Surface Mobility on Small Bodies
Why explore small bodies anyway?
A challenging environment to traverse
So how can we utilize the mobility subsystem (flywheels, motors, and brakes) to move from one point on the surface to another in a deliberate way? This problem is well understood for wheeled rovers, however remains largely unexplored for internally-actuated platforms, especially in microgravity. Using an iterative approach, we have studied the hybrid’s dynamics, created simulations, and performed experiments, to refine our control techniques and achieve targeted mobility.
Having three orthogonal flywheels allows the rover to produce torques and manipulate angular momentum around arbitrary axes. This enables a wide variety of possible maneuvers, or motion primitives. Hopping is the primary maneuver for covering large distances, where the direction can be controlled using a combination of flywheels. Tumbling and pointing maneuvers, on the other hand, produce small movements for precise control and pointing. We also inadvertently discovered a “tornado” escape maneuver that allows the rover to leap vertically, to escape from sandy sinkholes or fissures.
While hybrid rovers can reliably execute these motion primitives, some randomness and uncertainty remains in their final landing position, mostly due to uneven terrain and surface properties. Golf provides a nice analogy; large hops drive the rover down the fairway towards a target, while small tumbles are used like a putter on the green (the tornado maneuver is like a sand wedge.) Like a golfer, our hybrid control strategy compensates for these random errors while maneuvering.
Communication latencies and blackouts are unavoidable in deep space, and with limited windows for interaction with the ground, traditional teleoperation approaches are inefficient. Thus, increased autonomy is desirable to ensure efficient deployments. Hybrid rovers are designed to perform motion planning onboard, enabling navigating without interaction with the ground. We are designing computationally efficient algorithms that, combined with recent advances in microprocessor technology, can plan sequences of maneuvers to navigate hybrid rovers towards a target, while avoiding obstacles and considering uncertainty in our dynamic and environmental models.
Simulations are invaluable for studying many aspects of mobility, from controlling motion primitives on a local scale, to the execution of planning algorithms on a global scale. The video on the left shows one of our simulations; an interactive tool that models hybrid rovers on the surface of small bodies (Asteroid 25143 Itokawa is shown). The user can directly input torques to the flywheels or define targeted destinations and let the motion planning algorithms determine how to actuate the platform.
Note: (1) the surface color reflects the local slope of the terrain and (2) the simulation is run at 100 times real-time (motion on small bodies is slow!)
So how can we actually test a rover specifically designed for microgravity here on Earth? There are three ways in which we have approached this problem: (1) over-designing a prototype so that it can operate in a 1g environment, (2) emulating microgravity in the lab with gravity-offloading test beds, and (3) reduced gravity parabolic flights. No single approach can exactly replicate the asteroid environment, but each approach has complementary benefits.
It turns out that the transfer of energy from the flywheels into a hop is a terribly inefficient process (~1%), which is fine on a small body with barely any gravity to overcome, but highly restrictive on Earth. Nonetheless, we have developed a prototype optimized for efficient mobility, which can indeed perform short hops and tumbles in 1g. For comparison, the maneuvers you see in this video would translate into hops over a football field on Phobos!
Testing on Earth does not truly capture the dynamic interactions we would expect in a microgravity environment. In order to “turn down the gravity knob” in the lab, we have developed a first-of-a-kind 6 DoF gravity-offload test bed, capable of emulating microgravity down to 0.0005g and within a 1.5m x 3m x 1m workspace. The active gantry control together with passive compliance allows smooth motion tracking even during impulsive force inputs (as seen during hopping and ground collisions).
Test beds allow for iterative testing of control algorithms in a controlled pseudo-microgravity environment, but they inevitably introduce exogenous dynamics. A more accurate microgravity analog can be achieved in a “reduced-gravity airplane,” which flies in parabolic trajectories to provide 20 second periods of near-zero g’s inside the cabin. We took our prototypes aboard 4 of these flights, for a total of nearly 200 parabolas, in which we tested various maneuvers on several surfaces within a sealed experimental chamber. Of particular interest were the experiments on granular media (i.e. sand), which behaves very differently in microgravity.
- Dynamics and control: [JFR 16], [ISER 16], [FSR 15], [iSAIRAS 14], [ICRA 14], [ICRA 13]
- Autonomy – Motion Planning [ISRR 17], [iSAIRAS 18] and Localization [Aerospace Conf 18], [ICRA 18]
- Mission architecture and science case: [Aerospace Conf 13], [Aerospace Conf 12]
- A detailed report about the project [report]
- Slides presented at the NASA NIAC 2015 Symposium [Presentation]
- Slides presented at the NASA NIAC 2012 Symposium [Presentation]
In the press
This project brings together a strong team of experts in astronautics, human-space flight, science, and engineering from Stanford, JPL and MIT, and engages graduate students at both Stanford and MIT.
Julie C. Castillo-Rogez
Planetary Scientist, Co-I and JPL lead
NASA Jet Propulsion Laboratory
Systems Engineer, Co-I
NASA Jet Propulsion Laboratory