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

76 In operation | EarthSense TerraSentia they were sceptical that agricultural UAVs would tell them anything they didn’t know – and even if so, by the time they found something wrong with their field using unmanned aircraft or satellites, it would be too late to do anything as the damage might be irreversible at that point.” However, growers and breeders were attracted to the concept of a compact UGV augmenting crop scouts in the fields, collecting data in larger quantities for longer periods and at lower cost than labourers and UAVs. That enthusiasm effectively prompted the development of the TerraSentia. Development In its current form, the TerraSentia is 13 in wide, roughly the same height and 20.5 in long, its width being key to fitting between rows of corn and sorghum. These crops have occupied much of the initial focus of real-world tests, Chowdhary says. “Last year we had about 30 units in fields across the US, and a couple overseas, mostly with major breeding companies and universities. We wanted to get these early versions of the UGV out into the fields, and learn from that. “Nobody has actually put a sub- canopy UGV out there like this, and from these we were able to work on getting the TerraSentia to be autonomous and durable, with our early adopters paying money to test it out. “We’ll repeat the same process this year, but with an improved version of the UGV based on what we learned in 2018. And we’ll start expanding into herbicide- resistant breeding for polycultures, which is where TerraSentia’s data collection holds its biggest value. We anticipate that this will be the last year for iterating early adopter prototypes; 2020 will bring the fully productionised variant.” The vehicle weighs around 30 lb, with most of the weight taken by a 5500 mAh lithium-polymer battery. This provides a minimum three hours of endurance, but the company anticipates that rising to as much as eight hours as the system’s design is optimised for more energy storage. “With a UAV, a lot of the power goes into fighting gravity. With a UGV, we can devote most of our power to running multiple [Intel i7] computers on board, with inference chips on them for machine learning to improve analytics and behaviour over time,” Chowdhary says. The TerraSentia runs on four brushless DC (BLDC) hub motors, one on each wheel, the only moving components in the powertrain. The chassis is a thermoformed thermoplastic shell, with a metal frame body and metal struts to bear most of the stress loading. The previous design had used additively printed PLA (polylactic acid, a biodegradable and bioactive thermoplastic) for the hull, and brushed DC motors. However, the heat from these motors and some fields in Australia in the summer caused the material to overheat and melt. Switching to metal struts and thermoformed plastic however has resolved that issue. The TerraSentia also has an integrated temperature sensor and internal cooling, with a fan on the back that pulls air through from a vent on the front. Individual cooling systems are integrated on the internal computers. Three high-resolution cameras and two Lidars provide data collection and some navigation in GNSS-obscured spaces. Start-up To get the TerraSentia up and running, the user first connects with it via a tablet- based app. They then select which type of crop they want it to examine, then the kind of data they want to collect. “For phenotyping, users can select multiple variables that run concurrently with each other,” explains Chowdhary. “For example, stem count might compete with stem width because the given side camera has to look at a different location, so they can’t select both. Or a user might want to record several preset configurations of crops, data and behaviour for the vehicle to focus on.” These mission configuration profiles will also be where the user plans their next operation, depending on the degree to which they can use GNSS for automated navigation. Waypoints can be plotted, along with entry and exit points for each field. Initially this path will rely on an input of map data, such as Google Maps, UAV aerial imagery or satellite imagery. As Chowdhary explains, “When it’s possible to receive a good GNSS signal, the user can either plot a path using their geo- tagged files or remotely drive the robot along the path, and the robot will record the waypoints along the way.” Collecting data However, using GNSS signals can be troublesome owing to the multi-path effect from the canopy cover, as well February/March 2019 | Unmanned Systems Technology The TerraSentia runs off a 5500 mAh lithium- polymer battery, which powers multiple onboard computers, cameras, Lidars, and the brushless DC hub motors