Reduced Gravity Testing

Four ground methods reduce the apparent weight of a test article: free fall in a drop tower or aircraft [1], mechanical offload of a fraction of the weight through a cable or counterweight [3], flotation on an air bearing over a level floor [7], and submersion in water at neutral buoyancy [5]. Each removes weight along a different axis, for a different duration, and each leaves a different residual that shows up in the data.

Duration and quality

Method	Duration per run	Gravity level	Axes offloaded
Drop tower, NASA GRC 2.2 Second Drop Tower	2.2 s, 27 m free fall inside a drag shield in open air, 8 to 12 tests per day [1]	Below 1e-5 g; the package still meets residual drag inside the shield [1]	All three
Evacuated drop chamber, NASA GRC Zero Gravity Research Facility	5.18 s, 132 m drop inside a 145 m evacuated chamber, one test per day	About 1e-6 g, with drag nearly eliminated	All three
Reduced-gravity aircraft parabola	20 +/- 5 s per parabola, about 45 parabolas per 2 to 3 hour flight [2]	Selectable 0, 1/10, 1/6, 1/3 or 1/2 g, each parabola bracketed by a 1.8 to 2.0 g pull-up [2]	All three
Active offload gantry, NASA JSC ARGOS	Unlimited [4]	Lunar, Martian or microgravity, set in software [4]	Vertical only, at one attachment point [5]
Air bearing floor	Unlimited	1 g normal to the floor is carried by the bearing; motion is free in the horizontal plane [7]	Two translations and yaw; three degrees of freedom [10]
Neutral buoyancy	Limited by consumables and crew, hours	Weight canceled by buoyancy in all orientations [5]	All three, at the cost of drag and added mass [5]

Package size follows from the facility: the 2.2 Second Drop Tower takes hardware up to 125 kg, while the Zero Gravity Facility takes up to 450 kg in a one meter drop bus, and deceleration at the end of a drop tower run can exceed 70 g [1]. Parabolic task duration cannot exceed roughly 30 s, and cabin volume constrains what can be flown [5].

Gravity offload rigs

A partial-gravity rig supports a fraction of the test article weight through a vertical line while leaving the remainder to be carried by the article itself. Pogo, the Johnson Space Center rig built from Apollo-era hardware, is the worked example: a pneumatic lifting cylinder hangs the subject from a gimbal support assembly through a chest harness and seat, and a two-stage vertical servo, a flapper-nozzle pressure amplifier driving intake and exhaust servovalves, holds the commanded lifting force as the subject moves [3]. Measured against walking and running subjects at lunar and Martian levels, Pogo held the simulated body force to within 2.2 to 10 percent, the variation depending on the type of locomotion input; mean stride frequencies at running speeds were 12 percent below the Earth values, and foot to ground reaction forces were 62 percent lower at lunar gravity and 51 percent lower at Martian gravity than on Earth.

The Active Response Gravity Offload System at NASA JSC is the current implementation. It is a robot built like an overhead bridge crane: the active machine supplies the three translational degrees of freedom, vertical and the two horizontal axes, while a passive gimbal at the attachment point supplies the three rotational ones, and every translational axis is actively driven so that the machine’s own inertia is kept off the subject [4].

Human testing has covered the MKIII prototype planetary suit and the xEMU with a mass-representative PLSS mockup [4][5].

What an offload rig distorts

The offload is applied at a single gimbal pivot point, so only the mass acting through that point is offloaded. Limbs and appendages that move outside the line of action are still accelerating against full terrestrial gravity, which is most severe in prone postures where the lower torso lies entirely outside the region served by the cable [5]. The system center of gravity is computed for a static neutral posture and moves when the subject bends, climbs or goes prone, and the gimbal cannot follow it dynamically or be readjusted in real time. Mapping the reachable pivot point envelope of the spreader bar gimbal showed large dead zones in which the pivot cannot be placed at all, so a pivot coincident with the system center of gravity was not achievable for every subject, and where it was achievable the resulting neutral equilibrium was hard to control and allowed the subject to rotate upside down about the pivot during initial suspension. Incorrect center of gravity alignment, a high inertia lifting mechanism, or too few degrees of freedom each produce false data: the earlier POGO rig gave only two translational degrees of freedom and carried substantial overhead inertia, with direct effects on measured task performance, which is why ARGOS drives all three translational axes and its gimbals carry an adjustable center of gravity [4].

The offload fallacy in terramechanics

Offloading a vehicle, or reducing its mass, changes the load the soil carries but does not change the gravitational body force acting on the soil grains. Terrain under Earth gravity has a higher shear yield point than the same terrain under lunar or Martian gravity, so a lightened vehicle on Earth soil is supported by material stronger than it will meet in flight, and the trafficability result is optimistic [6]. The Curiosity case is explicit: the rover was stripped to 340 kg from a nominal 907 kg so that its Earth weight matched its Mars weight, leaving the Mojave terrain under full Earth gravity.

Quantified against VIPER using continuum simulation correlated with Glenn SLOPE lab physical tests, on GRC-3b at a friction angle of 47.8 degrees and bulk density 1839 kg/m3 the 73 kg MGRU3 test rover climbs a 30 degree slope at about 42 percent slip on Earth, while the 440 kg VIPER on the same terrain in lunar gravity would run at about 85 percent slip, above the 80 percent threshold associated with dig-in [6]. On the softer case, friction angle 37.8 degrees and density 1627 kg/m3, MGRU3 climbs 10 degrees at 18 percent slip and VIPER on the Moon would see roughly 75 percent. The nominal, unlightened vehicle produced the same slip against slope curve on the Moon as on Earth, so the correct procedure is to test the flight-mass vehicle and transfer the result through granular scaling laws rather than to offload it [6].

The effect has been measured directly, in the one facility that changes the body force on the soil as well as on the vehicle. A single ExoMars flexible wheel was run in ES-2 Martian simulant aboard the NRC Falcon 20 parabolic flight aircraft at 10 to 70 percent slip, with wheel load held equal between gravity levels [13]. Maximum sinkage rose 38 percent in lunar gravity against 1 g at 164 N wheel load, and 27 percent in Martian gravity at 164 N and 47 percent at 225 N, all statistically significant at p of 0.01, 0.001 and 0.02 respectively. A second document from the same group reports the same 38 percent for the lunar case [14], so the two agree and there is no spread to reconcile. The authors record that the sinkage effect was far more conclusive than the drawbar pull effect at every gravity level.

The load sensitivity moves too, which is the part that does not transfer. Raising wheel load 37 percent, 164 to 225 N, raised maximum sinkage 14 percent in 1 g but 31 percent in Martian gravity [13]: sinkage grows almost proportionally with load in Martian gravity and much less than proportionally in 1 g. A pressure-sinkage exponent fitted at 1 g is therefore not the exponent that applies in flight, and neither is a load sweep run at 1 g. Read the magnitudes with the method in mind: sinkage was inferred from hub displacement on a flexible wheel assuming constant average deflection, not measured at the soil surface.

Air bearing floors

An air bearing floor cancels weight normal to the floor by floating the article on a gas film, leaving motion in the horizontal plane nearly frictionless. The Flight Robotics Laboratory at NASA MSFC was built around a 4000 square foot precision air bearing floor carrying an air bearing motion base with six pressurized air tanks feeding 32 thrusters and limited six degree of freedom capability, and a dynamic overhead target simulator consisting of an eight degree of freedom precision overhead X-Y crane with a six degree of freedom arm able to traverse the whole floor [7]. The Flat Floor Robotics Laboratory at the same center uses a 44 by 86 foot self-levelling epoxy floor [8].

What the floor reproduces is planar free-floating motion in three degrees of freedom, two translations and yaw [10]. Out-of-plane rotation, out-of-plane translation and orbital dynamics are absent, and the idealization assumed in control work is a perfectly flat floor with zero friction; any residual slope of the floor appears as a constant in-plane acceleration. Combining a spherical, a cylindrical and a flat air bearing with a counterweight has been used to add vertical freedom that a conventional air bearing system cannot provide [9].

The residual is measurable and it bounds every planar result. Three air pads carrying a 6.3 kg satellite model on an 1800 by 1800 mm glass table leave a horizontal acceleration of order 1e-3 g, the ratio of horizontal to vertical acceleration set by air film viscosity and described as unavoidable [12]. In the free-flying manipulator experiments run on that table, the drift of the system center of mass and the error between measured and computed base attitude were both attributed to that film rather than removed from the data, so a control law validated there is validated at 1e-3 g and against zero initial system momentum, which is what an air bearing emulates. It is not microgravity: full weight is still carried by the pads, and out-of-plane translation and two of the three attitude axes are absent.

Neutral buoyancy

Neutral buoyancy cancels weight in every orientation and gives untethered freedom of movement over a large work volume with several suited operators at once, which no other method offers [5]. It substitutes water drag and added mass for weight, so dynamic motions are damped in a way orbital motion is not, and slow motion is favored over fast.

The residual 1 g inside the suit is a measured consequence rather than a theoretical one [11]. The EVA community holds Neutral Buoyancy Laboratory training to a target of no more than ten hours of pool time per hour of EVA flown; across International Space Station assembly the ratio averaged 11.6 to 1, and the maximum training rate is three six hour runs in a five day week [11]. In the tiger team review of the EVA astronaut corps, suspected causes were collected for seventeen cases of training-related shoulder pain, and of the thirty-one causes named the inverted working position accounted for 25.8 percent and the planar hard upper torso for 22.6 percent; three astronauts in the study group had shoulder surgery, two of which were judged training related. The planar torso replaced the pivoted design between 1997 and 2000 after a pivot failure during a manned test [11]. It dropped the gimbal that had let the arm bearing rotate about two axes, and the fixed scye that replaced it cut overhead and downward reach. The suit that flew was the same in both cases; the injury mechanism came from loading the body in a 1 g orientation inside it for hundreds of hours on the ground [11].

Method selection

Free fall gives correct physics for a short interval and is the only ground method that offloads a granular bed as well as the hardware sitting on it, which is why the case against gravity offset rests on parabolic flight data and scaling laws rather than on gantry results [6]. Offload rigs give unlimited duration at the cost of a single point of application [5]. Air bearings give unlimited duration in a plane but only three degrees of freedom [10]. Neutral buoyancy gives unlimited orientation at the cost of water drag on dynamic motion and of a 1 g load path through the operator [11].