Computer vision and map registration enable high-precision lunar landing that wasn’t possible with the radio-based techniques during the Apollo missions. Digital cameras, computers, algorithms, and global imagery unavailable to Apollo make this now possible. Map registration is the process of matching camera image features to a pre-built database of features from a map of the Moon. This allows for an absolute position reference which is critical for the minimization of drift that results from other techniques used to determine position. Furthermore, this database could be loaded onto a spacecraft such that it could navigate without communication to the Earth, allowing for landing on the far side of the Moon where radio-location from Earth isn’t viable. The database of key points is constructed from NASA Lunar Reconnaissance Orbiter image data and processed on Earth into a compact database optimized for location lookup. During orbit, the lander’s computer can ask for the points that it should be observing, and then compare those to the points that it actually sees in order to compute its actual absolute position.

Under drastic lighting conditions or in locations where map data is not available at a sufficiently high quality, we use image analysis between pictures taken by the lander over time. Looking at pairs of successive images, we identify the “features” of each image—the critical points of the lunar surface that are likely to be identifiable in a picture taken under varying conditions (i.e. changing light, rotation, scaling, etc.). These are marked by the multicolored circles in the video. Each of the features of the first image are matched against those of the second image to determine pairs of corresponding features. Once the location of specific features in both images is determined, the amount of motion and rotation can be calculated. These values are relative to the images themselves, but given an estimate of the relationship between the size of the image and the area of the lunar surface that it represents, the motion of the camera can be computed in meters.

This graph represents an estimated and actual X and Y trajectory. The red line in this graph represents a simulated trajectory of a lunar lander where the lander is at a constant altitude of 300 meters above the lunar surface and is moving away from a skylight visible at the start of the video. The blue line represents the estimated trajectory of the lander as computed by the software demonstrated in the first half of the video. As the software computes the change in motion based upon its own current position estimate, the drift of the estimated position relative to the actual lander position is compounded over time. This will be corrected by the map registration process to provide an accurate position over time.

The simulation of the Moon’s gravity is key in testing the performance of our rovers. On the Moon, an object weighs one sixth of what it does on Earth. The gravity offloader lifts five sixths of the rover’s weight by hoisting the rover from above its center of mass.

Real-time tracking ensures that the lift is always directly over the rover’s center of mass, enabling unfettered mobility. The tracking is currently accomplished using a camera and fiducial markers. These fiducials are unique black and white patterns that are easily printed and mounted on a rover or test rig. The current tracking experiment utilizes a cross-shaped rig mounted to a robotic arm.

In this experiment, the robotic arm moves while a stationary camera continuously snaps photos. The photos are processed using tracking software from reacTIVision. The fiducial movement data is then run through motion estimation code, producing a position estimate for the test rig.

These position estimates will control the movements and position of the crane hoisting the rover, enabling the crane to always remain over the rover’s center of mass.

First, carbon fiber sheets are cut into strips (figure 1). The carbon fiber strips are laid out on a table and covered in resin in figure 2. The resin hardens when it cures at high temperatures in an oven. The carbon fiber and resin are covered in a sheet of plastic and the resin is spread using a thin piece of plastic pushing against the plastic sheet in the direction of the fibers. The extra resin is pulled out of the carbon fiber in this fashion. When all of the fibers are soaked in the resin, the plastic sheet is pulled off (figure 3).

Next, the sheets are laid up against the foam mold (figures 4 and 5). The mold consists of two long foam pieces on either side of a metal piece (figure 6). The mold is constructed in this way so that the metal piece can easily be pulled out and the mold separated non-destructively from the carbon fiber after it has been baked. The carbon fiber strips are wrapped on the part at 30 degrees and 150 degrees so that the unidirectional fibers provide extra strength.

Last, the part is covered in plastic (figure 7 and 8). Tape is applied to seal the air in. A hole is put in the plastic so that the air can be sucked out of the plastic. When all of the air has been removed, the part is put in the oven (figures 9 and 10).

The final piece is in the picture below (figure 11).

Polaris (shown in the picture above) prospects for water at the lunar poles by using a drill to sample lunar soil and scientific instruments that detect water. The rover is capable of driving and avoiding obstacles autonomously including traverses into dark regions in the lunar pole’s long shadows. Polaris suspension includes raise and lower capability to vary chassis ground clearance to lower for drilling and raise for driving on rough terrain.

The Polaris Rover actuates a swing arm at each corner to raise, lower, and tilt the chassis. This vastly improves the ability to drive, work, and get out of trouble. Each swing arm is cut from a rectangular composite tube described in this blog.

Fabrication of the swing arm requires a high temperature cure resin, which limits the mold selection to metal or a high-temperature, high-density foam. Metal (aluminum) releases easily, creates a good surface finish, can produce many parts, but is expensive. Foam is cheap, easy to machine, but cannot withstand higher temperatures and will produce fewer parts as a result.

Thus, a 3-part mold was designed which consists of 2 foam blocks with a metal insert in the middle (shown in the picture above). The metal insert is removed after the part cures in the oven which provides space to extract the foam blocks.

In the picture below the high-density foam for the Polaris swing arm mold is machined on a 3-axis CNC machine.

The mold is sanded to remove waviness caused by the CNC bit, cleaned, and coated with two thin layers of 5-minute epoxy. Next, it is baked at 150 degrees F for an hour to fully cure the resin. Afterwards, the epoxy layer is sanded down to create a very smooth surface. The mold is cleaned again with mold cleaner or acetone and sealed with a mold sealant. Finally, multiple layers of release agent are applied to the surface of the mold to prepare for the composite layup. The release agent prevents the resin from sticking to the surface of the mold. The next step is to test the composite swing arm to determine what force must be applied in order for the part to fail.

A composite wheel has been developed. The wheel is designed and molds are created. A CAD model is used to create molds that are cut from foam. The molds are glued together and cut by a CNC machine, or robot arm. The mold is sanded down and a layer of 5-minute epoxy is applied. The epoxy is cured in an oven at 150 degrees F for an hour. The layer of epoxy is sanded down to a smooth surface and a second layer of epoxy is applied and sanded. The mold is cleaned and a layer of sealant is applied. The sealant dries for 24 hours. 3 layers of mold release are applied to ensure the composite-epoxy mixture does not stick to the foam mold.

A layup is performed when pieces of carbon fiber or Kevlar are soaked with epoxy and the excess is removed. The strips are laid in a pattern in the mold to make the part the right shape and thickness. The time-lapse above details this process.

The part on the mold is vacuum bagged so that it cures at high temperature and pressure. The part is placed in the oven at 250 degrees F for 8 hours. Once it cools, the composite is hard. It is removed from the foam mold and trimmed to create the final wheel.

Below is a video presentation of composite wheel design and manufacture.

Ramps have successfully shown deployment and egress of the rover from our lander in full Earth gravity.  This comes at the cost of substantial over-design and mass too great for flight.  The need is to lighten the design for flight and test that at the lower dynamics and impact forces that will be experienced in moon gravity.   Since ramp hinges bear the off-axis forces and moments, it is common to simulate a reduced-gravity deployment by tilting a swing-deployment at a an inclination to Earth gravity consistent with the desired proportion of gravity reduction.  To simulate moon gravity at 1/6 that of Earth gravity, the inclination angle is 9.5 degrees.

The three-meter aluminum ramps are u-channels of .04 inch thickness.  Lightweighting was accomplished by reducing the thickness of the webs and and flanges, and by perforating with triangular cutouts.

An additional bend shown in Fig x would add a remarkable stiffness to the ramp, however, this would make it difficult to observe deflections and deformations without proper instrumentation.

The walls of the channel are under compression loading at the deployment impact and during the rover egress. As the thickness of the profile decreases for mass optimization, buckling is the dominant failure mode and it becomes much more significant than yielding. The purpose of this test was to observe the buckling problem and to compare ramp deployment impact load and the static rover egress load.

The static load created by the rover is 50N which is equivalent to placing a 5kg mass in the middle of the ramp. To simulate moon gravity for the deployment impact load, the ramp hinge is mounted at a 9.5deg tilt from the vertical axis as shown in Fig x. This creates the desired gravity component at the rotation axis of the ramp.

Preliminary tests showed that although the impact load is less significant than the static load, it creates noticeable deformations on the ramp. Thus, a method is developed to mitigate the impact. A thin strip of aluminum is mounted to the end of the ramp. This acts as a spring and an energy absorber. Some of the energy is dissipated by plastic deformation, and the rest is stored and released. This creates a bouncing motion that settles by natural damping. Tests exhibited the effectiveness of this method for reducing the ramp’s ground impact force to a negligible level.