Small farm robots – How to get them moving

In a field just outside Salisbury, England, Tom moves carefully but briskly over rows of wheat plants, examining each set of leaves. Tiny differences between the narrow green stems might indicate that the hated blackgrass has infested the crops, reducing yield and slashing financial returns from this field. If blackgrass is found, this patch might need a dose of herbicide, or a vigorous hoeing.

Such a level of crop care is difficult in a commodity crop like wheat, grown on a vast scale in fields that can stretch to the horizon. That such attentiveness can be shown to the little plants in every part of this field is due to Tom’s inexhaustible patience. That owes much to the fact that Tom is a small farm robot, created by the Small Robot Company (SRC).

Indeed, a number of companies are already developing clever innovations that allow other small farm robots to perform hard farming practices with high precision and low impact. Activities such as planting, watering, spraying, hoeing, weeding, making furrows, monitoring and data collection, and even picking soft fruits, are all falling to automation.

These functional innovations form each of the individual USPs of the robots, but regardless of the function, small farm robots also need to move around the farm to do their jobs. Every one of them will need to solve this autonomy challenge. TTP and SRC have started a collaboration to achieve this.

Tom counts as a small farm robot even though he is nearly the size of a Mini, at ~1.5m x 1.3m. He is “small” by comparison with the enormous machines typical for sowing, tending and harvesting broadacre crops.

Deliberately created to be lightweight and with large, low-pressure tyres, he can roll over emerging plants without damaging them; with a pressure on the ground that is lower than a human wearing snowshoes, so as to avoid compacting or damaging the soil.

Small farm robots are billed as the answer to many a farmer’s desire to produce more food with fewer chemicals, and less waste, with long term sustainability.

This presents a commercial niche for “sustainable intensification” of food production, which leads to a requirement for better tending of crops. Plant-by-plant crop-care practices that would be more at home in a cottage garden are beginning to be thought of at scale in commercial horticulture – and even in broadacre crops.

SRC are proof that this what the market is demanding, their investors are mostly farmers.

But small farm robots will not deliver for farmers if their developers do not deliver on autonomy, enabling robots to make their way around the farm while recognising and responding to edge cases – the limit of field, a bounding dog, a pylon, hikers, uneven ground, holes – hail or shine.

In a bid to support the industry in the development of safe autonomous robots, a new standard has recently been launched; ISO 18496. This standard specifies the principles for design, verification, and validation of highly automated machines during field operations.

It includes a requirement to design perception systems that mitigate the risks of mis-detection of, for example, obstacles obscured by crops, dust, fog, snow or rain; light, dust or vibrations impacting perception systems; and uneven ground causing the sensing beam to vary as the vehicle pitches and tilts. In understanding what we’re up against, this is a good place to start.

Sensor fusion will be part of the answer, combining the input of several sensors (such as cameras, lidar, and radar) to see in different light conditions, overcome dusty and muddy environments, and detect hazards behind foliage.

There are also lessons to be learned from farm robotics innovation to date, as well as from other industries with more experience grappling with autonomy.

SRC are sharing their time, robots and challenges, so we can work collaboratively to address them. We hope this project will grow through more partners, who can also benefit from our autonomous vehicle platform.

We are sharing TTP’s know-how in solving complex challenges in autonomous vehicle perception systems. One example involved optimising lidar performances for bad weather by developing algorithms to correct for laser light absorption, scatter, and reflection in rain and fog. Another looked at innovative cleaning methods to eliminate signal interference from dirt on the sensor.

For the sake of sustainable food production, this is a new industry that needs to learn fast. We too are keen to learn more. We would love to hear from agricultural robot companies who are willing to share their challenges so we can solve them together.

Contact us to find out more.