Spirit

NASA/JPL-Caltech. Public domain (NASA / US government work).

Overview

Spirit, formally MER-A, landed in Gusev crater on 4 January 2004 on a mission planned for 90 sols and operated for more than six years [3]. It drove 7.73 km, returned over 124,000 images, ground the surfaces off 15 rock targets and brushed 92 more [3]. It last communicated on 22 March 2010.

Two mobility failures define its engineering history and separate it from its twin: loss of the right-front drive motor in 2006, after which the rover drove as a five-wheel vehicle dragging a dead wheel, and the embedding at Troy in 2009, which ended mobility [3], [4].

Specifications

Parameter	Value	Source
Configuration	six driven wheels, four steered, rocker-bogie	[1]
Wheelbase length / width	about 1.4 m / 1.2 m
Width across solar panel	1.8 m
Height with Pancam mast deployed	just over 1.5 m
Ground clearance	at least 0.3 m
Delivered mass	185 kg	[1], [10]

Mission profile

Parameter	Value	Source
Launch	10 June 2003, 10:58:47 PDT	[10]
Landing	3 January 2004, Gusev crater, Columbia Memorial Station	[3], [10]
Landing point	14.5692 S, 175.4729 E	[10]
Landing dispersion	about 13.4 km from the planned target
Planned mission	90 sols	[3]
Operating life	six years, two months and 19 days	[10]
Total traverse	7.73 km	[3], [10]
Last communication	22 March 2010
Mission declared complete	25 May 2011	[10]
Probable cause of loss	cold beyond what the survival heaters could hold

The mission page adds the landing coordinates, the 13.4 km landing dispersion inside Gusev, the 25 May 2011 date on which NASA declared the mission complete more than a year after the last contact, and the attribution of the loss to cold defeating the survival heaters [10]. None of these is in the retrieved literature.

Mobility

The right-front wheel became immobile in 2006 [3]. The rover thereafter drove predominantly backwards, dragging the failed wheel, and covered nearly another kilometer in that configuration. A dragged locked wheel is a continuous shallow trenching tool, and in 2007 the furrow it cut exposed bright white soil of nearly pure silica, which implies hydrothermal alteration at Gusev [3]. No planned trenching campaign had targeted that location.

Soil-covered slopes of more than a few degrees produced high slip, progressive sinkage and exposure of subsurface soil throughout the mission. On 23 April 2009, driving south beside Home Plate, the rover broke through a thin crusted surface into soft material beneath and embedded [4]. The right-rear wheel stopped working in November 2009, leaving four functioning drive actuators. Extraction attempts continued until NASA designated Spirit a stationary research platform on 26 January 2010 [4].

Terramechanics of the embedding

A driven wheel on granular soil develops thrust from shear at the wheel-soil interface. The shear stress the soil can sustain is bounded by its cohesion and internal friction angle acting on the normal stress, and the normal stress is set by the wheel load spread over the contact patch. The standard formulation couples the Bekker pressure-sinkage relation, in which contact pressure is a power law in sinkage through the moduli k_c and k_phi and a sinkage exponent n, to the Janosi-Hanamoto shear-displacement relation, from which wheel torque, thrust and sinkage follow [7], [8]. Slip ratio is defined from the difference between commanded wheel surface velocity and measured vehicle velocity, and runs from 0 to 1 in driving.

The failure is regenerative rather than a single threshold crossing. When available shear falls below resistance the wheel rotates without advancing, displacing material rearward and downward. That enlarges sinkage, which enlarges both contact patch and the bulldozing resistance ahead of the wheel, which further reduces net thrust [7], [8]. On slopes the coupling is worse because longitudinal and lateral slip both grow, and immobilizing slip is the specific hazard mobility prediction on deformable slopes exists to avoid.

Bekker-Wong semi-empirical modeling remains the flight-planning workhorse, but its accuracy against measured wheel forces in dry granular media is bounded, and resistive force theory and continuum plasticity by material point method reproduce forced-slip experiments across three wheel types and three granular materials to better accuracy over the same input range [7]. Resistive force theory is cheap enough to evaluate inside a planning loop, which is the property a tactical cycle needs.

Three site-specific factors made Troy severe. The surface was a thin indurated crust over low-cohesion material, so the failure occurred below the surface that had been imaged and assessed. The underlying material was a sulfate-rich sand, fine-grained and weakly cohesive. The rover was already down one drive actuator, so the dragged right-front wheel contributed resistance and no thrust, biasing the vehicle toward the runaway above [4]. Cohesion of Martian sand is at or below 1 kPa and its friction angle near 30 degrees, against 4 to 25 kPa for cemented duricrust, so a crust over sand presents a strength contrast of an order of magnitude across a few centimeters of depth.

The extraction campaign ran against physical and numerical models rather than direct attempts. At JPL a test rover was embedded in a prepared soil bed by spinning each wheel individually so that it sank without translating, reproducing the flight vehicle’s configuration, and candidate escape maneuvers were driven in the test bed before uplink [4]. The Rover Sequencing and Visualization Program then used to plan MER drives is purely kinematic and cannot represent soil response, so the physical test bed carried the burden of predicting how the wheels would behave in deformable material.

Power and energy

Spirit is solar powered [5] from triple-junction GaInP/GaAs/Ge arrays, with rechargeable batteries carrying the vehicle overnight [6]. Energy rather than time bounds surface activity, and the bound tightens in winter at Gusev’s southern latitude. The telecommunications record shows the effect directly: during the winter of the primary and first extended mission, with battery state of charge low, all telemetry transmission except the afternoon Odyssey relay pass was canceled [1]. Morning passes were given up first, because waking the rover in the coldest part of the sol costs disproportionate heater energy.

Array output is degraded by two separate dust processes. Suspended atmospheric dust modifies the solar spectrum and reduces intensity as a function of optical depth and time of day, and deposited dust obscures the cells cumulatively [6]. Array performance is the primary operational constraint on allowed landing site latitude and on the power available once landed. Wind events that clean the panels reset the budget, and a series of such events coincided with the onset of summer while the rover sat embedded at Troy.

After the embedding, surviving mobility was used solely to adjust vehicle tilt toward the winter Sun, because array tilt determines whether a stationary rover survives the season [4].

Thermal

Parameter	Value	Source
Enclosure	warm electronics box: composite honeycomb lined with aerogel, closed by the equipment deck	[1]
Aerogel composition	99.8 percent air
Surface peak daytime temperature	-15 C	[6]
Surface nighttime minimum	-127 C
Transponder and SSPA allowable flight temperature	50 C (protoflight 60 C)	[1]
UHF transceiver allowable flight temperature	55 C (protoflight 70 C)
X-band SSPA dissipation	16.8 W RF out of 58 W DC in
UHF transceiver input power	1.9 W

The box holds the rover warm overnight with no heaters running [1]. The binding thermal constraint is on the hot side. During the day the X-band and UHF transmitters run three or four times in succession, and because the heat-producing elements sit close together on the rover electronics module, internal temperatures climb toward those limits. The cruise-stage heat rejection loop was severed at Mars arrival by design, so surface thermal control is passive plus heaters.

Cold carries an operational cost beyond survival. As winter wake-up temperatures fell below -20 C the transponder’s voltage-controlled oscillator drifted enough that uplink acquisition strategies changed: sweep rate reduced from 200 to 100 Hz/s, received power at the spacecraft capped at -130 dBm, and the 70 m stations asked to transmit at 10 kW rather than 20 kW [1]. On Spirit, temperature-dependent coherent leakage could move the oscillator frequency by as much as 15 kHz in 10 minutes, and by the time wake-up temperatures approached -30 C an uplink frequency reference offset of up to -6 kHz was needed for morning acquisitions.

Compute and avionics

Element	Specification	Source
Processor	RAD6000, 20 MHz	[9]
Bus to transponder	MIL-STD-1553, low power	[1]
Power distribution	rover power distribution unit
Inertial measurement unit	Litton LN-200, three-axis accelerometers and rate sensors
Wheel odometry and IMU position update rate	8 Hz	[9]
Visual odometry image size	256 x 256 NavCam stereo pairs
Visual odometry computation time per step	nearly 3 minutes

The 20 MHz processor and low-throughput camera bus are what set the visual odometry step time, and that step time is why visual odometry could be run on only 52 of Spirit’s first 414 sols rather than continuously [9].

The most disruptive avionics event occurred on sol 18, when a fault in the flash memory file system left the rover cycling and unable to complete a boot [1]. Diagnosis was possible only because a high-priority communication window overrode the 10 bps default downlink mode and supported 40 or 300 bps through the low-gain antenna, which returned the repeating event reports that identified flash as the source. An unrelated geometric problem compounded it: the Pancam mast assembly occluded the high-gain antenna from local noon onward, degrading most sessions by as much as 8 dB and once by 14 dB, and with the instrument deployment device deployed the rover attitude could not be changed to clear the mast from the antenna’s field of view [1]. Occlusion severity for the following sol had to be estimated empirically from the current sol’s data because no model predicted it.

Autonomy

Spirit drives under a mix of directed sequences and onboard autonomous navigation with hazard detection from stereo imagery, and uses visual odometry to measure actual against commanded motion.

Visual odometry performance to March 2005	Spirit	Opportunity
Convergence rate	97 percent (590/609)	95 percent (828/875)
Mean features tracked	73.4 ± 29.3	87.4 ± 34.1
Mean iterations to convergence	6.4 ± 1.7	8.4 ± 5.2
Mean rover tilt while running	14.6 ± 4.4 deg	18.0 ± 4.6 deg
Tilt range	2 to 30 deg	0.8 to 31 deg
Sols used	52 of 414	75 of 394
Distance driven	4161 m over 184 driving sols	3158 m over 172 driving sols

Performance figures from [9].

Slip ratios as high as 125 percent were measured successfully, on sol 206 while Spirit drove up a slope steeper than 25 degrees, and changes as small as 2 mm were resolved [9]. Absolute position error on a 24 m Marsyard course was under 2.5 percent, with under 2 mm position and 0.2 degree attitude precision per step. Slip above 50 percent occurred on sand, in one case with 50 m of commanded wheel rotation producing 2 m of ground progress [9].

Slip limits are the fault protection that stops a drive before sinkage becomes regenerative, and visual odometry is what makes slip observable to the vehicle [9], [8]. Troy defines the limit of that protection: it measures slip at the surface the rover is on and cannot anticipate a weak layer beneath an unbroken crust [4].

Opportunity’s embedding at Purgatory ripple is the counterexample that shows the protection working in reverse. After 50 m of commanded blind driving the rover embedded on sol 446 and did not clear the ripple until sol 484 [9]. During the first eight sols of extraction, commanded drives of 2 to 20 m per sol produced 1 to 7 cm of actual progress, and visual odometry measured that progress to 1.3 ± 0.5 mm, which is what allowed each attempt to be evaluated before the next was designed.

Communications

The X-band subsystem is a small deep space transponder with a solid-state power amplifier mounted inside the warm electronics box, feeding either a steerable flat-panel phased array high-gain antenna or a near-omnidirectional rover low-gain antenna on the equipment deck [1]. Bulk return is by UHF relay from a 19 cm monopole on the equipment deck through a CMC Electronics transceiver designed for compatibility with the matching unit on Mars Odyssey.

Link	Parameter	Value
X-band LGA	Minimum command rate	7.8125 bps [1]
X-band LGA	Minimum telemetry rate	10 bps [1]
X-band HGA	Best uplink / downlink at shortest range	2 kbps / 28.8 kbps [1]
UHF	Supported rates either direction	8, 32, 128, 256 kbps [1]
UHF	Forward link, as flown	fixed at 8 kbps [1]
UHF	Return link, per pass selection	128 or 256 kbps [1]
X-band SSPA	RF out / DC in	16.8 W / 58 W [1]
UHF transceiver	Input power	1.9 W [1]

Return rates of 128 or 256 kbps were selected per pass according to which returned more data, and 256 kbps came into regular use during the extended missions [1]. Odyssey uses the CCSDS Proximity-1 protocol with Go-Back-N retransmission and achieves about 97 percent throughput at high signal-to-noise ratio despite up to 15 loss-of-lock gaps per pass. Mars Global Surveyor used the older Mars Balloon Relay protocol, which transmits for only 13.3 to 13.8 s out of every 16 s and lost roughly two rover frames every 16 s at 128 kbps, so only low-priority data was routed through it; MGS relay was abandoned for Spirit after sol 137 [1].

Passes above 20 degrees elevation average about 1.8 per sol per orbiter, lasting two to eight minutes, with a planned average return of about 56 Mbit per sol through Odyssey and 49 Mbit per sol through MGS [1]. To September 2005, about 92 percent of returned data went through Odyssey, 5 percent through MGS and 3 percent over the X-band direct link. Over the same interval the rover had driven 4161 m across 184 driving sols, so relay volume rather than drive distance is what bounded the science return per sol [9].

Rover orientation is a first-order term in relay performance because the UHF antenna pattern is strongly asymmetric. The same pass can return 50 Mbit at a favorable azimuth and half that if the high-gain antenna assembly blocks the line of sight, and at tilts near 20 degrees the rover deck itself can occlude the orbiter for most of a pass [1]. Mean rover tilt while running visual odometry was 14.6 degrees, with excursions to 30 degrees, so pose-dependent relay loss and pose-dependent slip are simultaneous constraints on the same drive [9].

Payload and instruments

The Athena payload comprises the panoramic camera and miniature thermal emission spectrometer on the mast, and on the instrument deployment device a Moessbauer spectrometer, an alpha particle X-ray spectrometer, a microscopic imager, and the rock abrasion tool that grinds through weathered rind to expose fresh rock [5]. Navigation and hazard cameras support driving; the NavCams have a 45 degree field of view and supply the 256 x 256 stereo pairs used for visual odometry [9]. Through sol 2210 the mission had ground the surfaces off 15 rock targets and brushed 92 more [3]. The five degree of freedom arm can place all three in-situ instruments and the abrasion tool on the same spot on a rock or soil target.

Modes of operation

Surface activity is organized around the sol and around communication windows. Most sols included a five-minute carrier-only downlink, called a beep when transmitted stationary and a honk when transmitted while driving, used to confirm that the morning master sequence had uploaded and started when a full direct-to-Earth session was not affordable in power, thermal or activity terms [1]. The master sequence is uplinked each morning blind, without downlink confirmation that the command sweep succeeded, and the timing of the returned beep is what tells the ground the sequence took. Beeps are detected first by open-loop fast Fourier transform and are reliably detectable down to 12 dB-Hz [1].

Communication behavior is governed by parameterized windows, including high-priority windows that override the default low-rate downlink mode during anomalies, as used on sol 18 [1]. Driving and imaging cannot be scheduled during a UHF pass, because pre-launch testing showed those activities can interfere with the UHF receiver. Visual odometry is itself a schedulable mode rather than a continuous one, since each step costs nearly three minutes of processor time [9].

A deep sleep mode was developed during the first Martian winter to eliminate overnight battery drain, at the cost of preventing the rover from waking for early morning relay passes [1]. It was introduced on Opportunity in July 2004 to counter a stuck instrument deployment device heater, and Spirit’s own winter power constraints led to cancellation of most of its morning passes on both orbiters.

Ground operations

Spirit was operated from JPL. Relay planning runs on two timescales. Orbiter passes are requested and sequenced onto Odyssey and MGS a week or two ahead, with unrequested passes left in a default configuration [1]. Tactical planning then works sol by sol over the next pass or two, optimizing UHF data return against rover attitude, available energy and the time the data is needed on the ground.

Vehicle pose is treated as a communications parameter. Data volume capability forecasts identify desirable parking yaw angles, which are handed to mobility planning as a constraint on where and how the drive ends [1]. On occasions the rover was commanded to turn between two afternoon overflights purely to place the next orbiter in a high-gain region of the antenna pattern. Where tilt was large enough to invalidate the zero-tilt forecast, the generalized telecom predictor was rerun with the estimated attitude, and mean tilt while driving on the Columbia Hills ran 14.6 degrees with excursions to 30 [9]. Coordination with the orbiter teams ran over a standing tactical email list, with pass keep and delete decisions confirmed after the daily command load was received.

Drive planning is the other half of the same cycle. Because a visual odometry step costs nearly three minutes of processor time, planners decide per drive segment whether to spend that time or drive blind on wheel odometry, and the Purgatory embedding followed 50 m of commanded blind driving [9]. Extrication planning after Troy added a third element: candidate maneuvers were driven in the JPL soil test bed before uplink [4].

Technologies developed

The extraction campaign coupled planetary rover operations to terramechanics: recovery maneuvers were developed against a physical deformable-soil model before being commanded [4]. The approach carries directly to lunar vehicles on low-cohesion regolith, where the Bekker pressure-sinkage and Janosi-Hanamoto shear formulation applies [7].

The mission established the physical test bed with a mass-matched rover in prepared soil as a standard tool for planetary mobility anomalies, including embedding the test vehicle by spinning wheels individually so the initial condition matches the flight vehicle rather than resembling it [4].

Flight visual odometry on the MER vehicles produced the first large statistical record of slip on another planet: 97 percent convergence over 609 stereo pairs on Spirit, slip measured up to 125 percent, and displacement resolved to 2 mm [9]. That record is the calibration data against which subsequent slip-prediction and wheel-soil models are tested [7], [8].

After mobility was lost, the stationary vehicle was assigned measurements a moving one cannot make, including long-baseline radio tracking of the planet’s rotational wobble to constrain whether the core of Mars is liquid or solid [4].