I was under the impression (when asking this question) that over the past four years that experience by ground operators and programmers had lead to a configuration where the Mars rover Curiosity could do a significant amount of "work" or movement autonomously - meaning not necessarily requiring somewhat-near-real-time comms. link with Earth, where "somewhat-near-real-time" means as fast as the current Earth-Mars separation will allow at the time.

For example, if a plan for a movement by a few meters, or tens of meters is decided, is Curiosity able to do it? Must the path be pre-planned and validated by people on Earth, or is Curiosity able to plan a path - including obstacle avoidance, and execute it.

What about drilling and sample collection, or laser ablation and chemical analysis - once a plan is in place, can these run autonomously without comms. for some period of time?

If so, I'm also interested in how commonly these are actually done autonomously, but the my main question is if it has been shown to have this capability.


3 Answers 3


AutoNav is simply given a set of waypoints, which the software then has to, on its own, plan and execute a route. You can imagine the waypoints as lawn darts on the 3D terrain map built from the previous sol's images. In fact, in the rover planning tool, the waypoints are displayed as actual lawn darts.

The plan is executed about a meter at a time, heading in the direction of the waypoint as permitted by the terrain. At each step the rover will build a safety map of the terrain ahead and go around obstacles as needed. The drive can terminate before reaching the final waypoint if either no progress can be made safely, or the allotted time has been used up. Regardless of how the drive ended, images are always taken of the final state for use by the rover planners on Earth for the next sol's plan.

Yes, arm, drilling, sample movement, instrument operations, etc. can all be executed autonomously over a sol or longer. Those are generally well-specified, deterministic actions (unlike waypoint driving), and the sequence terminates if any unforseen problems are encountered. Drilling and sample movement activities are done over several sols so that the operators on Earth can be in the loop and verify proper progress along the way.

The rovers are designed to be autonomous enough to keep themselves busy for a full sol. They normally get only one set of commands direct from Earth when they wake up in the morning, and then send back the results the sol's activities in the afternoon through a relay orbiter, before going to sleep.

  • $\begingroup$ OK excellent - this is exactly what I wanted to know. Amazing! $\endgroup$
    – uhoh
    Commented Sep 18, 2016 at 5:50

Not only has it autonomous driving capability, just very recently in 2016 it acquired new features for 'autonomously' targeting ChemCam and autonomosly re-schedule some of the science observations by itself. All based on pre-defined priority rules.

The auto-targeting software component is called AEGIS. Previously tested and enabled autonomous driving component is called AutoNav, which is being used since 2013. EDIT: as per comment below, AutoNav goes back to MER's, Spirit team first used the software piece in February 2004.

It is important to understand there are different levels of autonomy in these processes. For example, important component of AutoNav is simply estimating how far the rover actually moved during driving. Previously this was done back on earth by processing navcam images and wheel odometry data, and now AutoNav is pretty much able to keep an accurate estimate during the drive. This enables longer, bolder drive plans to be uplinked as the rover can be trusted not to be too far off its planned path, by 'closing the control loop' locally. This level of autonomy does not mean a lot of high level planning done by robot.

  • $\begingroup$ AutoNav has been in use since 2003, on the Mars Exploration Rovers. $\endgroup$
    – Mark Adler
    Commented Sep 18, 2016 at 1:45
  • $\begingroup$ Amended the post. I remember Curiosity switched this on in 2013, according to this paper Spirit got the software in 2005 $\endgroup$
    – kert
    Commented Sep 18, 2016 at 2:08
  • 1
    $\begingroup$ Great! Can you help me understand more about driving - does Curiosity ever choose, or at least modify a recommended path from point A to point B on its own? For example during a drive, close inspection of an obstacle may reveal a problem - does Curiosity ever decide to go around something and just do it on it's own? Also can the ground issue a command "drill here, collect a sample of power, and analyze it using method X" and Curiosity executes this autonomously? $\endgroup$
    – uhoh
    Commented Sep 18, 2016 at 2:40
  • $\begingroup$ I'm looking for an answer to "How much can the Mars rover Curiosity do autonomously" where the operative word is "do", rather than a list of names of software. $\endgroup$
    – uhoh
    Commented Sep 18, 2016 at 3:20
  • 1
    $\begingroup$ @kert No, the 2005 upload was just a bug fix and hooking in visual odometry, per the paper. The AutoNav software was there and validated on the rovers on Earth in 2003. I directed the first use of Autonav on Mars in 2004 on Spirit on Sol 36. $\endgroup$
    – Mark Adler
    Commented Sep 18, 2016 at 4:47

Supplementary answer; a day in the life of Curiosity, from Phys.org's How to drive a robot on Mars:

Around 9:30 Mars time, a message arrives from California, where it was sent 15 minutes earlier.

"Drive forward 10 meters, turn to an azimuth of 45 degrees, now turn on your autonomous capabilities and drive."

The Curiosity rover executes the commands, moving slowly to its designated position, at a maximum speed of 35 to 110 meters (yards) per hour.

Its batteries and other configurations limit its daily drive span to around 100 meters. The most Curiosity has rolled on Mars in a day is 220 meters.

Once it arrives, its 17 cameras take shots of its environs.

Its laser zaps rocks. Other tools on board drill into a particularly interesting rock to study small samples.

Around 5 pm Martian time, it will wait for one of NASA's three satellites orbiting the planet to pass overhead.

Curiosity will then send several hundred megabytes of scientific data via large ground antennae to its human masters on Earth.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.