Earlier this year, WeRobotics, in association with Peru Flying Labs, completed medical cargo delivery trials in the Peruvian Amazon. They released a very thorough report on their experiences, which can be found here. Event 38 participated in the trials by customizing aircraft for longer endurance flights, higher gross weights, and dual R/C control capability, though we were not able to participate in the direct operation of the aircraft. Cargo deliveries by drone are still a very new application with little field experience available from which to draw new development upon. I applaud WeRobotics for undertaking such a challenging project and for documenting their experiences so thoroughly. Cargo delivery missions are very different from mapping missions, and so demand a different set of equipment and processes. There are many options and performance tradeoffs to consider for any group looking to make cargo deliveries over relatively long distances. Considering specifically the type of cargo deliveries made by WeRobotics, I’ll address some of the options for such a mission profile and make recommendations for what features an operator should look for when selecting an aircraft. Each of the following phases of the mission must be considered: Launch, Transit, Delivery and Recovery.

Launch

The aircraft must be powerful enough to accelerate to a safe airspeed then climb quickly to avoid obstacles, even at relatively high takeoff weight. Taking off without launching equipment limits the maximum takeoff weight to whatever the operator can carry and requires the aircraft to accelerate more quickly to safely ascend. Fortunately, oversized electric powertrains do not have the same penalty that a similar gas engine would have, so electric aircraft can easily be powerful enough to accomplish this at even moderately high gross weight, up to about 12 pounds.  At low airspeeds, the propeller may be stalled, reducing the thrust output at the most critical stage of flight, so it is important to select the propellers carefully for takeoff thrust and efficiency in cruise.

Adding an extra motor is another way to address low takeoff thrust, but flying with two motors or more reduces efficiency for the rest of the flight. Direct failure of a motor is rare, but having an extra could theoretically improve reliability as long as the electronics are able to isolate a failure of one from the other. The autopilot also need to be able to safely fly the aircraft on one motor, something that is notoriously dangerous for human pilots. Using launch equipment on the other hand requires bulky equipment and increases investment and transportation costs. For a delivery program based out of designated locations though, it can increase the maximum payload significantly.

VTOL aircraft can certainly simplify the launch and recovery procedures, but come at a significant cost in terms of equipment, maintenance, and risk. VTOL aircraft by their nature require 5 times as many motors, ESCs and propellers as other aircraft, each of which must be maintained and checked for each flight, and each of which poses an additional failure risk. VTOL capability also adds relatively significant weight to a plane and typically are fixed externally, increasing drag and decreasing lift over the nearby wing section. With the right value of goods and mission profile, VTOL can and will excel in the future, but such opportunities may not be widely available today.

Transit

Because mapping aircraft usually fly equally with and against the wind, they can recover some lost energy by traversing downwind segments in less time. A cargo aircraft must be designed to operate an entire flight completely upwind if necessary. For such missions, the aircraft must be able to optimize forward progress instead of time aloft. To do this, it must balance groundspeed against aerodynamic drag, which increases with the square of airspeed. An example of this type of feature is the Wind Resist function on the E384-LR.

Wind speeds at flight altitude also tend to be stronger than on the ground. Operators must take this into account when planning missions. I’ve found Windy.com to be a helpful tool for planning higher flights as it shows estimated wind speeds at various altitudes, but any such forecast has only limited accuracy. Wind poses a major risk to aircraft flying BLOS and out of communications range because conditions can quickly change during flight and local terrain may amplify prevailing winds to the point where they exceed the maximum speed of the aircraft. If operators are unable to maintain a telemetry link with the aircraft for the entire flight, the aircraft needs the capability to determine whether or not it can reach its primary landing zone and to automatically reroute to alternative LZs if not.

For any flight covering more than a few kilometers, it becomes necessary to use terrain following in order to safely maintain the planned altitude above ground level. For cargo drones this not only protects them from colliding with terrain, but it also helps maintain the lowest wind speed possible at lower altitudes and keeps them free from endangering other aircraft assuming proper permissions have been arranged.

Delivery

The only two realistic delivery mechanisms for logistics drones at this time are landing the aircraft or dropping the cargo from the air with a parachute. Landing completely has the benefit of reducing aircraft range requirements by half, which also significantly reduces aircraft cost. It will increase operating and infrastructure cost however, and requires personnel at both the sending and receiving end of a delivery mission. Setting up a system in which the aircraft can land at all locations also opens the possibility of 2-way deliveries, something which is mentioned as a need for isolated communities. An aircraft capable of out and back flights for the longer distances described in the WeRobotics report will likely require larger launch equipment, more space to land, and a much larger investment.

Recovery

Aircraft recovery in the WeRobotics trials was performed manually using specially outfitted aircraft to be operable by two separate R/C remote controllers. This method worked well for trials but may be difficult to scale and over time will have some added level of risk associated with human operators. It also requires a skilled operator to be present and ready for the arriving aircraft. Two other methods should be considered for aircraft recovery, auto-landing and parachute recovery. The former would likely require RTK GPS for accuracy and even then would still require a relatively large open space, something which was not conveniently located in all the Peruvian trial locations. This would increase infrastructure costs and may still pose some technical challenges with the aircraft switching between RTK base stations mid-flight. The last option, parachute recovery, is likely the best choice for cargo deliveries such as those performed by WeRobotics. It would have a payload or range penalty associated with it, but it significantly reduces the level of expertise needed by remote airfield operators. Aircraft could be landed automatically, and no operator would need to be immediately present. The parachute can also act as a safety device to improve the survivability of the aircraft if any other failure occurs during flight. Event 38 is continuing to test parachute recovery options, but some commercially available ejection mechanisms have serious drawbacks such as unreliable deployments or packing difficulty that need to be overcome.

Recommended Setup

Based on the WeRobotics cargo delivery mission profile and the desired capabilities and budget, I would recommend the following configuration for further testing. The aircraft should use a single, forward mounted motor and large propeller optimized for hand-launches and flight speeds up to 16m/s. Due to the high expected utilization, high temperatures, humidity, and difficulty of obtaining spare parts, an investment should be made into higher spec components including servos, motor controllers, and auxiliary power supplies. The aircraft mission software should be configured primarily for one-way flights, to include a ground advancement optimization, terrain following, and automatic mission abort based on estimated energy remaining. The aircraft should use a parachute for recovery, with an additional form of protection for the bottom of the aircraft in case of landing in brush.

Most of these features are commercially available although not necessarily all in the same aircraft. Event 38 has recently made some of these features standard, like Wind Resist, terrain following, and adding further safety margin to certain components like servos and power supplies. Ultimately, it is possible to achieve reliable operation of cargo drones within the requirements of a humanitarian mission operating in remote areas. A drone outfitted with the equipment described here has the best chance of success, while balancing simplicity and cost. I’m excited by the progress WeRobotics has made already. I hope to support their and other teams’ future efforts in building links to remote rivers and trails by taking to the sky.