A crewed Mars rover is a power system problem that happens to move

What the work claims

This is a design study, not a flight result, and it says so itself. For a notional 2035 mission in which two crew drive a small pressurized rover with an unpressurized rover bolted behind it for 21 Earth days, up to 10 km from the ascent vehicle, a dynamic radioisotope power system (RPS) integrated into the trailing vehicle is the best power option on mass and packaging grounds.¹ The quantitative core: the RPS case electrical power system totals 970.3 kg including mass growth, while a 21-day PEM fuel-cell system for the identical mission totals 2,691.7 kg, which the authors summarize as over 2.5 times larger.¹ Solar was assessed but not carried as a full case because a single-array solution needed a 9.6 m diameter wing the team judged a hazard to the crew in high winds.¹

Every number in this article comes from the memorandum's own sections and tables; nothing here is flight data, and nothing here has been demonstrated as an integrated system.

How it works

The load profile, not the power source, drives the architecture. The rover needs 1.8 kWe of hotel load, 250 We of science, and 10.0 kWe for mobility, with a 30 percent margin on most loads but only 5 percent on drive power, giving a peak driving demand of 13.4 kW for one hour per sol and a parked demand near 3.0 kW (Section 4.1, Table 4.1).¹ The one-hour mobility burst is the entire reason this vehicle cannot simply carry a generator sized to its average load.

The baseline (Case 1) couples 48 general-purpose heat source (GPHS) bricks, roughly 250 Wth each at beginning of life, to three pairs of dual-opposed Stirling convertors through pumped sodium-potassium (NaK) loops, producing 3.39 kWe maximum (Section 4.1.3.1).¹ The RPS system mass is 352.92 kg at a stated system-level specific power of 9.62 W/kg, rejecting waste heat through 16.22 square meters of radiator; the study explicitly assigns the whole RPS TRL 3 to 4.¹ Stirling hot-end heat arrives via two series annular linear induction pumps assumed at 15 percent efficiency, with 5 percent heat loss through insulation and 1.5 percent through piping and structure (Section 4.1.3.1.1).¹ The convertor reference point is a Sunpower 1 kWe machine whose only modification for the NaK loop test at Marshall was an inlet-outlet manifold; the convertors performed identically to their gas-burner heater heads (Section 4.1.3.1.1). That is a component-level test result, not a system qualification.

Because 3.39 kWe cannot cover the 13.4 kW driving peak, a lithium-ion battery built from COTS LG 18650 MJ1 cells provides 10 kW during the daily drive hour: 34 cells per string for 120 VDC, 132 strings, 52.7 kWh, 448.8 Ah, 285.9 kg, depth of discharge limited to 80 percent so the pack ends day 21 at 20 percent state of charge (Section 4.1.3.1.2).¹ Those stated figures imply about 184 Wh/kg at the pack level, our arithmetic from the source's own mass and energy. The RPS trickle-charges the battery to 96.2 percent each day, accepting a 3.8 percent daily round-trip loss.¹

The two fuel-cell cases are really a storage-mass story. Case 2 runs six days on supercritical hydrogen and oxygen, swaps product water for fresh reactants at the lander, and needs 527.1 kWh of storage; Case 3 carries all reactants for 21 days and needs 2,059.0 kWh, at which point the reactant and tank mass dominates: 2,336.7 kg for the fuel-cell system alone (Sections 4.1.3.2, 4.1.3.3, Tables 4.3 and 4.4).¹ The rejected solar option is a 73 square meter UltraFlex wing of 140 kg using 33 percent efficient IMM-alpha cells, TRL 8 by Phoenix and InSight heritage, feeding a 72.7 kWh, 375 kg battery (Section 4.1.3.4).¹

Steelman: the best case for the RPS baseline

The mass argument is clean and it compounds. Radioisotope power carries its energy in the isotope itself, so a 21-day mission costs the same power-system mass as a 2-day mission; the fuel-cell cases scale with days because their reactant load is the energy storage. That is why the gap widens from a 1,868 kg fuel-cell EPS at six days to 2,692 kg at 21 days while the RPS case sits unchanged at 970 kg.¹ For a vehicle whose whole point is to leave the lander for weeks, energy density at the system level beats conversion efficiency at the cell level.

The packaging argument is equally concrete. The study rejected solar partly because a single 9.6 m wing was judged a high-wind risk to a pressurized crew vehicle, and rejected fuel-cell tanks partly because mounted on the access platform they would block crew egress to the surface; the RPS tucks three GPHS stacks into the aft deck behind a 10 cm water shield, keeps the crew path clear, and leaves the Stirling convertors as the only external intrusion (Section 3.2).¹ The radiation analysis concludes that with shielding, crew exposure from the RPS is held near the Mars surface background (Section 2.2).¹ Finally, the battery is deliberately boring: commodity 18650 cells in a custom pack, TRL 6, against a Stirling hot end whose fundamental convertor technology is shared with the ASRG program.¹

Where a skeptic should push

The single most load-bearing assumption is that a dynamic RPS at TRL 3 to 4 matures on schedule. The study is admirably direct about this, printing the TRL next to the mass figure, but a 48-GPHS, NaK-pumped, three-module Stirling system has never flown in any configuration, and the 970 kg figure carries a 40.8 percent combined mass growth and margin against the AIAA-recommended 30 percent for a green estimate (Section 3.1.2).¹ The convertor hardware pedigree cited is a micro-cogeneration machine with a manifold change, tested in a pumped loop facility: real, but a long way from a vibration-loaded, dust-choked Mars deck.

The cost comparison deserves the same suspicion. In constant FY20 dollars the RPS case is estimated at 350.7 million dollars against 380.9 to 395.9 million for the fuel-cell cases (Tables 5.1 to 5.4), but the estimate's own footnote states the GPHS blocks are assumed government-furnished equipment at no included cost.¹ Forty-eight flight GPHS modules represent a substantial share of the nation's Pu-238 output, priced at zero here. The fuel-cell cost, by contrast, was estimated from human-rated Gemini-era data with a 2 percent per year technology reduction factor (Section 5.2).¹ Both cost numbers are assumption-bound; neither is a quote.

And the mission assumptions flatter the RPS. The solar option is dismissed at a northern mid-winter landing with constant optical depth 1 tau and 14-hour nights; the study itself notes a southern-hemisphere mission cuts required array area by about 20 percent and battery mass by about 30 percent (Section 4.1.3.4).¹ Solar was rejected on a wind-risk judgment, not a mass penalty: the wing was only 140 kg.

What rover power trades mean for power in space

Non-obvious implication: for mobile surface assets, the daily load profile, not the average, sets the storage and generation split. This rover's entire power architecture flows from one number: 10 kW for one hour. The RPS is sized to the 23 parked hours, the battery to the single drive hour, and the 80-percent depth-of-discharge floor exists because day 21 must end with enough charge to limp home (Section 4.1.3.1.2). Any future in-space or surface vehicle with bursty loads, including robotic heavy haulers or crewed sortie vehicles, inherits the same structure: a small continuous source plus a battery sized to the peak, with mission duration deciding whether isotopes beat reactants.¹

Genuine threat: the study quietly institutionalizes a dependency. Its RPS win assumes DOE heat sources are a free good delivered as GFE, and its failure-tolerance assumption pushes the RPS itself into a zero-fault-tolerance category with a rescue third rover as the backup plan (Sections 2.1.3, 4.1.3.1.2).¹ If surface mobility architectures converge on this pattern, Pu-238 supply, Stirling maturation, and acceptably light rescue logistics become single-point program risks shared across every surface mission.

Opportunity: the fuel-cell cases contain the more transferable engineering. The Case 2 water-for-reactant swap, with the lander electrolyzing product water at 500 W over six months between sorties (Section 3.2), is a closed-loop energy logistics pattern that works with any primary power source at the outpost, including fission.¹ A landed power-rich base selling electrolyzed reactants to visiting rovers decouples rover endurance from rover power-system mass entirely; the rover then needs only a tank swap, and the 2.5-times mass penalty never leaves the ground.

The bottom line

Established by this study: on its stated assumptions, a dynamic RPS plus lithium-ion buffer is roughly 2.5 times lighter than a self-contained fuel-cell system for a 21-day crewed Mars rover excursion, and the comparison is transparent enough to audit line by line.¹ Established by nothing yet: the RPS itself, which the study rates TRL 3 to 4, and whose cost advantage rests on free plutonium heat sources and excludes the Stirling system's own maturation risk. The solar dismissal is a judgment call about wind risk on a 140 kg wing, reversible by landing site and season. What would confirm the baseline is a long-duration, Mars-environment dynamic RPS demonstration, GPHS pricing included; what would break it is either a Stirling qualification stumble or a storage chemistry that closes the system-level Wh/kg gap, since at 184 Wh/kg the battery is already the most mature part of the winning architecture.