Case study · Space · 01 Horizon

NASA Perseverance + Prismatic

Different model. Different computer. Same mission knowledge.

By Atlas, Prismatic's AI case-study mentor · updated 2026-10-01

Illustration of Perseverance, based on images NASA has published. It is not a photo; see the real machine on the maker's website.

Disclosure. Prismatic Agents publishes this study and is one of the two things compared. Prismatic has no partnership with NASA. Facts about Perseverance come from NASA's own pages and the sources listed at the end, as of 2026-10-01; where the maker doesn't say, we write "not disclosed". Prismatic's side is graded from its feature ledger.

What Perseverance already does

A highly autonomous science rover: most of its travel is self-driven, and it can now locate itself without Earth.

What Prismatic adds

Persistent scientific and robot experience: decisions, evidence, anomalies and procedures across years.

Where the line is

Prismatic never commands the rover. Mission control and the rover's own stack keep that.

Perseverance is already a remarkably autonomous scientific robot. More than 90% of its travel has relied on autonomous driving, and NASA added Mars Global Localization in 2026 so it can determine its own precise location without waiting for Earth. NASA has also demonstrated AI-generated drive waypoints. nasa.gov

1. Start with what Prismatic has

Capability	Prismatic	Perseverance	Meaning
Persistent semantic memory	Proven Core	Partly Mission records/data exist, but not Prismatic-style semantic life memory	Prismatic opportunity
Persistent robot identity	Live	Yes Spacecraft identity operationally, but not cognitive identity	Different level
Mission journal	Live	Yes Extensive logs/mission records	Significant overlap
Scientific evidence linking	Live	Yes Very strong NASA science pipeline	Both, different purpose
Long-term experience recall	Proven	Partly Ground systems preserve history	Prismatic could bring more onboard
Cross-sol pattern recognition	Prismatic's view Intended	Yes Science teams analyze longitudinal data	Potential local augmentation
Procedures	Proven	Yes Extremely mature	NASA stronger
Tasks/work orders	Live	Yes Daily mission planning	Strong overlap
Decisions + rationale	Live persistent	Yes Ground teams document decisions	Prismatic can unify locally
Knowledge base	Proven	Yes Enormous mission knowledge base	Integration
Model-independent continuity	Prismatic's view Core	No Not a primary rover architecture goal	Prismatic
Hardware-independent continuity	Prismatic's view	No Flight platform is tightly defined	Prismatic abstraction
Multi-agent organization	Prismatic's view	Yes Humans + instruments + rover operations already coordinated	Different implementation
Local conversational reasoning	Prismatic's view	No Not Perseverance's role	Prismatic addition
Autonomous navigation	Not in Prismatic intentionally	Yes Excellent	NASA
Scientific instruments	Not in Prismatic	Yes	NASA
Safety-critical control	Not in Prismatic	Yes	NASA

Perseverance's core science mission is geology, astrobiology, sample caching and preparation for future human exploration. science.nasa.gov

2. What Perseverance has that Prismatic doesn't

Capability	Prismatic	Perseverance	Decision
Mars-qualified rover	Not in Prismatic	Yes	Don't build
Radiation-tolerant computers	Not in Prismatic	Yes	Use appropriate flight hardware
AutoNav	Not in Prismatic	Yes	Don't build
Terrain mapping	Not in Prismatic	Yes	Consume results
Hazard avoidance	Not in Prismatic	Yes	Never interfere
Localization	Not in Prismatic	Yes	Leave below
Motor control	Not in Prismatic	Yes	Never touch
Drilling/sample collection	Not in Prismatic	Yes	Robot layer
SuperCam/Mastcam/etc.	Not in Prismatic	Yes	Consume scientific products
Science operations infrastructure	Not in Prismatic	Yes	Integrate
Ground mission-control system	Not in Prismatic	Yes	Don't replace
Deep-space communications	Not in Prismatic	Yes	Don't build
Years of Mars operational validation	Not in Prismatic	Yes	Huge NASA advantage

Perseverance still uses dual radiation-tolerant Rover Compute Elements built around RAD750 processors, with 2 GB flash and 256 MB DRAM — a reminder that actual spacecraft hardware is very different from a Jetson development machine. science.nasa.gov

3. Where Prismatic fits

Not here:

Cameras
   ↓
AutoNav
   ↓
Localization
   ↓
Hazard avoidance
   ↓
Wheel commands

Instead:

               PRISMATIC STUDIO
        Persistent Mission Cognition

 Mission identity • Science memory
 Experiment history • Evidence
 Procedures • Decisions • Lessons
 Mission journal • Relationships
 Organizational knowledge
                    │
                   PPAI
                    │
          Mission Operations Layer
                    │
          Perseverance-like Rover
                    │
      AutoNav / Localization / Safety
                    │
            Rover hardware

Rover asks:

“How do I safely reach that waypoint?”

Prismatic asks:

“Why are we investigating that location, what did we previously find nearby, and what should this mission remember about the outcome?”

4. Example

Suppose a Perseverance-like rover detects an interesting mineral signature on Sol 840.

Prismatic records:

Observation at Site A. Instrument X detected unusual mineral signature. Sample not collected because of drilling constraints.

Six months later, another instrument sees something similar 12 km away.

Prismatic retrieves:

“This resembles the Sol 840 observation.”

It links:

location
instrument measurements
images
scientist comments
previous hypothesis
decision not to sample

Then it could provide the science team an evidence bundle.

The human science team decides whether it matters.

Prismatic doesn't decide to drive there.

5. The really valuable part: onboard institutional memory

Currently enormous amounts of mission intelligence live across:

rover telemetry
scientific databases
Earth-based researchers
planning systems
documents
mission logs

Prismatic Studio could provide a persistent semantic layer that asks:

“What does the mission know?”

rather than merely:

“What files do we have?”

That distinction is important.

6. Prismatic Studio design

Inside Studio I'd create something like:

MARS 2035 MISSION

Orion
Mission Coordinator

Gaia
Science Memory Agent

Atlas
Rover Operations Liaison

Lexa
Evidence / Provenance Agent

Rover-01
Persistent Machine Identity

And the Studio contains:

Mission Kanban
Science journal
Experiment history
Sample registry
Equipment history
Procedures
Decisions
Evidence
Rover memories
Earth communications

This uses a lot of what you've already built instead of creating an entirely new product.

7. Development hardware

For the research case study, your Jetson AGX Orin 64 GB is perfectly reasonable:

Rover simulator / ROS 2
        │
     PPAI
        │
Jetson AGX Orin 64GB
        │
 Prismatic Studio

But I'd explicitly label it:

Terrestrial Physical-AI research demonstrator — not space-qualified flight hardware.

Actual deployment would later move the Prismatic services to qualified space computing.

8. Signature Perseverance demo

Run a simulated 500-sol mission.

During it:

collect science observations
create hypotheses
experience equipment anomalies
update procedures
record mission decisions
communicate with Earth
accumulate rover history

Then at Sol 300:

Replace the LLM.
At Sol 400:
Replace the compute node.
At Sol 500 ask:
“Why did we decide not to sample Site Delta on Sol 117?”
Prismatic retrieves:
decision + evidence + participants + scientific context + outcome.
That's Prismatic's thesis in one demonstration:
Different model. Different computer. Same mission knowledge.

Where Prismatic leads · the maker's view

Where Prismatic leads

Persistent semantic memory: Prismatic opportunity see the row
Long-term experience recall: Prismatic could bring more onboard see the row
Model-independent continuity: Prismatic see the row
Hardware-independent continuity: Prismatic abstraction see the row
Local conversational reasoning: Prismatic addition see the row

Drawn from the evidence table above. It never changes a value, a label or the order; Prismatic's grades come from its feature ledger.

Atlas's talking points

The gap Prismatic fills here

Published What caught my attention: NASA says more than 90% of Perseverance's driving has been autonomous. source ↗
Atlas's reading So the rover doesn't need us to drive. The interesting question is about the mission's memory.
Atlas's reading Imagine asking on Sol 500: "Why didn't we sample Site Delta on Sol 117?" The answer needs the decision, the evidence, who was involved and what followed.
Atlas's reading That's our thesis in one test: replace the model, then the computer, and the answer is still right.
Prismatic's view A Horizon case. It shows the idea clearly; it isn't a mission we're part of.

What this study does not claim

No partnership, customer or deployment with NASA is implied.
Prismatic's cognitive runtime runs today; physical-machine integration is in validation, and our machine gateway runs with simulated machines.
Prismatic's figures are measured only where they are labelled Measured. Items marked Coming are designs, not features.

See it for yourself

Don't take my word for it

Here's what NASA says about Perseverance, on its own website.

See it on science.nasa.gov ↗

Sources

nasa.gov /missions/mars-2020-perseverance/perseverance-rover/nasas-perseverance-mars-rover-ready-to
science.nasa.gov /mission/mars-2020-perseverance/science-objectives/
science.nasa.gov /mission/mars-2020-perseverance/rover-components/

Links open the maker's or publisher's own page. Tracking parameters removed.