The US space agency's rovers will get a software upgrade to allow them to make "intelligent" decisions in the study of Martian clouds and dust devils.
The new algorithms will give the robots' computers the onboard ability to search through their images to find pictures that feature these phenomena.
Only the most significant data will then be sent to Earth, maximising the scientific return from the missions.
Nasa says its robotic craft will become increasingly autonomous in the future.
"An instrument can acquire considerably more data than can be down-linked - this is a recurring theme on all spacecraft," explained Rebecca Castano, from the agency's Jet Propulsion Laboratory (JPL).
"The idea now is to collect as much data as the instrument can, analyse them onboard for features of specific interest, and then down-link only the data that have the highest priority," she said.
Currently, the rovers are allocated time to look for clouds and dust devils, which may or may not appear - they are naturally transient events. And getting humans to sift the images is time consuming.
The software upgrade, due to take place in the next month, will make the whole process much more efficient.
"Clouds typically occur in 8-20% of the data collected right now," Castano said.
"If we could look for a much more extended time and select only those images with clouds then we could increase our understanding of how and when these phenomena form. Similarly with the dust devils."
Leaving the robots to "get on with it" - to do the decision-making - is the way ahead, Nasa believes.
The agency's Mars Odyssey orbiter, which has been mapping the Red Planet since 2001, will get new autonomous flight software later this year.
This will give the satellite the ability to react to sudden changes on the Martian surface. It will be "tuned" to look for temperature anomalies, rapid changes in the polar caps, the emergence of dust storms and the formation of water-ice clouds.
If its algorithms mark an event of interest, the spacecraft will be able to break into its routine and take more images, without waiting for commands from Earth.
Scientists say this will capture short-lived, but highly significant, events that might otherwise have been missed.
The approach has been pioneered on Nasa's Earth Observing-1 satellite, which has now made thousands of autonomous data collects since 2003.