Phoenix Robotic Arm

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

Overview

The Phoenix Robotic Arm was a four-degree-of-freedom backhoe-configuration manipulator built by NASA JPL with Alliance Spacesystems as arm subcontractor and Honeybee Robotics supplying the rasp, flown on the Phoenix Mars Lander to the north polar region of Mars [1]. Its function was to excavate trenches, acquire dry and icy soil, and deliver samples to the Thermal and Evolved Gas Analyzer and to the Microscopy, Electrochemistry and Conductivity Analyzer, and to place the Thermal and Electrical Conductivity Probe and the Robotic Arm Camera. The arm operated for over 150 sols after landing on 25 May 2008 [1]. It was the element that made the mission’s central measurement possible, since both analytical instruments depended on it for material [6].

The design objective that separates this arm from earlier Mars lander arms is hard icy soil. Loose regolith at the site was only a few centimeters deep over ice-cemented material that a scoop blade could not cut, so a powered rasp was carried as the primary icy-soil acquisition tool [1]. The arm could reach far enough to dig about half a meter deep, deeper than the ice table was expected to lie [3].

Specifications

Parameter	Value	Source
Degrees of freedom	4: shoulder azimuth, shoulder elevation, elbow, wrist pitch	[1]
Length	2.35 m	[3]
Link material	aluminum links, titanium end fittings	[1]
Joint actuators	brushed DC motors, planetary gears plus harmonic drive; wrist uses a bevel gear at the planetary output
Motor type	Maxon RE020 brush DC, copper/graphite/MoS2 brushes to a JPL formulation	[2]
Braking	motor leads actively shorted, rotor captured by magnetic detents	[1]
Position sensing	quadrature encoders at the motor shaft, potentiometers at the joint output
End effector	scoop with primary and secondary blades, powered rasp, TECP, RAC
Secondary blade	tungsten carbide, on the underside of the scoop, for scraping indurated material
Rasp drive	brushed DC motor, spur gears plus miter gear pair, 1.25:1 reduction motor to bit
Design environment	-90 to -20 C diurnal, CO2 atmosphere, pressure to 5 Torr
Planetary protection	Category IV-B bioburden, sterilized and enclosed in a bio-barrier

Mission profile

Parameter	Value	Source
Landing	25 May 2008, 68.16 N, 233.35 E	[3]
Primary mission	90 sols	[1]
Major arm operations	through sol 150
Last contact	sol 157	[3]
Dig sequences executed	53, in 10 areas	[1]
Samples acquired and delivered	18: 6 to TEGA, 4 to MECA wet chemistry, 8 to MECA optical microscope
Icy-soil samples delivered to TEGA	1
Deepest trench	Stone Soup, about 18 cm at the end nearest the lander
Snow White trench	about 25 cm wide by 60 cm long, 8 cm maximum depth, 7 digs and 12 scrapes
Icy sample yield by rasping	about 2.5 cc from a 4 by 4 grid of 16 rasp holes
Rock push	0.5 m guarded move with the scoop front, sol 117

Indurated material was encountered between about 3 and 10 cm below the surface across almost the whole workspace, identified in operations by increased joint torques, engagement of the surface accommodation algorithm and motion-impeded events, and sloping deeper towards the lander [1]. The ice table reached in the Snow White trench lay about 5 cm below the surface. The site’s habitability case rests on results the arm enabled: segregated light-colored ice in the surface materials, calcium carbonate at 4 to 5 percent by weight in the dry soil above the ice, and perchlorate, presumably magnesium perchlorate, at about 0.5 percent by weight [4].

Mechanism

The arm is a backhoe layout with two degrees of freedom at the shoulder, azimuth and elevation, and one each at elbow and wrist pitch [1]. Each joint actuator is a brushed DC motor with multi-stage speed reduction through planetary gears and a harmonic drive; the wrist substitutes a bevel gear at the planetary output. There are no brakes: the motor leads are actively shorted to slow the rotor until magnetic detents capture it, and detent holding torque is sufficient to prevent slippage with power off. Each joint carries a heater and a temperature sensor to keep the motor at or above its minimum operating temperature.

Joint position is measured by quadrature encoders on the motor shaft and potentiometers at the joint output. Encoder counters are initialized either from potentiometer data or by driving each joint against a mechanical hardstop at the end of its travel, and counts are written to flash memory at the end of each sol for use in the next sol’s initialization [1]. The scoop is divided by an internal partition into a front chamber, which collects material excavated by the front blade, and a rear chamber, which houses the rasp and collects rasp cuttings. Material is transferred from rear to front by rotating the whole scoop about the wrist axis. Scoop and funnel are aluminum. A tungsten carbide secondary blade on the underside penetrates harder material and is used mainly for scraping indurated surfaces. A machined depression in the scoop blade holds small samples for close-up imaging by the Robotic Arm Camera before delivery. The camera is mounted on the arm immediately above the scoop [3].

Rasp and icy-soil acquisition

The rasp is the primary icy-soil tool: a high-speed cutting bit on the back of the scoop, driven by a brushed DC motor through spur gears and a miter gear pair with an overall reduction of 1.25:1 from motor to bit [1]. The bit is mounted in a pivoting, spring-loaded housing so that it retracts into the scoop when the scoop is preloaded against the surface, then protrudes under spring force during cutting, which forces the bit into the material. A knurled contact plate grips the icy soil to stop the scoop moving during rasping. A current-limiting board mounted externally on the arm electronics limits rasp motor current. The rasp mechanism carries resistive strip heaters and a motor temperature sensor, and its powered components are thermally isolated from the scoop body. The rasp doubles as a vibrator. Driving the pivoting housing so that its spring-loaded body strikes its lower hard stop once per revolution produces high-frequency vibration used to move cuttings from the rear chamber to the front.

Acquisition of an icy sample took a sequence rather than a single motion: repeated scraping to flatten the trench floor for rasp load-plate placement, a light scrape immediately before rasping to clear loose material, sixteen rasps in a 4 by 4 grid, transfer of cuttings from rear to front chamber, and careful scooping of the surface ejecta, yielding about 2.5 cc [1]. Acquisitions were run early in the morning to keep the scoop as cold as possible and limit sublimation of ice in the sample.

Three of the first four icy-sample attempts failed for reasons specific to icy material. On the first two the sample congealed in the scoop during delivery, attributed to direct solar illumination on the sample making it stick to the scoop during the dump [1]. The third succeeded by scraping rather than rasping, which produced a lower-ice sample that did not congeal, and TEGA confirmed water ice in it. A fourth was acquired by rasping with the scoop motions sequenced to minimize solar energy on the sample, but delivery failed because the target TEGA cell doors did not open fully. Phoenix confirmed the presence of ice-cemented ground at the site [4].

The failure mechanism is thermal, not mechanical, and it acts between acquisition and delivery rather than during either. Insolation on ice-bearing regolith held in an aluminum scoop warms the sample enough that it adheres to the scoop wall and will not fall out when the scoop is inverted over the inlet [1]. Neither the rasp nor the scoop is at fault: the material changes state in the interval between them. The corrections applied on Mars were operational, moving acquisition to the coldest part of the sol so that the scoop starts cold, and sequencing scoop motions during transfer to minimize solar energy on the sample. A sample containing less ice, obtained by scraping rather than rasping, did not congeal and was delivered successfully. This is a ground-test fidelity gap rather than a design defect: reproducing it requires a chamber that holds Mars pressure, Mars-relevant sample temperature and a representative solar flux on the tool simultaneously, through a transfer sequence of realistic duration.

Thermal

The arm was designed for a landing site with diurnal excursions from -90 to -20 C, a CO2 atmosphere and pressure as low as 5 Torr [1]. Every joint carries a heater and a temperature sensor so that the motor is operated only at or above its minimum operating temperature. The actuator temperature limits and the warm-up heater sizing were published only as conference slides, so they are not stated here.

Compute and avionics

Arm flight software ran on the lander Command and Data Handling computer rather than on a dedicated arm processor [1]. A payload electronics board provided power conditioning, motor voltage control and drivers, heater drivers, joint encoder counting, and digitization of potentiometer and temperature sensor voltages, motor currents and total heater current. Firmware on that board executed low-level motor commands to move joints to commanded positions and controlled heaters and the rasp. Processor part and memory size are not published.

The arm motors are Maxon RE020 brush DC units with copper/graphite/MoS2 composite brushes manufactured to a JPL formulation and supplied to the motor vendor as government-furnished equipment, the same motor and brush lineage used on the Mars Exploration Rovers [2]. Brush tip and pivot breakages found in Phoenix qualification life testing at -80 C triggered an independent NASA Engineering and Safety Center review. The review found that all environmental loads except pyroshock produce negligible loads on the brushes, that even the weakest brush tested had a quasi-static capability of 945 g against a 5.5 g rms random acceptance level and a 58 g landing qualification load, and that relative brush movement inside the motor requires peak accelerations above 200 g radial or 75 g axial, beyond which internal impacts can generate loads well above the external excitation [2]. A broken brush was assessed as still allowing motor function with possible reduction in torque margin or life. Molybdenum disulfide in the brush material is sensitive to humidity at or above 40 percent relative humidity, and chemical conversion of MoS2 raises the friction coefficient, which made storage conditions of the brush stock part of the review [2].

The choice of brush DC motors over brushless or stepper types traces to the same JPL line of Mars actuator development. Brush motors were selected for the Sojourner rover and for the four-degree-of-freedom Mars Polar Lander arm because they need no commutation electronics and give a high power to mass ratio, at the cost of brush wear as the life limit [5]. Qualification of that lineage established the failure modes that matter in a 1.0 kPa CO2 atmosphere at -80 to -10 C: rotor shorting from conductive debris in the commutator slots, galling or micro-welding of brush to commutator that bends or breaks the brushes, and commutator grease bonding brush to commutator below -60 C. Testing of 29 Maxon RE016 units across air, vacuum, low-pressure CO2 and low-pressure N2 at -70 to -80 C returned average lifetimes under 10 million revolutions against 30 to 40 million in the earlier Sojourner campaign, which forced a change of motor and of brush material to the silver-graphite compound qualified on Cassini. High-load mechanism life then reached 102 million revolutions at -70 C in 1.0 kPa CO2 against 36 million in 1.0 kPa N2 and 62 million in ambient air [5].

Autonomy and control

Arm software expands single high-level commands, such as dig a trench or acquire a sample, into the sequence of motion commands needed to execute them, which saves uplink bandwidth and removes sequence construction from the operator [1]. The software tracks time and energy consumed during execution and terminates the activity when allocations are exceeded, a feature that mattered most in trenching, where soil properties made duration unpredictable. Motion commands cover coordinated joint motion and Cartesian motion of the end effector. Joint moves are absolute or relative; Cartesian moves are specified in the payload frame at the arm base, in the local frame of the currently selected tool, which may be the scoop, the scraping blade, the TECP, the rasp load plate or the RAC. The four Cartesian degrees of freedom are three translations plus the angle between the selected tool approach vector and the plane of the lander deck, except for the RAC, whose orientation cannot be commanded independently. Each command is decomposed into via points sent sequentially to the arm electronics for firmware execution. Contact is handled by guarded moves: the arm drives toward the commanded position while motor currents and computed joint torques are compared against preset thresholds, stopping on contact. Guarded moves were used for TECP insertion, preloading the rasp load plate on hard ice, sample acquisition and trenching, which made operation tolerant of uncertainty in surface location.

Faults, defined as an inability to complete a command caused by hardware failure, are distinguished from events, an inability to complete a command from any other cause such as a rock in the digging path [1]. Both are reported in telemetry, and the software either attempts recovery or moves the arm to a safe configuration. A surface accommodation algorithm adjusts the scoop trajectory when motion is impeded during digging and, if necessary, dumps the scoop contents and re-executes the dig. Because indurated material impeded the arm frequently, autonomous recovery from these events was enabled in flight software and in command sequences, which let subsequent operations continue without ground intervention and saved sols that would otherwise have gone to recovery.

Communications

Phoenix was a UHF-relay-only lander with no surface direct-to-Earth link, relayed by Mars Odyssey and the Mars Reconnaissance Orbiter with Mars Express held as backup [3]. The requirement was 1 Mb per sol forward in a single overflight between 02:00 and 09:00 local solar time and 60 Mb per sol return, with 30 Mb in a single overflight between 12:00 and 19:00. Actual performance over the 157-sol landed mission was 860 relay passes, more than five per sol, of which Odyssey provided 71 percent, and more than 38 Gb returned, an average of 242 Mb per sol [3]. Forward-link traffic totaled 574 Mb, 97 percent through Odyssey, and Odyssey morning passes were raised to a 32 kbit/s forward rate, giving about 12 Mb of forward capability [3]. Use of MRO was capped at two passes per sol because MRO was still in its primary science mission.

Ground operations

Arm operations were planned with the Rover Sequencing and Visualization Program, carried over from Pathfinder and the Mars Exploration Rovers and adapted for Phoenix [1]. RSVP provided target designation from panorama imagery, macro-driven subsequence generation, arm motion simulation, collision checking, command duration estimation and output of a complete sequence file for uplink.

Every terrain-interacting activity was preceded by imaging from the Surface Stereo Imager and the Robotic Arm Camera and by generation of a digital elevation model of the workspace, which was imported into RSVP for target definition [1]. The operator verified the commanded motion in the RSVP simulator before uplink. After execution, image data were used to assess trench profiles, rasp placement and ejecta, post-scraping topography, TECP insertion impressions, sample quantity and delivery, and arm sensor data were imported into RSVP and played back to confirm completion. Playback was the diagnostic that located indurated layers, since it showed where the arm entered surface accommodation mode and where hard material impeded motion. Sample delivery used a photographic comparison procedure. The scoop was posed above the target inlet port, a RAC image was taken, and it was overlaid on an equivalent image from the Payload Interoperability Testbed on the ground so that positioning corrections could be made before the dump. RAC imaging before and after delivery documented port readiness and delivery outcome, which mattered because the TEGA doors did not open fully.

The RAC on the arm was also used for lander inspection. Before any terrain interaction the three lander footpads were imaged, the rear one by looking underneath the lander, to verify that digging could proceed safely [1]. The camera also imaged the MECA and TEGA inlet ports before and after each delivery, which established instrument readiness and, for TEGA, which doors had opened.

Technologies developed

The powered rasp with a spring-loaded pivoting bit, a knurled reaction plate and a two chamber scoop is the program’s specific contribution: a means of acquiring ice-cemented ground with a small lander-class arm and no percussive drill [1]. Its use as a vibrator to move cuttings inside the scoop is a second function obtained from the same actuator. On the operations side, the combination of guarded moves against current and torque thresholds with a surface accommodation algorithm that re-plans a dig when motion is impeded produced an arm that could keep working through unexpected hard ground without a ground cycle, which is the capability the mission’s sol count depended on. The delivery-side lesson is that the difficult part of icy-soil sampling was not cutting but retention: samples that survived acquisition were lost to sublimation and adhesion during transfer, which drove acquisition into the coldest part of the sol and drove scoop motion sequencing to minimize insolation.