CNSA Extraterrestrial Body Landing Comprehensive Test Field

China Aerospace Science and Technology Corporation.

The central Chinese facility for surface-contact testing is the Extraterrestrial Body Landing Comprehensive Test Field at Huailai County, Hebei, described by CNSA as the largest of its kind in Asia [2]. The Chinese name is 地外天体着陆综合试验场; the English form used here is the one carried in the reporting of the 2019 test and has no operator-published equivalent. Its function is gravity offload: a three-dimensional follow-up system hangs the spacecraft from a cable and pulls upward with a constant force so that the net downward force matches the target body, while tracking the vehicle in three dimensions so that the cable stays vertical and adds no horizontal disturbance [1]. The system was built for Chang’e-3, extended for Chang’e-5 take-off testing, and rebuilt on full cable-parallel drive for the Tianwen-1 Mars lander. The Lanyue crewed lunar lander was verified there in August 2025 [4].

The problem the facility solves is not new. The Apollo program’s own answer, the Lunar Landing Research Facility built at Langley in 1965, hung a lander simulator from an overhead gantry on cables that supported five sixths of its weight so astronauts could practice descent under an apparent lunar gravity [5]. Every full-scale terrestrial gravity-offload facility built since inherits the same tradeoff: a suspension line can follow a moving vehicle and add little inertia to it, but the suspension hardware itself has mass, friction and stiffness, and none of that goes away just because the line is carrying less of the vehicle’s weight than gravity would.

Laboratory data

Parameter	Value
Operator	Not established in open sources. See the coverage note [1][2][3]
Location	Huailai County, Hebei Province, China [2][3]
Commissioned	Chang’e-3 era, upgraded for Chang’e-5, rebuilt for the Mars lander [1]
Type	Outdoor gravity-offload landing and take-off test field
Floor area	Not published. Ground drive units sit on a 60 m radius circle
Capabilities	Follow-up system, earlier builds, terrain
Simulant or terrain	Not published as a material
Instrumentation	Suspension-point encoder, gyroscopic inertial table, motor-torque rope tension
Ground truth	Cable tension and platform pose, measured independently of the article
Fidelity limits	1 g inside the vehicle; last 100 m of descent only. See below
Access	National program facility. No external user route published [2]
Cited by	zhurong, chang-e-5-sampler

Capabilities

Three-dimensional follow-up system

Parameter	Value
Working volume	20 x 20 x 80 m, driven at speed and stably [1]
Test article limits	Tension 0 to 20 kN, continuously adjustable. Dynamic capacity 3x static
Vacuum	Not applicable. Outdoors at ambient pressure
Temperature	Not controlled. Outdoor Hebei conditions [1][2]
Illumination	Not published. The published campaigns are daylight tests
Simulant or terrain	See terrain field
Slope	Not published
Gravity offload	Constant tension, accuracy 3 parts per thousand or better; tension error below 20 N [1]
Instrumentation	Suspension-point encoder and gyroscopic inertial table for rope inclination

Two subsystems do the work: a constant tension system, that is the gravity compensation system, using a servo motor and drum in a closed loop through a load cell so the compensating force can be set arbitrarily; and a horizontal follow-up system that tracks the article in the horizontal plane so the rope stays vertical and produces no appreciable lateral component [1]. A universal gimbal at the rope end preserves the article’s pitch, roll and yaw freedom, mounted so the rope tension line passes through the article center of mass. The Mars gravity acceleration simulated is about one third of Earth’s [2][3].

The Mars lander system replaced the gantry crane form with cable-parallel drive throughout. The main hoist has six ground drive units, each with two motors, two reducers and two rope drums, each drum pulling two ropes through sheaves to two swing wheels at 133 m height on the main structure, forming a parallelogram stay mechanism [2][3]. The upper stay system has six units guided by swing wheels at 84 m, the lower stay system six ground units guided by horizontal swing wheels at about 7 m. In total 18 cable-parallel drive units and 36 wire ropes connect to the platform, driven by 24 AC servo motors [2][3]. On the platform itself sit six AC synchronous servo motors: three in the fast horizontal follow-up device, a two-layer carriage with two motors on the upper layer and one on the lower, and three in the tension fine-adjustment device driving the drum that raises and lowers the suspension rope. The tension fine-adjustment stack carries three motors, two for unloading over a 0 to 20 kN range and one for closed-loop control [2][3].

The design requirement is stated as five items:

Requirement	What it covers
Structural load capacity	Static capacity above article, tooling and propellant weight; impact capacity 3 times that
Running volume	Set by the landing and take-off maneuvering of the article
Velocity and acceleration	Horizontal and vertical, sized by obstacle avoidance under radar and image guidance
Constant tension control	Range, accuracy, control angle, disturbance force amplitude and sinusoidal period
Control response	Response time, tracking accuracy and overshoot per axis, bus delay, sensor bandwidth

Earlier builds

Parameter	Value
Working volume	Chang’e-3: 100 m rail travel, 16 x 16 m horizontal following
Test article limits	Chang’e-5 test masses 1.5 t landing, 0.8 t take-off, on a 20 t platform
Gravity offload	Same constant-tension principle, on a gantry crane form
Instrumentation	Six main hoist wire ropes; six ground drive units on a 60 m radius circle

The Chang’e-3 system used a three-level hierarchy in which stroke and mass fall and speed and precision rise at each level. The first level covers 100 m of coarse travel on ordinary rails and wheels [1]. LandIR uses the same coarse-to-fine staging, across drop towers whose rail lengths run from 15 to 72 ft rather than one mechanism sized for every test [5]. The second, on servo motors with gear and rack, covers 16 m by 16 m with real positioning accuracy [1]. The third is the fast follow-up platform itself, hanging below the vertical main hoist on six wire ropes, carrying a two-axis table on servo motors, large-lead ballscrews and precision rails, and a tension fine-adjustment unit made of two motors rigidly in parallel: a large one that carries most of the article weight as a coarse unloader, and a small one running closed-loop on tension so that the fine control has a usable range [1]. Platform stiffness comes from six ground drive units on a 60 m radius circle pulling six stay cables up to the platform, with tension controlled in real time.

For Chang’e-5 the same architecture was extended for take-off tests. Upper stay ropes leave their sheaves at 80 m height and 70 m horizontal radius, lower stay ropes at 5 m height and 60 m radius, both in six symmetric sets at 60 degree spacing, and the platform connects to 18 ropes in total: six upper stay, six lower stay and six main hoist [1]. The addition recalls how ARGOS’s own EVA campaign grew by adding a suit-adapted gimbal rather than redesigning the whole gantry [7]. The vertical main hoist runs on two servo motors of 845 kW at 1250 rpm; the upper stay drive on six AC servo motors of 540 kW at 1750 rpm; the lower stay drive on six of 359 kW at 1750 rpm [1]. Wire rope is 22 mm diameter with 217 square millimeters of effective section [1]. Average set tensions at platform heights of 80, 50 and 10 m are 2.17, 2.18 and 1.36 t on the upper stays, 1.80, 2.10 and 2.10 t on the lower stays, and 10.00, 9.00 and 6.60 t on the main hoist.

Comparison to other active offload rigs

The Huailai follow-up system is a cable rig with the vehicle itself as the payload, which puts it in a different family from the offload rigs used for human EVA testing. NASA’s Active Response Gravity Offload System (ARGOS) at Johnson Space Center holds a constant upward force on a suspended subject through an overhead gantry robot over a test volume of more than 340 cubic meters, moving at up to 3.3 m/s and detecting faults in under 20 milliseconds [8]. Its three active translational degrees of freedom and three passive ones through a gimbal were chosen to remove the two-degree-of-freedom and high-overhead- inertia limits of the older POGO rig, the roughly 30-second task window of parabolic flight, and the drag of neutral buoyancy [7]. Getting the gimbal pivot point wrong is not a minor error: across 14 test subjects and 68 pivot configurations, ARGOS engineers found that placing the pivot exactly at the payload’s center of gravity, the intuitively correct choice, left the subject unmanageably unstable and able to rotate upside down under full suspension; the usable zone was a narrow band 2.5 to 6.4 cm above and 1.3 to 2.5 cm aft of the modeled center of gravity, with fore-aft placement far more sensitive than vertical [6]. ARGOS also displaced its predecessor as the setting for gross-translation EVA tasks because gait recorded under neutral buoyancy differed from both 1 g walking and ARGOS walking, while ARGOS reproduced the same swing-stance split as 1 g ground walking with only a shifted mid-swing point [7]. Cutting the suspended gimbal’s own mass from 105 lb to 53 lb, by moving from a generic claw fitting to one adapted to the spacesuit, measurably improved how subjects rated the simulation, which is the same lesson Huailai’s designers draw when they size the suspended platform’s own hardware against the article it carries [7]. The rig itself is not small: ARGOS covers a working volume of more than 340 cubic meters, is human rated and two-fault tolerant, detects a fault in under 20 milliseconds, and moves its gantry robot at up to 3.3 m/s, specifications sized for astronaut and full-scale rover testing rather than the vehicle testing Huailai does [8]. NASA’s older Lunar Landing Research Facility gantry at Langley, from which the ARGOS design lineage descends, was itself a large structure at 240 ft high by 400 ft long by 265 ft wide, with a maximum drop height of about 200 ft and a bridge crane load of 65,000 lb once repurposed for aircraft crash testing after the Apollo program ended, with a hydro impact basin added in 2011 measuring 115 ft by 90 ft by 20 ft [5].

ARGOS has been in development since 2007 and is used for human performance and space suit evaluation, EVA procedure development, astronaut training, and reduced-gravity testing of humanoid robots and robotic rovers, a broader user base than Huailai’s single-program remit [8].

The German Aerospace Center’s LAMA facility, built to qualify the Philae comet lander’s touchdown gear, shows the same coupling from a different suspension geometry: a heavy-duty robot arm on rail, linear guide pillars, tension springs and a carbon-fiber beam, ending in a cardanic joint trimmed to the test object’s center of gravity. There the suspension itself was shown to perturb the very touchdown dynamics it existed to measure, with a hood tilt limiter specified to hold the lander to plus or minus 3 degrees of tilt actually permitting at least 9 degrees regardless of landing speed, a result that had to be carried into the landing stability analysis rather than treated as rig error [9]. The same LAMA campaign found the touchdown detection boundary itself was velocity dependent, triggering at 0.5 m/s on soft ground but only 0.2 m/s on hard ground, a distinction that a rig reporting a single threshold would have missed entirely [9]. A later retest quantified the effect directly against the qualification campaign’s 1996-2002 pendulum data: the suspension’s own stiffness, 4080 N/m, was more than sixteen times the pendulum’s 242 N/m, and that difference had to be corrected for before the offloaded data could be trusted [10]. That correction was only possible because the qualification campaign itself spanned 1996 to 2002, giving the 2013 retest years of pendulum-rig baseline data to check against [10]. Huailai’s own coupling between suspension friction and horizontal-following vibration, described below, sits in the same family of failure.

Terrain field

Parameter	Value
Working volume	Beneath the suspended platform. Extent not published [1]
Gravity offload	Supplied by the platform above, not by the field

The site has a prepared terrain field beneath the suspended platform, visible in the operator imagery, but no source found gives its composition, depth, grain size, preparation procedure or whether it is a simulant at all. What the facility controls is the vehicle’s apparent weight, not the surface it lands on [1].

Constant-tension suspension of the whole vehicle sidesteps a bias that a different offload method carries. NASA’s standard practice for wheeled-rover gravity offload, reducing the vehicle’s mass so its weight on wheels matches the target body while the soil stays under full Earth gravity, has been shown by coupled granular-dynamics simulation to give systematically overoptimistic trafficability: the soil grains are still pulled at 1 g, so the terrain is stronger than the low-gravity terrain it stands in for, and vehicles that climb a slope on Earth at moderate slip can see far higher slip at flight mass under the target gravity [11][12]. The gap is not small: in that simulation a 73 kg test rover climbing a 30 degree slope at about 42 percent slip on Earth-gravity soil corresponds to a 440 kg lunar-mass rover at about 85 percent slip on the Moon, above the roughly 80 percent threshold associated with wheel dig-in, and the same soil and vehicle at a 10 degree climb go from 18 percent slip to about 75 percent [11][12]. Huailai’s follow-up system holds the article at its actual flight mass throughout and adjusts only the net force through the cable, so it does not import that particular error, but the tradeoff is that the terrain field’s own properties are undocumented, so whatever bias the ground itself contributes cannot be assessed from the published record.

Instrumentation

The system runs in two modes. In active motion mode the cable-parallel drive moves the platform to commanded motion parameters, the two-layer carriage executes commanded two-degree-of-freedom horizontal motion, and the tension unit pays rope out or in under position or velocity control. In following mode the tension unit runs in force control, holding rope tension constant while paying rope in and out as the article rises and falls; the horizontal unit runs in position-following control so the rope inclination stays small; and the 18 cable drive units move the platform in three dimensions so that the carriage and rope strokes do not run out of travel [1].

Rope tension planning is treated as an optimization rather than a setting. The 36 rope tensions are planned by minimum-variance optimization, the objective being to minimize the sum of squared deviations of the upper and lower stay tensions from their set mean values, subject to bounds on each rope [1]. Because rope tension is regulated through rope length and measured through motor torque, the difference between actual and modeled rope length is calibrated on a regular rectangular grid across the working volume and interpolated at the eight vertices of the enclosing cell with bilinear basis functions [1].

What it does not reproduce

The method’s limits are set out by its designers. Suspension gravity compensation must overcome the friction and inertia of the suspension hardware itself, horizontal following is vulnerable to rope vibration and swing, and the two effects couple and constrain each other, so raising system accuracy and dynamic response is difficult and mishandling can drive the system unstable [1]. Every active suspension rig in this comparison set reports the same coupling under a different name: LAMA’s cardanic joint traded pitch and roll freedom for a touchdown-tilt limit that did not hold under load [9], and ARGOS’s gimbal-placement study exists because the fore-aft and vertical axes of that same coupling are not equally forgiving [6].

What the operators claim for the Mars campaign is narrower than an environment simulation: the 2019 run simulated the process of hovering, obstacle avoidance and slow descent under a Martian gravity field of about one third of Earth’s, and verified the correctness of the design [2][3]. The 2025 Lanyue campaign is described in the same terms, as verification of the landing and take-off system scheme, the control scheme, the touchdown cutoff scheme and the interfaces between guidance, navigation and control and propulsion, rather than as a reproduction of the lunar surface environment [4].

Stiffness is bought, not given. Because the platform hangs on flexible cable, the article’s inertia makes the platform hard to keep steady through speed changes, and stiffness has to be imposed by pretensioning ground stay cables. The pretension cannot simply be raised: larger pretension gives more stiffness but demands more motor power and cost, so the design target is the smallest pretension that meets the stiffness requirement, and the achieved horizontal stiffness still varies with height, falling from above 1.8 x 10^6 N/m to about 1.4 x 10^6 N/m when the platform is above 80 m [1].

The same stiffness-versus-cost tradeoff shows up wherever a vehicle hangs from cable rather than sitting on a rigid mount. LAMA’s designers accepted a suspension roughly sixteen times softer than the pendulum rig it replaced and corrected for the difference after the fact rather than building a stiffer rig [10]; ARGOS instead spent mass and complexity on an overhead gantry robot precisely to keep gimbal-induced motion artifacts small enough that subjects could still be trusted to judge simulation quality [7]. Huailai’s choice, high ground-stay pretension paid for in motor power, sits at the same point on the tradeoff as both, decided by what the vehicle being tested can afford rather than by any property of the suspension method itself.

The envelope is the last 100 m. The tests cover the final landing phase from about 100 m above the target surface, that is hovering, obstacle avoidance, slow descent and touchdown, and on the take-off side the constant-tension hold before mechanical unlock followed by constant-force following after ignition [1]. Everything above that altitude, and the atmospheric entry and descent phases for Mars, is outside the facility. The Apollo-era LLRF at Langley was built to the same kind of envelope: astronauts trained on a lander simulator suspended over a prepared moonscape, offloaded to five sixths of its Earth weight, rather than over the whole descent trajectory [5].

Gravity is compensated but not changed. The article remains in Earth gravity with an upward rope force removing part of its weight, so internal masses, propellant slosh and any gravity-dependent behavior inside the vehicle still see 1 g. The authors note as the reason for choosing suspension over water and air-bearing methods only that it follows actively and adds no extra inertia, not that it is exact [1].

Campaigns run there

Chang’e-3 soft landing verification. The Chang’e-3 spacecraft begins hovering at 100 m above the lunar surface to select a landing site, translates to the chosen point once the navigation sensor has fixed it, descends slowly, and below 4 m shuts down the main engine and free-falls to the surface [1]. The three-dimensional follow-up system reproduced the hovering, obstacle-avoidance and slow-descent segments on the ground, balancing part of the vehicle weight through rope tension so that the net upward force matched lunar surface conditions.

Chang’e-5 landing and take-off verification. The Chang’e-5 campaign added ascent testing to the Chang’e-3 test set, which drove the addition of the upper and lower stay cable drives alongside the second-level horizontal following and main hoist [1]. See chang-e-5-sampler.

Tianwen-1 lander hover and obstacle avoidance, 14 November 2019. The lander for China’s first Mars mission was run through hovering, obstacle avoidance and slow descent under a simulated Martian gravity field of about one third of Earth’s, as an integrated check of design correctness [2][3]. It was the first public appearance of the Chinese Mars program, watched by ambassadors and envoys from 19 countries and by representatives of the European Union, the African Union and the Asia-Pacific Space Cooperation Organization [2][3]. See zhurong.

Lanyue crewed lunar lander, 6 August 2025. The comprehensive landing and take-off verification test of the Lanyue lunar lander was completed at Huailai [4]. China’s first extraterrestrial landing and take-off test of a crewed spacecraft, it covered many test conditions over a long test period and verified the lunar landing and take-off system design, the control scheme, the touchdown engine cutoff scheme, and the interface match between the guidance, navigation and control and propulsion subsystems. Lanyue consists of a lunar module and a propulsion module, carries two astronauts between lunar orbit and the surface, and can carry a rover and science payloads [4].

References