accuracy
Until recently, whenever precision agriculture robots were trialled to perform these tasks they have proved slow and wasteful compared to human labour.
In 2018, machines developed by the company CROO to pick strawberries were still only able to find and pick 50% of ripe berries in a typical field. This, compared to the 60 – 90% success rate recorded by teams of farm workers.
But harvesting technology has improved
But the huge advances in agricultural robotics over recent years have seen harvesting technology improve exponentially. The multi-disciplinary skill required to build ‘intelligent’ harvesting machines with complex moving parts has finally come together, including:
- Sensor tech (to assess sight, scent, and moisture levels of fruiting plants)
- AI (for machinery to learn and optimise picking skills in changing conditions)
- Precision of motion (mechanical engineering to mimic human dexterity)
- Software and mech-eng integration (to improve speed and meet quality deadlines)
And, at last, thanks to tireless R&D and huge levels of investment in precision agriculture tech, the robots are finally being commercialised.
How agribot R&D is bearing fruit around the world
In California ‘The Advanced Farm’ company now operates 10 robotic strawberry picking machines in addition to various apple picking and propagation technologies.
The wheeled berry pickers, with robotic arms, patrol the rows of plants in covered tunnels to harvest fruit at the optimum time:
Each time a ripe strawberry is identified, a silicone robotic hand with a suction cup in the middle moves in, grabs the strawberry, and then uses three fingers to twist it away from the stem and place it in a bin. Advanced Farm designed nearly 50 versions before deciding on the current design of its picking system.
Each of these machines can now pick around 50 kilos of strawberries an hour, joining battalions of soft fruit harvesting machinery which are operational across the West Coast of America.
Harvesting robots: speed is still the enemy
But soft fruit harvesting technology is being constantly iterated and improved across the world. Different companies have been racing to improve waste levels and, critically, increase their speed to compete with human labour.
3 years ago in the UK, the Plymouth University spin-out company Fieldwork debuted their raspberry picking robots to great fanfare:
Guided by sensors and 3D cameras, its gripper zooms in on ripe fruit using machine learning, a form of artificial intelligence. When operating at full tilt, its developers say the robot’s gripper picks a raspberry in 10 seconds or less and drops it in a tray where the fruit gets sorted by maturity, before being moved into punnets, ready to be transported to supermarkets.
In spite of the extraordinary mechanical engineering and mechatronic skill required to robotically pick such delicate fruit, it was just too slow. 10 seconds per berry was no match for the speed and dexterity of a human picker. Added to that, the machines were indiscriminate when it came to the ripeness of the fruit they were harvesting.
But the team at Fieldwork have now completely redesigned their offering, from the sensor tech which monitors ripeness, to the grippers which could cause damage to the incredibly delicate fruit. The result has been a faster product:
Robots are now picking 1kg of fruit an hour, with the company working to ramp this up to more than 4kg an hour. The firm is aiming to have a robot picking 25,000 raspberries a day, compared with 15,000 for a human working an eight hour shift.
Still, raspberries aren’t the most complex harvesting process that has been automated in the agricultural sector.
White asparagus: the final frontier?
Harvesting premium, white asparagus is especially complex because it needs to be picked at a particular moment when it is ripe but still under the ground. In Holland, the Cerescon company set about automating this process but it’s been a huge challenge.
“Selective harvesting is really complex. You need hi-tech sensors, you need electronics, you need robotics,” says its inventor Ad Vermeer.
To “see” the asparagus, the robot injects an electrical signal into the ground. Electrochemical sensors dig through the soil and pick up the signal the closer they get to the asparagus. Then the machine harvests selectively based on these results.
For Ad Vermeer, it took 14 years for the vision, available technology and investment in the sector to coincide to realise this vision. Now, the machines, which have been commercially available for 5 years, have replaced the need for 100s of workers, improving affordability of the end product and worries about labour shortages.
Investment and opportunities accelerate
We are still in the relatively early days of innovation in precision robotics in agriculture. Yet developments in sensor tech, miniaturisation and advanced mechanical engineering are seeing development and investment accelerate fast. There is strong belief in the potential of this tech to help develop more sustainable farming practices, beat labour shortages and increase efficiency and crop yield on the ground. Large amounts of money are being ploughed into the sector as a result to remove the necessity of human labour to undertake routine agricultural tasks.
Advanced Farming closed a $25 funding round in 2021, while the fruit farming group Bowery have raised an incredible $300 million for their automated harvesting solutions in the last year. The startup Oishii raised $50 million in 2021 for its vertical farming operation as it works to grow and harvest difficult-to-cultivate Omakase mountain strawberries in its indoor facilities.
Meanwhile, the UK government have announced a £25 million fund to help farmers access the tech that could transform their fortunes in the labour crisis resulting from Brexit.
Cost and complexity plagues commercialisation
But still, cost problems remain. These farming robots remain expensive and the complexity of maintenance means they are outside the reach of most farmers.
The technology itself is in a state of flux with new solutions being iterated all the time. For example, the use of cutting edge, re-generating materials that will organically ‘grow’ new coverings for ‘gripper mechanisms’ when they wear out.
The truth is, farmers won’t want to buy first hand when the technology could be obsolete in a matter of months.
Agtech OEMs need to keep up their development momentum
Right now, agtech OEMs need to keep up the development momentum, finding ways to improve performance, reduce costs and bring their solutions to a wider market.
Servitisation and leasing models are helping robotics developers monetise their inventions right now. And that in turn is bringing valuable data and real world experience to bear on their R&D programmes.
But as the tech cycle and demand continues to accelerate there are opportunities for agtech OEMs to partner with EMS providers to value engineer machinery and iterate their offerings more rapidly.
Sensor tech is hard to get right
Improving the resilience of sensor technology integrated into heavy machinery is proving particularly challenging for some OEMs. But working with third party experts in active lens alignment can bring huge gains in accuracy, where micro-millimetre precision in sensors make all the difference to crop yield.
This kind of support will help innovators ensure design, procurement and manufacturing processes are as efficient and cost-effective as possible.
EMS with mechatronic expertise can bring new players to market
The need for continued innovation in smart farming may also be attracting players from other sectors with new software and AI applications. And who can blame them? From autonomous tractors, to crop health management solutions, milking robots, spraying drones and weeding robots, there is opportunity everywhere.
According to Precedence Research the global agtech market size is expected to grow from $24.08 billion in 2024 to $43.37 billion by 2030.
But challenger companies will need to work with third-party component and mechatronic experts to fill specific gaps in their knowledge and help them apply their solutions to a new and unknown sector.
EMS with specific skills in these areas can help new entrants to the marketplace realise their ideas while avoiding expensive mistakes.
At the same time, they can help those with a foothold in the market already to iterate the next generation of their products faster and in more commercial ways.
From big players in the agricultural industry like John Deere, to the smallest farming drone technology start-up, there’s a common need to access third party micro-electronic and mechanical engineering expertise to scale the manufacture of cutting-edge new products.
As the need for speed in innovation increases, the next generation of agtech may only be realised with the support of EMS providers who have the right blend of design, sensor tech and mechatronic expertise.
Editor’s note: this post was originally published in June 2022, and republished in May 2024 for accuracy.
