Feifei Qian and her team field testing at Mars-like White Sands National Park, New Mexico (Credit: Justin Durner)
For decades, scientists have been looking for answers about Mars: Is there water on Mars? Where did the water come from? Who used to live on Mars? Can we live on Mars one day?
But one of the biggest challenges to finding answers lies in how to gather information on the Red Planet. From challenging terrains and harsh climate conditions to extreme communication delays, collecting data on Mars is incredibly challenging.
Meet Spirit, a Robot dog (Credit: Justin Durner)
Now imagine a real-world R2-D2 — a robot companion that can not only trot through the heat and survive the cold while carrying heavy scientific instruments and collecting data but also solve problems and even team up with astronauts or other robot friends to complete a mission.
This is no longer a Star Wars sci-fi fantasy. A team of USC-led researchers have developed a highly mobile, autonomous robot dog named Spirit through three interconnected NASA-funded projects:
- Legged Autonomous Surface Science In Analogue Environments (LASSIE)
- Legged Autonomous Surface Science In Analog Environments for Mars (LASSIE-M)
- Temporarily, Robots Unite to Surmount Sandy Entrapments, then Separate (TRUSSES)
USC Viterbi School of Engineering and USC School of Advanced Computing’s Ming Hsieh Department of Electrical and Computer Engineering Assistant Professor Feifei Qian and her research group teams up with researchers from NASA’s Johnson Space Center, University of Pennsylvania, Oregon State University, Texas A&M, Temple University, and Georgia Institute of Technology to advance legged robot technology that will ultimately help answer scientists’ biggest questions about Mars and the Moon, and support astronauts in future planetary missions.
With her background in bio-inspired robotics, Qian is the lead investigator for LASSIE, and co-investigators for TRUSSES and LASSIE-M, while other team members bring expertise in cognitive development, modular robotics, multi-agent motion planning, and planetary science.
As the projects approach the final stage this year, the team field-tested the latest technological developments at New Mexico’s White Sands National Park, an analog environment that mimics extreme terrain on Mars. The legged robot was previously tested in a Moon-like environment at Oregon’s Mount Hood National Forest in earlier stages of the study.
Learn more about the White Sands field test in the documentary below:
Wheeled Robot vs. Legged Robot
Rover (wheeled robot) vs robot dog (legged robot) (Credit: Justin Durner, Venice Tang)
Most current missions use wheeled robots called rovers, which carry instruments and sensors to collect data as they wander across planets.
With their wheel design, traditional rovers often face physical restraints mid-mission as wheels get stuck in loose soil and struggle to travel on steep slopes, making traversing surfaces like Mars difficult.
On the other hand, legged robots like Spirit have high mobility and can climb steep slopes and avoid getting trapped in soft sand, thanks to their animal-inspired legs with high degree-of-freedom. Robot dogs can reach places that rovers and humans cannot, allowing them to scout risky and scientifically interesting areas on foreign planets.
A New Way to Walk on Mars
Unlike rovers, which explore with their camera “eyes,” Qian’s robot dog has terrain-sensing legs that enable it to “feel the surface,” like the way animal legs can feel the ground and tell if the sand is soft or compacted when we walk on the beach. These “force signals” gathered from the robot dog legs can be more insightful when exploring soft sand pits, as compared to visual data from cameras.
Building on this design, Qian’s group developed a “crawl-n-sense,” a new gait that allows robots to detect and measure planetary terrain properties at each step, significantly improving the robot’s capabilities and spatial resolution in gathering environment data.
Different from the faster-moving “trotting gait” commonly used by legged robots, crawl-n-sense operates one leg at a time and is highly precise in sensing terrain mechanics. Three legs support the body while the fourth leg probes or “punches” into the ground, and then, the robot alternates between all four legs, continuing the same motion.
This type of ground-property data collection is key to helping scientists understand Mars’ history of landscape evolution, and answer questions about possible biological traces and signs of water, as well as map unsafe zones for rovers and astronauts, by analyzing the softness of the sand.
Training Robots to Think Like NASA Scientists
Currently, it would take up to 24 minutes for a message to reach Mars from NASA’s mission control on Earth, and another four to 24 minutes for a response, according to NASA.
Qian pointed out this delay makes developing a smarter, more autonomous robot a necessity. Since most existing robots are preprogrammed and have low-level cognitive ability, in collaboration with Cristina Wilson, decision scientist at Oregon State University, Qian’s team aimed to train Spirit to operate autonomously — to think and communicate like scientists — so robot dogs can become reliable members of a NASA mission, capable of not only executing tasks but also adapting and making independent decisions relevant to different situations at hand. This starts by embedding higher-level decision-making models that mimic how scientists prioritize data collection.
“We are training the robots to come up with ideas about what data to collect, where to go next on their own and how to communicate information with scientists on Earth,” explained Qian. “This could allow scientists and astronauts to shift focus back on important Extravehicular Activities activities, instead of having to guide robots’ actions step by step during a mission.”
Qian’s team has programmed Spirit to alternate between different gaits. An important part of Spirit’s training involves the ability to determine when to switch gaits to optimize performance: the trotting gait is faster and allows researchers to cover a larger region in the same amount of time, while crawl-n-sense is enabled for higher-quality data collection in scientifically interesting areas. The key is for Spirit to independently decide when to sense and when to move on — without humans directing every step.
Robot Dog Can Now Carry Additional Sensors
As part of the LASSIE-M project, Spirit can now carry additional sensor payload in a backpack — the same type of sensing equipment that rovers typically carry. These sensor signals complement the leg-force measurements, allowing the robot dog to evaluate scientific hypotheses in real time.
Since robot dogs can potentially reach places that are difficult or risky for rovers and astronauts to access, the ability to carry this payload becomes an important capability to ensure critical measurements are taken in challenging areas. These additional payloads include multiple stereo cameras, which have two lenses to capture 3D images and “depth” information in addition to standard red, green and blue imagery. These cameras can be used for the robot dogs to plan their path, and look at the foot-ground interactions as it steps into the sand.
Feifei Qian and her team field testing at Mars-like White Sands National Park, New Mexico (Credit: Justin Durner)
An additional element of the study focused on teamwork between robots to collaboratively complete a mission and assist one another when facing failures. Mainly explored in the TRUSSES project, researchers incorporated a team consisting of one rover and two robot dogs, each playing to its strengths. The robot dogs are specifically trained to pause what they’re doing to help any team member when they get stuck.
The findings from all three projects will be published in a few upcoming papers. Among them, the paper titled “Scout-Rover cooperation: online terrain strength mapping and traversal risk estimation for planetary-analog explorations” reports the team’s deployment results at White Sands and NASA’s Ames Research Center. The team is also performing a technology demonstration at the Research Center in late March, to showcase the robot collaboration outlined in TRUSSES.
Qian plans to continue refining the legged robots’ mobility, sensing and cognitive intelligence to further strengthen their role in future planetary missions in upcoming research projects.
Published on February 24th, 2026
Last updated on February 25th, 2026