Iris

Carnegie Mellon University.

Overview

A university that wants to fly a lunar rover cannot buy mass, power or bandwidth in the quantities a national space agency can. Every gram of chassis, every watt of heater power and every bit of downlink has to be justified against a payload that, in Iris’s case, is two cameras. Iris is the Carnegie Mellon answer to that constraint: a 2.2 kg nanorover in a shoebox-sized carbon fiber chassis running a custom single board computer [1], built as a technology and mobility demonstration rather than a science mission, and to do it on a commercial lander’s terms rather than a flagship’s.

It shares its origin with Astrobotic’s CubeRover product line, which packages the same constraint into a CubeSat-like sizing convention: one U is a 10 cm cube of structure supporting roughly a kilogram of payload, with 2U, 4U and 6U variants sharing one structural, power, thermal, avionics and software design [3]. CubeRover’s own numbers, a 4.6 to 10.6 kg rover mass across those sizes and an advertised delivery cost near 4.5 million dollars per kilogram of payload [2], [3], set the design space Iris’s own mass figure sits inside. Iris was the first flight vehicle out of that CMU effort, ahead of the product line it shares a lineage with.

Iris was carried on Astrobotic’s Peregrine lander, launched 8 January 2024 on the first flight of the Vulcan rocket [5]. A propulsion anomaly prevented the landing, and Iris was operated in cislunar space instead, then destroyed with the lander on controlled re-entry on 18 January 2024. What changed after the anomaly was the mission, not the rover: Iris completed the checkout its designers had planned, only against a lander that could never put it on the ground.

Specifications

Parameter	Value	Source
Mass	2.2 kg	[1]
Chassis	carbon fiber monocoque	[1]
Flight computer	single board computer, custom	[1]
Flight software	JPL F Prime	[7]
Sensors	IMU, two cameras, odometry, UWB module	[6]

Carnegie Mellon news material gives 2.2 kg [1], and the 2020 preprint written during development describes a four pound rover [6]. No source reconciles the two. Wheel mass, ground clearance and obstacle height are not published in any source held here and are omitted. For scale, Astrobotic’s own CubeRover line runs 4.6 to 10.6 kg across its 2U to 6U sizes [2], so Iris sits below the smallest of Astrobotic’s own commercial products despite drawing on the same CMU heritage.

Mission profile

These describe Peregrine Mission One, not the rover.

Parameter	Value
Lander	Astrobotic Peregrine
Launch vehicle	United Launch Alliance Vulcan, maiden flight
Launch	8 January 2024, 02:18 ET, Cape Canaveral
Anomaly	oxidizer tank rupture after PCV2 actuation
Mission duration	10 days 14 hours
Peregrine distance traveled	more than 860,000 km (535,000 miles)
End of mission	controlled re-entry 18 January 2024, 16:04 ET, South Pacific

Figures from [5].

Peregrine activated its avionics and power management system at 03:16 ET on 8 January 2024 and established communications through the Deep Space Network [5]. During propulsion commissioning, actuating the oxidizer-side pressure control valve PCV2 let helium flow uncontrollably into the oxidizer tank, which over-pressurized and ruptured, and the tank leaked oxidizer for the rest of the mission. With landing impossible, mission priorities were changed to collecting propulsion data for the investigation, powering and communicating with payloads, and raising subsystem technology readiness levels. Nine of the twenty payloads were designed to communicate with the lander and all nine returned data; the other eleven were passive. On 13 January, in consultation with NASA, Astrobotic decided to end the mission to avoid a debris event in cislunar space [5].

Two-way communication with Iris was established and telemetry returned; the rover survived launch loads and the space environment [1]. Peregrine re-entered with Iris still attached [5].

Mobility

Iris is a wheeled rover carried bolted to the lander deck; because Peregrine never landed, none of its wheels ever turned on the Moon [5]. Published photographs show carbon fiber wheels in a bottlecap form, a flat disc face with a cylindrical rim carrying molded grousers, rather than a spoked or compliant metal wheel. Wheel mass, ground clearance and obstacle performance for Iris itself are not published.

CubeRover’s published mobility figures for the related product line, a 30 degree slope and a nearly 15 cm obstacle capability with independently actuated per-wheel drives and a top speed near 10 cm/s [4], describe a larger vehicle and are not a claim made for Iris itself. Ground testing of that product line, more than 150 traction and slope tests across eleven wheel sets in a 120 ton lunar regolith simulant bin at Kennedy Space Center, found that some wheel sets climbed 30 degree slopes while others turned successfully in deep simulant, and that the same rover crossed simulant dunes and drove into and out of a dug trench [9], [10]; those are per-wheel-set best cases in Earth gravity, not a demonstrated product capability, and nothing comparable has been run on Iris’s own wheels.

Power and energy

Custom power management electronics were flown, but power source, generation and battery capacity for Iris are not published. Iris was a short-duration surface mission and no night survival capability is claimed.

Thermal

The chassis carries multi-layer insulation over the carbon fiber monocoque [1]; no temperature limits or heater powers for Iris itself are published. Carnegie Mellon reports that the rover endured extreme temperatures and high radiation levels during the ten day flight with primary systems performing normally [1]. The sister CMU nanorover MoonRanger, built on the same low size, weight and power philosophy but targeting a permanently shadowed lunar polar site, was modeled to need at least 9.5 W of lander-supplied heater power just to survive dark cislunar transit and to hold a surface operating band of about -10 to 35 C [8]; no equivalent figures exist for Iris, but the comparison shows what a lunar nanorover’s thermal margins look like when they are published.

Compute and avionics

A single board computer, custom built, is the flight computer; power management is also custom. Flight software is an implementation of JPL’s F Prime framework, an open source component and port architecture with active, passive and queued component kinds, that by 2018 had already flown on the ISS RapidScat instrument and the ASTERIA CubeSat and was in development for Lunar Flashlight, NEA Scout and the Mars Helicopter [7]. Choosing an existing flight-proven framework rather than writing one is itself a low-resource-team decision. Processor part, memory and radiation tolerance approach for Iris are not published.

Autonomy

Iris was designed to be driven from the ground. The published architecture puts localization off the vehicle: onboard localization relative to the lander uses an ultra-wideband module, and the SLAM and visual-inertial odometry computation runs on the ground rather than onboard, because using RF for both communication and state estimation saves mass and power the rover does not have [6]. The onboard sensor suite is an IMU, two cameras and odometry sensors, plus the UWB module. This is the opposite choice from the onboard navigation JPL has flown on its own deep-space missions, where a dedicated processor computes position and guidance onboard because round-trip communication delay makes real-time ground control impossible [12]; a nanorover’s mass and power budget instead pushes that computation to the ground whenever a short-range link to a lander allows it.

The ground-side perception pipeline proposed for the mission is object-level SLAM that optimizes the pose and shape of the lander jointly with the rover camera trajectory, using the known lander location as a prior, with a semantic segmentation front end fine-tuned from DilatedResnet-101 and UperNet-101 trained on ADE20K to produce the lander mask [6]. A video frame interpolation stage synthesizes intermediate images between camera samples, using a depth-aware flow projection layer, to compensate for the low frame rate the link allows. This pipeline was proposed during development; no result shows it running on flight data, since Iris never had the chance to be deployed independently of the lander.

Communications

Iris has no direct-to-Earth link. It communicates through the lander, and the same ultra-wideband link that localizes the rover relative to the lander is the data path, with the lander acting as the way station [6]. The related CubeRover product line instead budgets a wired RS-422 and 802.11n wireless interface at 10 kbps per kilogram of rover mass for its lander link [2], a different point in the same design space of getting data off a mass-constrained rover through its host lander rather than direct to Earth.

Payload and instruments

Iris carries no science instruments. The two cameras, plus the IMU, odometry sensors and UWB module, are the payload [6]; no stated science objective for the imagery is published.

Modes of operation

No mode set beyond stowed, deployed and driving is published, and none beyond stowed was exercised, since Peregrine never landed [5].

Ground operations

Operated by a student mission control team at Carnegie Mellon [1]. Three hundred students across seven Carnegie Mellon colleges worked on Iris, with about thirty in Florida for launch. When the landing became impossible, the team continued operating the rover from a rented house near the launch site for the remainder of the ten day flight. The scale of that effort is not unusual for a CMU flight nanorover: the university’s other lunar nanorover project, MoonRanger, reports more than 150 students, staff and faculty across its life, and a mid-program migration of its flight software from ROS to NASA’s core Flight System driven by an avionics change [11], the kind of program-level churn a student-staffed effort has to absorb that a professional flight program budgets around from the start.

Technologies developed

The result that outlived the vehicle is the carbon fiber structure: a chassis and wheels built from carbon fiber that survived launch and the space environment [1]. It also fed directly into Astrobotic’s CubeRover product line, which now sells the same low size, weight, power and cost approach as a commercial mobility service on later CLPS landers [2], [3], [4].

Context: nanorovers and small lunar rovers

The nanorover idea predates Iris by decades. JPL’s 1997 technology survey listed nanorovers beside long-range rovers, sample retrieval and aerobots as a thrust area [13]. A 1998 paper described the MUSES-C nanorover prototype alongside the Sojourner performance data and the shared planetary and urban military miniature robot work [14]. That prototype had four independently actuated wheels on independently posable struts plus a pitchable body, and a 1999 position paper proposed discovering its self-righting behaviors by genetic programming against a dynamics simulator rather than designing them by hand [15]. These were ground prototypes and proposals, not flown vehicles.

Of the small rovers aimed at the Moon since, Iris is the one that left Earth and was operated. MoonRanger, the other CMU nanorover, passed NASA’s Key Decision Point review with Astrobotic as prime, CMU as subcontractor and NASA Ames as partner, carrying a neutron spectrometer to search for water [16]. Its project logs report a custom Wi-Fi board that transmitted and received at 160 m with a 30 percent drop from peak short-range performance [17], and a power switching board with twenty switching groups built on automotive grade smart switches [18]. Lunar Outpost’s MAPP rover on Intuitive Machines IM-2 shared Iris’s outcome in one respect: the lander came to rest on its side in a crater, MAPP was never deployed, and it was operated in transit and on the surface from its stowed position [19]. The Canadian Space Agency reports a Canadian Lunar Rover and a much larger Lunar Utility Vehicle as its two rover programs [20]. JPL’s TULIPS tethered power system proposes 1 kW at 1.5 kV over a 1 km tether with fiber Ethernet for CLPS-class rovers, a route to power and bandwidth that a battery-only nanorover does not have; it has been demonstrated on Earth, not on the Moon [21].

On flight software, F Prime on Ingenuity ran as three separate deployments on three processors joined by a hub pattern, which shows the framework Iris used scaling beyond a single computer [22].

Context: wheels, soil and thermal environment

Iris’s wheels never touched regolith, so the relevant evidence comes from other programs. The ISRU Pilot Excavator wheel trade ran ten cleat and grouser configurations through 83 tests on RASSOR 2.0 in the Kennedy Space Center regolith test bed, measuring slip against motion capture [23]. The IPEx TRL-5 design record uses one common actuator design for wheel, arm and bucket drum roles [24], and its predecessor RASSOR 2.0 is a 66 kg vehicle whose counter-rotating drums cancel horizontal reaction force [25]. These are excavators on Earth, tested in simulant, and bound Iris’s wheels only as a reference for method.

Lunar Terrain Vehicle wheel thermal modeling at the south pole used a 140 km elevation model and a 150 hour eclipse node to set wheel temperatures [26]. VIPER was to measure wheel normal force and sinkage directly and invert them to soil properties [27]; the rover was not flown. The Lunar Sourcebook gives the geotechnical, electrical and thermal surface properties a lunar machine is designed against [28], and the 2023 to 2032 planetary decadal survey sets the technology priorities, including autonomy and extreme environments, that small commercial landers now serve [29].

What is not established

Iris’s mass is reported as 2.2 kg by Carnegie Mellon and as four pounds (about 1.8 kg) in the 2020 development preprint, and no source reconciles the two [1], [6]. Wheel mass, ground clearance and obstacle performance for Iris itself are not published. Power source, generation and battery capacity are unpublished, as are processor part, memory and radiation tolerance approach for the flight computer. No thermal limits or heater power figures for Iris itself have been published; the MoonRanger figures given for comparison describe a different vehicle on a different mission profile [8] and do not stand in for Iris’s own numbers. A deployment concept study and a mission control practices retrospective for Iris are listed only as titles on the project’s own publications page, with no fulltext available to check what they say [30]. Because Peregrine never landed, nothing about Iris’s actual surface mobility, wheel-terrain interaction or ground autonomy performance was ever demonstrated, only what was validated aboard the lander and in ground testing beforehand.