Robotics and Autonomous Systems (RAS) can revolutionize all agricultural sectors. Their contribution will vary in nature from crop, animal, and aquaculture to primary production activities. There are many ways robots can contribute to this spectrum, both economically (e.g. growing and harvesting more efficiently and cheaply), ecologically (e.g. reducing and eliminating chemical use while ensuring soil health) and ethically (e.g. increasing animal welfare through monitoring and early intervention). Parallel to this, robotics can enable automation in livestock and aquaculture care and alternative growing systems such as’ vertical’ farming (protected environment) or agroforestry systems combining farming and forestry on the same land.
Technology challenges in agri-robotics can be divided into 2 classes: (1) breeding/phenotyping and (2) agriculture / primary production. The challenges can be illustrated by the use of conventional terrestrial arable agriculture as a farming example.
Input traits and complementary output characteristics (e.g., conversion of a nutrient to biomass, shelf life, flavor or nutritional qualities) are traditionally identified and chosen for genetics with useful abiotic characteristics (e.g. drought or saline tolerance) or biotic (e.g. resistance to fungal viruses or bacterial diseases). In recent years, these breeding activities have seen a degree of robotic integration in order to reduce the dependence on manual intervention, but their costs and completeness, and the questionable reliability and technological readiness of existing systems have limited their use.
Although phenotypic laboratory screening may be important to identify beneficial breeding lines to be crossed, determining how this plant can thrive under real-world conditions is just a proxy. Robotics open up mass direct crop phenotyping potential in real-world farms. Such uncontrolled “non-laboratory” systems pose major challenges in identifying the specific feature leading to a beneficial phenotypic response. The robotic capacity for repetitive and detailed environmental assessment of individual plants opens the potential for a paradigm shift in developing agricultural genetics.
Establishment and Seeding
Ploughing is one of the main primary cultivation processes involving topsoil reversal or mixing to prepare the appropriate seedbed. Modern agriculture currently uses huge amounts of plugging energy: an estimated 80-90% of traditional farm energy is used to repair the damage caused by large tractors. Small, smart, electrical robots offer an alternative solution by avoiding excessive soil compaction and micro-slaughtering on-board devices. Nutrients could also be more precisely targeted at the local seed environment. To optimize plant density and seed pattern with reference to air, light, nutrients and individual plant moisture requirements, seed placement and mapping could be further automated. Robotics will also play an important role in managing primary production inputs, including soil and water specific monitoring and interventions.
Scouting timely and accurate information is a major crop management activity. This allows for cost-effective data collection by autonomous robots carrying sensors that evaluate crop health and status. Use or combine both aerial and ground-based platforms. Fusing data from different devices or sources with a wide range of temporal and spatial resolutions and interpreting data automatically poses several interesting research challenges. Weed mapping includes machine vision recording location and density (biomass) of different weeds.
A treatment map can interpret the resulting weed map. Robotic weeding is an active research field that examines alternative methods of killing, removing or retarding unwanted plants without harming crops. Intra-row weaving is harder than inter-row weaving, as it requires precise plant positioning. Alternative weed control techniques include mechanical weeding, selective (micro-) scraping, and laser weaving. Irrigation is another area where robots can help target water at the right time. Pre-harvest evaluation and robotic sensory system yield forecasts will also help select the right time for harvesting.
Selective harvesting only involves harvesting parts that meet certain quality or quantity thresholds. There are two requirements: the ability to detect the required quality factor before harvesting, and the ability to harvest the interested product without affecting the remaining crop. Selective harvesting presents challenges for modern robotic technology. Perhaps the main challenge is how to conduct independent sensorimotor coordination with noisy, incomplete sensory data in the complex agricultural environment. This would likely require a better vision for machine recognition, segmentation, spatial location, and tracking, as well as robust and accurate robotic technology. Precision was traditionally provided by rigid, easy-to-model robot mechanics.
However, the increased computing resources on robot platforms could enable the precise software and sensing burden to be handled while enabling more passive, safe and robust robotic harvesting devices. Another challenge is how much to adapt the robotic system to a particular crop and growing environment and how much more efficient selective harvesting of robotic systems should be adapted to the growing environment. Interesting tradeoffs can differ from crop to crop in this space. One related question is how to maximize year-round use of expensive robotic hardware, especially in seasonal crops like soft fruits. Possibilities include developing adaptive technologies that can switch from sharing tasks like picking fruit to common device capabilities.