Zhurong

CNSA, via China News Service.

Overview

Zhurong is China’s first Mars rover, delivered to Utopia Planitia by the Tianwen-1 entry module and released onto the surface on 22 May 2021 [1]. Its suspension, unlike the passive rocker-bogie carried by every American Mars rover, is an active suspension that can change its own geometry on command: a main and auxiliary rocker arm linkage with a clutch that, once released, lifts a sunk wheel or contracts the whole chassis rather than relying only on wheel torque to climb out of trouble [1]. The rover had operated 347 Martian days against a 90 day planned life by 5 May 2022 [11] and entered hibernation on 18 May 2022 without waking again [2].

Where the design sits among Mars rovers

Mars surface rovers fall into a few families by mobility, autonomy and operations. The flown NASA rovers share a passive six-wheel rocker-bogie: Sojourner, the Mars Exploration Rovers, Curiosity and Perseverance [12], [25], [31]. The JPL FIDO rover was the field prototype for the 2003 Mars Exploration Rovers, with a six-wheel rocker-bogie, a mast and a front arm, and was exercised in blind field trials in 1999 and 2000 that reported no quantitative traverse or localization results [18]. Zhurong departs from that lineage in the suspension, not in wheel count. Its power system, thermal design and operating cycle are documented in the sections below from a single mission-design source [1]; the comparison sources are mostly NASA descriptions at design-nominal level [31], [25].

Autonomy is the family where the held record is least balanced. Perseverance has flown Enhanced Autonomous Navigation since 2021, with two-stage path selection, collision checking, wheel-slip bounding-box expansion and FPGA-accelerated stereo, and reached 32.1 km at an average of 92 m/h by sol 1312 [16]. Its onboard planner, Mars 2020 Simple Planner, was deployed on 5 October 2023 [23]. Sojourner was driven from the ground with stereo waypoints and a flying-camera terrain model, with every command sequence built in one ground workstation [27]. Zhurong is recorded as supervised, driving one command load per sol, and no published onboard navigation architecture for it is held in the corpus. Perception limits are documented only for the American system: flight block-matching stereo leaves holes large enough to stop autonomous driving at solar phase angles under 20 degrees and on low-texture terrain, which 16,768 Perseverance NavCam pairs showed and zero-shot deep stereo models were then tested against [28]. Whether Zhurong’s navigation cameras share those conditions is not established.

Specifications

Parameter	Value
Dimensions	approximately 3.3 x 3.2 x 1.85 m
Power source	solar, four deployable panels
Suspension	active, main and auxiliary rocker arms with a releasable clutch
Payload	six instruments

Specifications from [1]. Rover mass is not published.

Mission profile

Parameter	Value	Source
Landing	15 May 2021, Utopia Planitia	[1]
Rover egress from the platform	22 May 2021	[1]
Planned mission	90 Martian days, completed 15 August 2021	[11]
Odometer traverse to 5 May 2022	1921 m over 347 Martian days	[11]
Reconstructed true-path traverse	2009 m over 206 site operations, 360 sols	[2]
Data returned by rover and orbiter to 5 May 2022	about 940 gigabytes	[11]
Relay test with ESA Mars Express	November 2021	[11]
First solar conjunction outage	mid September to late October 2021	[11]
Status	did not wake from hibernation, 18 May 2022	[1]

The 88 m gap between the 2009 m of true path reconstructed from site imagery and the 1921 m odometer reading is attributed to forward wheel slip on a generally downslope, southward traverse [2].

Mobility

The suspension’s clutch, once released, converts a normally passive linkage into a commandable mechanism with four functions: controlled lifting, controlled lowering, wheel lifting, and suspension contraction [1]. That is a deliberate departure from the JPL rocker-bogie lineage, which excludes springs and active joints altogether because an elastic or actuated link raises the normal force on a climbing wheel while unloading the others, exactly wrong for obstacle climbing on a passive chassis [12]. That design was tuned against six or seven modeled obstacle positions, where six wheels climb an axle-high vertical face at a friction coefficient near 0.8 against 0.78 to 1.4 for four wheels depending on which end meets the wall, a fixed-geometry optimum Zhurong’s clutch trades away [12]. Zhurong trades that climbing optimality for a self-extraction capability the passive linkage cannot offer.

Failure condition	Recovery action
Single wheel sunk	lift that wheel
All wheels caught	contract the suspension
Body in contact with a rock	raise the suspension to clear it

Recovery actions from [1].

Suspension contraction with sequential wheel lifting produces a walking gait that extracts the vehicle from soft ground by changing which wheels carry load rather than by increasing drawbar pull. That addresses slip-sinkage directly: a wheel that is unloaded and repositioned does not have to develop thrust against soil whose shear strength has already been exceeded, the same failure mode a rigid rocker-bogie has no recourse against beyond driving through it [12]. Terrain along the traverse at Utopia Planitia is flat, covered by reddish soils and dust, with dark-toned material exposed where wheels or wind removed the dust layer [3].

Localization during the traverse combined onboard dead reckoning with cross-site visual matching: dead reckoning against a HiRISE-visible rover track carried an error of 3.11 percent of traverse length, cross-site Affine-SIFT visual localization reduced that to 1.75 percent, and matching site imagery to orbital imagery at four waypoints reduced it further to 0.50 percent [2]. Site digital elevation and orthophoto models were built at 0.02 m grid spacing out to 15 m range, with total stereo measurement error of 0.108 m at that range [2].

Power and energy

Solar-powered Mars rovers have flown twice at scale. The MER pre-launch design used a direct energy transfer bus and GaAs triple-junction arrays sized for 300 W at Mars and 140 W on sol 1, which are design figures, not flight measurements [19]. No Zhurong array output or battery figure has been published.

Solar radiation intensity at Mars orbit is approximately 43 percent of that near Earth [1]. Atmospheric dust adds to that deficit: background column optical depth over the aphelion half of the year sits below 1, and dust storm activity recurs seasonally, so array output carries a seasonal floor independent of surface deposition on the panels themselves [7], [8]. Two measures respond to the combined problem. Maximum power point tracking improves conversion efficiency from the cells, and an ultra-thin solar film with an anti-dust coating gives a dust removal efficiency over 86 percent, addressing deposition rather than the seasonal atmospheric floor [1].

A dormancy and wake-up mode covers the range of surface lighting conditions. On the onset of a dust storm the rover powers off to prevent battery drain and remains asleep until generated power and temperature reach their thresholds simultaneously, a coupling that makes the mode a survival state rather than a low-power state [1]. During a global dust storm, column optical depth reaches a global mean near 4 with local values of 5 to 10, so a storm subtracts from generation at the same time as it adds to deposition on the arrays [8]. The dust climatology behind that figure was itself validated against independent measurements from CRISM in orbit and from PanCam and Mini-TES on the ground, which is why it is used as the dust forcing in the Mars Climate Database [8]. The 2018 global storm encircled the northern hemisphere in its first 7 to 10 sols in a data assimilation reanalysis, with winds and dust above 4 km inferred by the model and not observed [22].

Thermal

Surface temperature below -100 C sets the design case [1]. Measured air temperature at landed Mars sites spans roughly 140 to 300 K depending on season and latitude, with a diurnal range large enough that a surface vehicle crosses most of that span every sol [7]. At Gale crater, the diurnal amplitude of the ground temperature cycle runs on a thermal inertia of roughly 170 to 550 SI units depending on terrain, the same site-dependence that makes any fixed thermal design a compromise across a traverse [7]. The architecture combines active and passive control.

Element	Function
Solar window	collects heat during the day, raises cabin temperature at night
Aerogel plus CO2 isolation	passive insulation: reduces conduction, isolates radiative leakage, eliminates in-cabin convection
Active heating at night	brings in-cabin equipment to its required ambient temperature
Active control on external equipment	navigation and obstacle avoidance cameras, multispectral cameras

Thermal elements and their functions from [1].

The passive layer is what makes the active budget affordable: with limited active thermal resources, leakage is attacked on all three transfer paths at once rather than compensated by heater power [1]. That approach to the cold is not the only one flown to Mars: the InSight HP3 mole, a hammering probe rather than a thermally isolated cabin, failed to reach its planned depth because the cohesive duricrust at its site gave no lateral soil pressure and therefore no hull friction to react the hammer’s recoil, a soil-mechanical failure mode with no counterpart in Zhurong’s cabin-insulation design [9]. The nuclear-powered Mars 2020 rover takes a third route: a heat rejection system fluid loop and actuator warmup heaters sized from a Holden Crater design environment, with the results at critical design review being model predictions rather than flight or chamber measurements [14].

Communications

Mars surface links are a relay problem. A mission-by-mission survey inventories the radio links of every orbiter and lander and adds ray-tracing simulation over Jezero terrain showing terrain-dominated surface propagation, with no Mars surface radio measurement to validate it [21]. For the NASA fleet the relay network works through manually scheduled overflights, with a planned move to delay-tolerant networking that is an expectation and not a demonstrated capability [30]. Curiosity’s downlink envelope was the constraint on its data management until about sol 2200, when MAVEN and TGO relay support raised it [32]. The Zhurong record is smaller: about 940 gigabytes returned by rover and orbiter together to 5 May 2022 [11].

Link	Parameter	Value
X-band direct	Rover-to-Earth visible arc	approximately M04:30 to M16:00 each sol
X-band direct	Uplink rate to the omnidirectional antenna	as low as 7.8125 bps
UHF relay	Architecture	loop relay mode, two-way, through the Tianwen-1 orbiter
UHF relay	Near-Mars arc pass around noon	approximately 10 min
UHF relay	Martian night pass	approximately 5 h
UHF relay	Near-Mars arc rate	Mbps class

Link parameters from [1]. The 7.8125 bps direct uplink is a command channel, not a data path, so the relay carries everything of volume: telemetry, images and instrument data, all within the same near-Mars pass. The orbiter’s own magnetometer, MOMAG, continued returning solar wind and interplanetary magnetic field data through the relay period, cataloging two coronal mass ejections and three stream interaction regions at Mars between November and December 2021 in coordination with NASA’s MAVEN [10]. Each event’s magnetic field measurements were compared directly between the two spacecraft rather than merged, since Tianwen-1 orbits Mars while MAVEN samples the solar wind further out, and the two records agree on timing even where local field magnitude differs [10].

Modes of operation

The rover is in standby by default, with computers on and other equipment operating intermittently against a mission cycle template run onboard [1]. Standby rather than sleep is affordable because the passive thermal stack holds cabin temperature without continuous heater draw.

Interval	Action
Martian afternoon, approx 10 h	X-band omnidirectional receiver on, holding a basic uplink channel
Martian noon, approx 1 h	UHF receiver on for the near-Mars arc
Martian night, approx 5 h	UHF receiver on for the remote arc
M11:00 to M15:00	primary work period, ambient temperature highest and no additional heating needed

Daily template from [1].

Scheduling the work period against temperature rather than against illumination is a direct consequence of the thermal budget: heating an actuator costs energy the arrays would otherwise deliver to the activity itself.

Ground operations

Delayed instructions are uplinked each Martian morning via the orbiter, relayed to the rover by UHF around noon for execution, and telemetry is returned through the orbiter by UHF during the Martian night [1]. The cycle is therefore one command load per sol with a single execution window and a single return window, whereas the NASA rovers relay through the Mars Relay Network, whose orbiter overflights are scheduled manually [30].

Ground testing

The landing platform’s hovering and obstacle-avoidance test was completed on 14 November 2019 at a test field in Huailai County, Hebei, simulating Martian gravity, per a Chinese space-industry news item that gives no achieved offload accuracy [26]. No rover-level terramechanics data for Zhurong’s wheels is held. The nearest evidence is from lunar and laboratory wheel studies: a Tohoku single-wheel testbed with a discrete element model reproduced traction and sinkage for a Rashid-1 style grousered wheel, with the lunar gravity results coming from the model alone [24], and tests at speeds up to 1.0 m/s, about a hundred times flown rover speed, found traction speed dependent in a way low-speed testing missed [29]. Neither has flown, and neither describes Zhurong’s wheel.

Payload and instruments

Six instruments are carried [1]. Four measure the environment rather than the rover’s own state, and the meteorology package is the Chinese counterpart to REMS and MEDA [7].

Instrument	Measurement	Source
Navigation camera	surface morphology, driving
Multispectral camera	mineral and rock types, material distribution
Subsurface detection radar	soil structure and stratification, water and ice content
Surface composition detector (MarSCoDe)	LIBS for quantitative chemistry, short-wave infrared spectroscopy for mineralogy	[5]
Magnetic field detector	magnetic field intensity, variation, ionospheric current
Meteorological instruments	temperature, pressure, wind speed and direction, sound

Instrument list from [1] unless noted.

MarSCoDe’s laser is a passively Q-switched 1064 nm unit at 23 mJ per 4.5 ns pulse, firing at 1 to 3 Hz with fixed energy, feeding a 106 mm Ritchey-Chretien telescope that both focuses the beam and collects the plasma emission, pointed by a worm-gear mechanism driven by two stepping motors [5]. Emission is dispersed across three LIBS channels covering 240 to 340, 340 to 540 and 540 to 850 nm, while a short-wave infrared channel covers 850 to 2400 nm [5]. Because the rover is solar powered and cannot keep the spectrometer at a constant temperature, its spectra drift with Martian temperature; the instrument carries an onboard titanium calibration target, and an elastic particle swarm optimization algorithm matches the shifted spectral lines against a standard wavelength set established in a Mars simulation chamber before launch, cutting the residual wavelength error after correction to under 0.006 nm at the ultraviolet channel’s coldest tested point of -40 C [6]. That correction was validated in a Mars simulation environment chamber holding Earth-lab pressure near 1 percent of Earth’s own and temperatures from -40 to 30 C, using standard element lamps for the initial calibration and the onboard target for the temperature-shift correction itself [6].

MarSCoDe measured LIBS and short-wave infrared spectra of six Vastitas Borealis Formation targets, soils, an outcrop and a coherent rock, in the rover’s first 300 sols, and the result reads as a low-alteration mix of igneous minerals with clay-like alteration phases formed under limited or ephemeral water and cold conditions [3]. The magnetometer campaign returned vector measurements at 16 sites along a 1089 m track; the field measured, under 10 nT over most of the track, is far weaker than the roughly 81 nT that downward-continued orbital models had predicted for the site, though InSight measured about 2000 nT at Elysium Planitia, so the discrepancy is a local rather than a universal one [4]. Separating that faint ambient field from the roughly 10,000 nT generated by a permanent magnet inside the rover’s own mast-mounted camera required rotating the mast and the whole rover to isolate sensor-frame, rover-frame and Mars-frame components of the reading [4]. The same discipline applies to sample return: a retrospective assessment of the Perseverance sample caching system against its 0.5 mT limit concluded exposure above 5 Gauss is unlikely, from bounds on two of four actuator types and not a compliance verification [17].

End of mission

Zhurong entered hibernation on 18 May 2022 and did not re-establish contact [2]. Dust accumulation on the arrays is the assumed cause, against an 86 percent dust removal efficiency demonstrated for the coating and a wake logic requiring both power and temperature thresholds to be met at once [1]. Removal of deposited dust by wind requires a surface shear stress at or above the saltation threshold, which is not a schedulable event, and the storm conditions that would generate that stress are the same conditions that raise column optical depth and cut array output at the same time [8].

The same loss mode ended Opportunity, a solar-powered Mars rover. It was last heard from on sol 5111 (10 June 2018), when the 2018 planet-encircling dust event cut array output below the level needed to communicate; more than 1000 recovery commands were radiated before the search ended on 12 February 2019, and its end state is inferred from thermal and power models rather than observed [13]. For Zhurong the cause is assumed, not published.

Dust protection for moving joints is documented for a lunar vehicle, not a Mars one: the VIPER mobility system used a 42-layer outer barrier over four mobility modules and graded labyrinth, felt and PTFE seals, qualified at ambient pressure with no combined dust, vacuum and thermal test reported [20]. Zhurong’s joint sealing is not described in the held sources.

Across NASA surface missions, 48 major anomalies of four or more consecutive sols of lost progress cost an average of 11 percent of prime mission duration and 6.6 percent over the full mission [15]. That catalog covers American missions only, and no equivalent exists in the corpus for Zhurong.

Technologies developed

The active suspension is the one mobility capability on Mars that no NASA rover has carried. Its value is not obstacle clearance, which a passive rocker-bogie already provides by design [12], but a recovery mode that does not depend on developing more thrust: lifting a sunk wheel, or contracting the whole suspension, changes the load distribution instead of demanding more shear from soil that has already failed. The thermal architecture pairs a passive stack that attacks conduction, radiation and convection simultaneously with a solar window that stores daytime heat for night use, converting part of the survival heating load from an electrical draw into a thermal one. MarSCoDe’s onboard particle-swarm wavelength recalibration is a narrower but transferable lesson: an instrument that cannot be thermally stabilized on a solar-powered platform can still hold calibration by matching against a known target rather than by controlling its own temperature [6].

What is not established

Rover mass, the exact panel count beyond what is visible in imagery, and a definitive account of the wake failure, dust accumulation versus a hardware fault, remain unpublished. The magnetometer discrepancy between ground and orbital field estimates is attributed to demagnetization at either the scale of the whole Utopia Basin or of the smaller ghost crater the rover landed in, and neither is ruled out [4].