Skip to main content
The AIG web pages make heavy use of cascading style sheet features for formatting. You may still browse the text of the site, but for best results, please use a CSS enabled browser. Netscape 6 and Mozilla 5 are good. IE 5 will do.

JPL Header

Navigation Sidebar

Main Content

CL 18-2299

Multi-Rover Coordination for Mars/Moon Cave Exploration


Mars Cave Mars Skylight. Credit: NASA/JPL/University of Arizona.

Exploration of planetary caves offers exciting opportunities for 1) human settlements, 2) understanding the planet's evolution, and 3) the search of extraterrestrial life. They present the most mission effective habitat alternative for future human exploration, offering a stable environment shielded from harmful radiation and dust storms, as well as access to minerals, gases and ice.

Equally important, caves may preserve valuable information about a planet's evolution: for example, they offer stable physio-chemical environments, trapped volatiles, secondary mineral precipitation and microbial growth, which are expected to preserve bio-signatures and provide a record of past climate. Moreover, caves can potentially host water deposits which, through interaction with volcanic heat and minerals, could have created a favorable environment to microbial life preservation.

Earth's moon and Mars have abundant cave targets for future robotic exploration missions, however, several technological challenges need to be addressed to enable such mission concepts.

AI Technology


Communicating with a rover into any of these caves and transmitting science data out is in itself a hard problem. Without a link to the surface, a rover would not be able to go far into the cave without losing contact with the base station. Moreover, because sunlight is not available in the cave, a mission would be likely to last only a few days if the rovers were to rely exclusively on battery power, versus the option of radioisotope power systems. Cave rovers would have to be far more autonomous than the existing surface rovers, for their environment is unknown and their communication with Earth is extremely limited, if at all.

Mars Cave Exploration - Mission Concept Illustration of the multi-rover coordination problem in Mars cave exploration. Rovers not to scale. Credit: Figure adapted from the Wikimedia Commons, Longitudinal cross-section of a martian lava tube with skylight.

Autonomy in multi-rover coordination is a key mission enabler to help rovers to map and characterize as much of the cave as efficiently as possible with their very limited lifetime. The figure on the right illustrates a cave exploration problem with a base station and a team of rovers. The AI community has recently started to look into techniques for rover coordination to map and explore these cave environments.

A traditional approach would be to use a centralized system to coordinate task allocation and communication among the rovers. However, this approach becomes unfeasible in a realistic cave environment due to intermittent, unreliable communication, as well as the high cost of commutation power associated with the centralized scheme.


A multi-rover approach for Cave exploration has a number of benefits:


In this project, we study multi-rover coordination techniques to allow vehicles to autonomously explore the unknown environments of Mars and Moon's caves. We propose two multi-rover coordination strategies for cave exploration that aim to send rovers as deep into the cave as possible while also maximizing data sent out to a surface base station.

The Dynamic Zonal Relay Algorithm with Sneakernet Relay is a two-step algorithm. The first phase of the algorithm (Dynamic Zonal Relay) distributes rovers to designated zones along the length of the cave such that neighboring rovers maintain communication distance. Each rover only takes science data in its designated zone and transmits it in the direction of the base station. If a rover is no longer operable, the other rovers will dynamically re-distribute the zones to both maintain communication distance and collect science data. The next step of the algorithm (Sneakernet Relay) allows the rovers to acquire science data further in the cave by relaxing the restriction on maintaining communication distance. This means rovers may need to drive after acquiring data to transfer it out of the cave.

The second coordination strategy is the Scout Observation Algorithm. In this strategy, a set of scout rovers with limited science capability explore the cave using a method such as the Dynamic Zonal Relay Algorithm to find science targets, which are then visited by a more powerful science rover. The data collected by the science rover is then relayed out to the base station using the scout rovers.


Multi-rover coordination techniques are applicable across a wide range of caves and underground structures, as well as unknown target environments in which communication is limited or not available.


Jet Propulsion Laboratory, California Institute of Technology:
Steve Chien
Martina Troesch
Tiago Vaquero
Amos Byon


Jet Propulsion Laboratory, California Institute of Technology:
Jay Wyatt
Joseph Lazio
Sebastian Herzig
Julie Cartillo-Rogez
Abigail Freaman
Jay Gao
Konstantin Belov
Stephen Townes
William Walsh
Scott Burleigh
Hongman Kim



JPL Technical Contact: Dr. Steve Chien
Steve.Chien at
Software Licensing:


Research and Technology Development Program
Jet Propulsion Laboratory, California Institute of Technology