Design Trade-offs: Endurance, Payload, and Cost

A DJI Agras T50 can lift forty kilograms of pesticide off the ground. It can also only stay airborne for seven to ten minutes before the battery needs swapping. Most people hear that and assume it is a limitation the manufacturer is still working to fix. Operators who actually run these machines treat it as a feature: they buy a stack of spare battery packs, run a nine-minute charge cycle in rotation, and keep the aircraft in the air almost continuously across a field, landing, swapping, and launching again in under a minute. Nobody is chasing a single long flight. The short flight, repeated fast, is the design, and understanding why takes a closer look at what a battery actually costs an airframe.

That workflow only makes sense once you understand why the T50 cannot simply carry a bigger battery and fly longer. Endurance and payload are not two independent settings you slide up and down. They are locked together by the one component that determines both: the battery. Understanding that link is the last piece of the constraint picture this chapter has been building, and it closes out the design thinking chapter before the series moves into what actually happens when a design leaves the drawing board.

Every additional kilogram of payload has to be lifted, which takes more power, which means either a bigger battery or a shorter flight. So far that sounds like an ordinary trade, the kind every vehicle designer has always made. The part that surprises people who are new to UAS design is what happens when you try to fix a short flight time by simply adding more battery: the battery itself is dead weight the aircraft has to lift too. Past a certain point, adding capacity stops buying you endurance, because the extra mass you just added costs nearly as much energy to carry as the extra energy it stores.

This is not a limitation of any particular battery chemistry, and it is not something a better cell design fixes on its own. Lithium polymer, or LiPo, and lithium-ion cells both run into it, typically somewhere in the range of 150 to 250 watt-hours per kilogram at the pack level. Higher energy density chemistries push the ceiling higher, but they do not remove the ceiling. It is a structural feature of flight itself: the thing that stores your energy has mass, and mass costs energy to lift. Every aircraft that has ever flown, crewed or uncrewed, has run into some version of this same relationship.

Picture endurance plotted against how much of the aircraft's total weight is battery. It is not a straight line climbing forever as you add more cells. It rises, peaks, and then falls, because every additional battery kilogram is also an additional kilogram the motors have to lift on every single flight. Research on multirotor power systems has found peak endurance typically lands around a battery weight fraction of roughly two-thirds of total takeoff weight. Push past that point chasing more flight time, and you actually start losing it.

That curve is the real shape of the endurance-payload relationship, and it explains why the T50 does not simply carry a much larger pack. At forty kilograms of payload, the aircraft is already spending most of its lift budget on cargo. Adding enough battery to meaningfully extend a ten-minute flight would push the weight fraction past the peak of that curve, buying almost nothing back in return. The better move, and the one DJI actually made, is to keep the battery small enough to stay near the productive part of the curve and solve endurance operationally instead, with a fast swap cycle rather than a bigger pack.

It is worth sitting with why that peak exists at all rather than treating it as a fixed number to memorize. Below the peak, each added battery kilogram stores more energy than it costs to lift, so endurance climbs. Above it, the aircraft is spending an increasing share of every watt-hour just carrying the pack that stores it, so endurance falls even as total stored energy keeps rising. The peak is not a hard limit imposed by regulation or hardware. It is the point where those two effects cross, and it moves depending on motor efficiency, frame weight, and aerodynamic drag, which is exactly why real platforms in different mission categories land in noticeably different places on the curve.

The shape of that curve changes depending on what the mission actually needs, and different missions land in very different places on it. A mapping platform built for hour-long survey flights carries a light sensor payload and pushes battery weight fraction high, deliberately trading payload capacity for time in the air, because the value of that mission comes from covering as much ground as possible before landing. An agricultural platform like the T50 does the opposite: it pushes payload as high as the frame and motors allow and accepts a short flight, because the mission is measured in acres covered per hour of ground operation, not minutes aloft, and a fast battery swap costs far less time than an underloaded aircraft making twice as many passes. A last-mile delivery platform sits somewhere between the two, carrying a moderate payload far enough to matter while still needing enough endurance margin to reach a diversion point if something goes wrong mid-route, which is a very different design goal from either extreme.

Weight fraction is not the only variable pulling on that curve. Speed does too, independent of how heavy the aircraft is. Flying faster, or hovering under a heavy load, pulls more current from the pack, and higher current draw makes the battery's voltage sag further under its own internal resistance. That sag matters because the flight controller triggers a landing failsafe once voltage drops below a safety threshold, well before the pack is actually empty. A high-throttle flight burns through its usable capacity faster than the raw watt-hour math would suggest, while a slow, steady cruise gets closer to the aircraft's full rated endurance. This is part of why survey platforms plan conservative cruise speeds and why headwinds cost more flight time than the extra distance alone would explain.

The redundancy floor from the last post is not a separate conversation from this curve. It sits on top of it. A delivery platform already carrying a moderate payload and needing diversion margin has less room left on the curve to also absorb the weight of dual GNSS receivers and coaxial motors, which is exactly why that mission category tends to run closer to the peak of the curve rather than pushing past it in either direction.

Cost does not track this curve smoothly either, and that is the part experienced teams learn to plan around and newer teams get caught by. Small increases in payload or endurance can usually be absorbed with a bigger battery or a slightly larger frame, and the cost rises gradually along with them. But certain thresholds force a structural step change instead of a gradual one: crossing from an airframe simple enough to build from aluminum and standard hardware into one that needs a carbon fiber structure and real stress analysis, or from a single motor per arm into a coaxial pair, or from a single flight computer into the flight-computer-level redundancy this series covered last week. None of those steps are gradual. Each one is a jump, and each jump usually costs more, in both money and engineering time, than the incremental payload or endurance gain would suggest on its own.

The step change is easiest to see in materials. An airframe carrying a light sensor payload can often get away with an off-the-shelf aluminum or plastic frame, because the loads involved stay well within what standard hardware is rated for. Push the payload high enough, the way the T50's forty-kilogram spray tank does, and the airframe crosses into territory where standard components are no longer adequate. The frame needs a purpose-built composite structure, the arms need real stress analysis instead of a rule of thumb, and the whole design cycle stretches from weeks to months. None of that shows up gradually on a budget spreadsheet. It shows up all at once, the moment the payload number crosses the line the original frame was rated for.

That is why the most expensive mistake in early UAS design is not choosing the wrong point on the endurance-payload curve. It is not realizing a step change is coming until the frame, the motors, and the flight controller are already chosen and the mission requirements shift just enough to force a jump nobody budgeted for. A payload increase of a few hundred grams can look trivial on paper right up until it is the increase that pushes the airframe across one of those thresholds, and the resulting redesign costs far more than the extra grams would ever suggest.

A delivery van runs into the same problem a UAS does. Add cargo capacity and the van needs a bigger engine and a stronger frame to carry it, both of which add weight, which eats into the fuel economy and range the extra capacity was supposed to serve. Fleet operators do not size vans by maximizing cargo space in isolation. They size them for the specific route: how far the van has to go, how much it has to carry, and how often it can return to base to reload. A cargo van meant for short urban loops with frequent depot returns gets built differently from one meant for long highway hauls between distribution centers, even though both are, at a glance, just delivery vans. A UAS mission profile is answering the exact same question, just measured in watt-hours instead of gallons, and it produces the exact same range of very different, very deliberate designs.

Endurance and payload are not independent dials. Both are set by the weight of the battery, and pushing battery weight past roughly two-thirds of total takeoff weight tends to reduce endurance rather than extend it.

• Peak endurance typically lands around a battery weight fraction of about two-thirds of total takeoff weight, not at the maximum battery size the airframe can lift
• Speed pulls on the same curve independent of weight: higher throttle draws more current, sags battery voltage faster, and trips the landing failsafe earlier than the raw watt-hour math would predict
• Mission profile determines where on that curve a platform should sit: mapping favors endurance, agriculture favors payload, delivery needs a working balance plus diversion margin
• The redundancy floor from the last post competes for the same weight budget as this curve, which is why delivery-class missions tend to sit closer to the curve's peak
• Cost does not follow the same smooth curve as endurance and payload. Certain thresholds force structural step changes instead of gradual increases, and the costliest planning mistake is not seeing one coming until the airframe is already locked in

That closes out how constraints shape the airframe on paper. The next post follows a design off the page and into the air, tracing what actually happens between a simulation model and a working aircraft on its first flight.

Curious to exchange some ideas? Reach out via the contact form or connect on Linkedin!