Unmanned Systems Technology 024 | Wingcopter 178 l 5G focus l UUVs insight l CES report l Stromkind KAT l Intelligent Energy fuel cell l Earthsense TerraSentia l Connectors focus l Advanced Engineering report

75 talking about how we could enable polycultures,” EarthSense’s CTO Girish Chowdhary says. “The key bottleneck to developing polycultures is that current automation technology couldn’t work in such a complicated stretch of arable farmland. Meanwhile, the main bottleneck in a profitable farm is sufficient labour, or sufficient automation to enhance the capabilities of that labour.” These two issues spurred the team to consider smaller robotic vehicles that could navigate fields more nimbly than modern tractors and other equipment. By way of background, to boost the development of polycultures, plant breeding companies need greater research and understanding of plant phenotyping – the measurement of plant traits in order to relate them back to their genetic make-up. By understanding the correlation between physical traits and genes at a large scale across entire species – known as ‘high-throughput phenotyping’ – plants can be bred to give far higher yields. Current agricultural scientific literature highlights the lack of high- throughput phenotyping as a primary bottleneck in agricultural progress. “So, in addition to wanting to achieve closer crop monitoring with small robotics, we had this phenotyping project – a part of ARPA-E, the Advanced Research Projects Agency-Energy in the US – which drove us to form a concept for small UGVs that could travel under the crop canopy and get phenotypic data,” Chowdhary explains. Phenotypic data includes variables such as plant height, width and leaf area, and is typically challenging to collect, as it is hidden beneath the canopies of crops and difficult to read from above. Crop scouts can be hired to walk among the crops, but these human workers cannot realistically be expected to crawl for hours through mud, sharp leaves and chemical residue, photographing plant parts. Chowdhary and his team had originally hypothesised that UAVs were the ideal tool for collecting data for parties interested in a solution to the labour and phenotyping shortfalls. However, as he explains, “We spent about four months on a programme called the National Science Foundation’s I-Corps, which supports fundamental research to get university professors to talk to real people, with business hypotheses based on their research and technology, to ask if they’d buy a UAV- centred solution. “We talked to more than 200 growers about UAVs that could carry NDVI and hyperspectral cameras, and fly them over fields, and basically we heard that none of that was really working for them.” Early indicators of problems such as nitrogen stress are more often found under the canopy, not at the top. The first leaves that show signs of nitrogen stress are at the bottom of plants, since the tops get the most sunlight, making a key case for a ground-based approach over an aerial one. “Also, NDVI data was really designed to distinguish between green and not-green; RGB images probably actually had more actionable information than the NDVI pictures,” Chowdhary notes. “So EarthSense TerraSentia | In operation Unmanned Systems Technology | February/March 2019 The UGV has also been designed to help farmers monitor for indicators of poor plant health and weeds closer to the crops and ground, where such problems become visible much sooner than they do from above