Beagle 2 PAW and ARM

Source: Bridges et al. 2017, Royal Society Open Science, figure 4 [5]; CC BY 4.0.

Overview

Beagle 2 carried a five-degree-of-freedom manipulator, the Anthropomorphic Robotic Manipulator, terminating in a fixed instrument cluster called the Position Adjustable Workbench [9]. The PAW is not an end effector in the usual sense: it is the payload, carrying the stereo cameras, microscope, X-ray spectrometer, Mossbauer spectrometer, rock corer grinder, wind sensor, wide angle mirror, torch, spoon and the launch tube for the PLUTO subsurface mole. The arm exists to point that cluster.

The lander landed on Mars on 25 December 2003 and was never heard from [5]. It was identified in HiRISE imagery in 2014 and 2015 and the identification published in 2017; the configuration seen is consistent with lid deployment and partial solar panel deployment, which would have left the UHF antenna covered [5].

The design driver is mass. Landed mass was 33.18 kg against a 69 kg probe mass limit, of which the scientific payload was 8.896 kg, and the program aimed at the highest ratio of payload to support systems flown to Mars, achievable only with an integrated design and minimal or zero redundancy [9]. The complete experiment package draws less than 40 W. An earlier design-stage figure had put the probe at about 65 kg and 0.95 m in diameter [14]. An independently published figure of about 33.2 kg for the landed mass agrees [11].

Specifications

Parameter	Value
ARM degrees of freedom	5, with the PAW permanently attached at the wrist
ARM length, fully extended	109 cm, body joint center to PAW wrist joint center
Joint actuator	DC brushed motor through a 100:1 harmonic gearbox
Joint position sensing	potentiometer on the output shaft
Typical joint speed	0.6 deg/s, axes 1 to 3
Forearm plus PAW length	400 mm
Contact sensing	piezoelectric force load washer, 3 pC/N
Stop delay after contact detection	about 100 ms
ARM mass	2.11 kg
PAW mass	2.75 kg
Rock corer grinder mass	348 g flight model, without translation table
Rock corer grinder envelope	about 30 by 60 by 100 mm
Rock corer grinder power	6 W
Grinding depth	up to 6 mm of weathering rind
Ground patch diameter	10 mm
PLUTO total mass	890 g: 340 g mole, 550 g deployment unit
PLUTO envelope	about 380 by 90 by 80 mm, mole 280 mm long
PLUTO power	3 W, 17 W for the post-landing pin-puller release
Mole sampling depth, vertical	1.5 m baselined
Mole temperature measurement depth	1.25 m maximum
Mole sample volume	about 0.20 cm3
Mole hammer interval	one shock every 5 s
Mole lateral travel per stroke	about 10 mm
Spoon sample volume	no more than 20 mm3
Onboard processor	ERC32
Comms bands and rates	UHF, forward 2 and 8 kbit/s, return 2 to 128 kbit/s
Battery	42-cell lithium ion
Solar array	four deployable panels, about 1 m2 effective, triple-junction GaAs on germanium
Digital elevation model working zone	0.6 to 1.2 m near-stereo range
Multispectral panorama rotation	360 degrees
Elbow-alone tool sweep at the rock corer grinder	up to 5 mm, about 0.7 degrees
Gas analysis package mass spectrometer	6 cm radius magnetic sector
Rock corer grinder qualification-model translation table	about 30 mm diameter flat area, not flown
Viking trench depth reached, for comparison	20 cm

Values from [9].

Mission profile

Parameter	Value
Landing site	Isidis Planitia, 265.0 W, 11.6 N
Landing date	25 December 2003
Planned primary mission	up to 180 sols
Planned complete sampling cycles	5 to 6 samples over about 80 to 100 sols
Rocks planned for full in situ analysis	about 3
Mole samples planned in the primary mission	about 3
Relay assets	Mars Express and Mars Odyssey
Outcome	landed, no contact; identified in HiRISE imagery

Landing site, planned primary mission, planned sampling cycles, rocks and mole samples planned, and relay assets are from [9]; landing date and outcome are from [5].

Manipulator

The arm is not a general-purpose manipulator: with the PAW fixed to the wrist, every instrument placement is an arm placement, and the requirements levied on the ARM are optical rather than mechanical. Digital elevation model construction of the working zone requires the PAW to view the scene from at least two or three perspectives while holding targets in focus; the multispectral panorama requires the PAW to point straight up for height advantage and rotate about that axis; the close-up lens standoff constrains PAW orientation to avoid surface contamination and navigational hazards. Contact is detected by the piezoelectric force load washer, which produces charge when the PAW meets a resistive surface: a rock, a calibration target, the gas analysis package inlet or a sample dish. The sensor is quiet during normal gentle arm motion, so it functions as a contact event detector rather than a force sensor. After contact is detected the arm continues to drive for a short interval set by joint stiffness, with the exact stopping time depending on the instrument being placed and the expected PAW orientation relative to the surface. Approach forces for each instrument are then derived from that point as a function of arm velocity, direction and time, and the weight of the PAW or its vertical component can be used at a suitably chosen site. Final approach precision comes from single-joint motion: moving the elbow alone sweeps the PAW forward to prepare a surface with the rock corer grinder, within the surface preparation accuracy the grinder achieves. The X-ray spectrometer was proposed for all first contacts with a candidate sample because it tolerates contact loads better than the other instruments, establishing a reference point from which subsequent positioning is more accurate.

Contemporaneous JPL work on planetary manipulators took a different path: the Planetary Dexterous Manipulators task pursued lighter-weight ultrasonic actuation and camera-guided autonomous control of rover-mounted arms for instrument placement and sample acquisition, rather than a fixed-cluster arm executing preplanned motions [6]. The same effort was developing new piezoelectric ultrasonic motors for light-weight, high-torque actuation in planetary environments, a different actuator path from the ARM’s brushed-motor and harmonic-gearbox joints [6]. Earlier JPL lander studies using a laboratory replica of a Mars lander had already demonstrated automated visual positioning control of robotic sampling and instrument emplacement with a lightweight, gas-deployed, two meter telescopic arm [10], the same problem Beagle 2 solved instead with predetermined maneuvers, safe-point configurations and a contact-timeout abort rather than vision-guided servoing. A set of configurations designated safe points is defined within the sphere of PAW and ARM operations, and advancement toward a target proceeds until contact is detected or a timeout expires, in which case corrective action may abort the sequence and return the arm to a safe or default position [9].

PAW instrument cluster

The PAW carries the in situ instruments and the tools that prepare samples for them and deliver material to the gas analysis package on the lander deck. That package is a magnetic sector mass spectrometer that runs in both static and dynamic modes, and it is the instrument that every PAW tool and the mole feed. Institutional responsibility was distributed: the X-ray spectrometer and the wide angle mirror, torch and spoon to the University of Leicester, the stereo camera system to Mullard Space Science Laboratory, the microscope to Bern, PLUTO and the sampling mole to DLR Cologne, the rock corer grinder to Hong Kong Polytechnic University and the wind sensor to Oxford, with Leicester supplying the PAW structure, control electronics and cabling.

Element	Function
Stereo camera system	panoramic and working-zone stereo imaging through filters, close-up lens at about 80 mm
Microscope	self-illuminating, contact required, images individual grains, 4 micrometer per pixel scale
X-ray spectrometer	elemental composition, Mg to Nb, Fe-55 and Cm-244 style excitation after Viking rather than the APXS proton mode
Mossbauer spectrometer	iron-bearing mineralogy of the prepared surface
Rock corer grinder	removes weathering rind and cores the ground patch for delivery to the gas analysis package
PLUTO mole	subsurface soil sampling, subsurface temperature, soil mechanics from intrusion behavior
Spoon	backup sampling of unconsolidated soil
Wide angle mirror	360 degree first view of the site through the right stereo camera before the PAW is unstowed
Wind sensor	boundary layer wind regime at the PAW

The wide angle mirror is a deployment safety device as much as an instrument: it is held under spring tension until the lid and solar panels open, then moves into the field of view of the right-hand stereo camera, giving a panorama that lets mission planners unstow the PAW without hitting anything. The rock corer grinder is dual function: grinding removes weathering rind to produce individual or arrayed fresh surfaces sized for the spectrometers, and after in situ analysis the same tool cores the ground patch by a combined hammering and rotating action of the main drive, retaining chippings in a micro end effector jaw device and delivering a reproducible volume suited to the gas analysis package ovens. A translation table developed for the qualification model would have swept the tool head to produce a larger flat area, but it was not adopted for flight on mass and other grounds. The elements, functions and this account of the wide angle mirror and rock corer grinder are from [9].

The stereo camera system is a pair of Space-X CCD micro-cameras with 48 degree optics and a 209 mm baseline between the CCDs, each behind a titanium filter wheel carrying twelve interference filters; one filter position holds the close-up lens, giving 60 micrometer per pixel resolution at the 80 mm standoff [7]. Generating the digital elevation model that the PAW’s placements depend on was the camera system’s primary design goal, built by merging overlapping stereo pairs through lens distortion correction, stereo matching and global adjustment into a single model free of artifacts at the seams between views [7].

PLUTO mole

PLUTO is a self-penetrating mole tethered to a support mechanism on the PAW, derived from a Russian self-burying penetrometer and scaled down for space under an ESA technology program. Once a small initial penetration is achieved it needs no reaction force from the lander: an internal inertial hammering mechanism drives it forward, powered through the tether, and it can reach depths several times the length of its casing. Reaching 1 m vertically was estimated at 30 minutes to 1 hour under Mars gravity and the expected soil resistance [9]. A commandable mechanism at the tip opens at the sampling depth to take a soil sample and closes to secure it. Retrieval reverses the cable reel, pulling the mole back into the launch tube on the PAW while the arm holds position, after which the arm moves to the gas analysis package inlet port and the tip mechanism opens again to discharge. Retraction was tested on a variety of analog materials, and can be assisted by running the hammer while rewinding the cable, an option reserved for deep sampling. Horizontal deployment uses the same hammering to crawl the mole across the surface, a mode verified in an earlier ESA technology activity. The intended use was to reach a boulder beyond the reach of the PAW and then divert downward to sample the protected soil beneath its overhang. Beyond sampling, the mole was to act as a platform for subsurface temperature measurement against time and depth, and its intrusion behavior was to yield soil mechanical properties and layering. The rationale for the mole’s sampling depth is stated directly: Viking’s trench depth remained inside a presumably highly oxidized layer in which no organic matter was measured, and the top meter of bulk soil is expected to have been exposed to atmosphere and ultraviolet flux by aeolian mixing, destroying organics on a decadal scale. There is no precise positioning requirement for the mole, unlike the instruments [9].

The same hammering approach was flown again on the InSight lander’s HP3 heat flow probe, where it penetrated only about 40 cm before an unexpected cohesive duricrust removed the soil friction the mechanism needed to advance [12]; PLUTO itself was never tested against Martian soil.

Compute, power and communications

Lander electronics carry an ERC32 processor and provide power conditioning, distribution and management plus the instrument and equipment interfaces. Onboard software is in four modules: a bootstrap loader in PROM that runs mission phase detection and executes the application image; common software providing low-level drivers and utilities; probe software covering entry, descent and landing and handover; and lander software providing subsystem control and experiment functionality. The last three are stored in EEPROM and loaded into RAM for execution. Power comes from the lithium ion battery kept warm overnight by insulation and heater power, charged by the deployable solar panels.

Communication is UHF only, bidirectional with Mars Express and Mars Odyssey, with receive and transmit antennas integrated into the lander lid structure [9]. A UHF relay to an orbiter overhead, rather than a direct-to-Earth link, was the standard approach for Mars landers of this era because it avoids the mass and pointing accuracy a direct high-gain antenna would need [8]. The UHF forward and return frequency pair later standardized for the Mars relay network, 401.585625 and 401.528711 MHz, is the same band used by Mars Express and Odyssey to talk to landers and rovers on the surface [8].

Modes of operation

The post-landing sequence is fully automatic: deploy solar panels, start battery charging for the remainder of the day, take a monochrome image with the right stereo camera and the wide angle mirror positioned automatically in the field of view, and enter an overnight low-power mode if no contact occurs on sol 1, continuing in low power with daytime charging until communication is possible. Once contact is established and the first image assessed, the PAW and ARM are deployed to a safe default configuration and an imaging phase builds the stereo coverage needed for a digital elevation model of the working zone. Thereafter the sol is split by power budget rather than by science priority: arm positioning and gas analysis package processes run during the day, and low-resource activities such as spectrometry and microscopy run at night. A complete sampling cycle was estimated at tens of sols [9].

Loss and identification

No signal was received after landing, and the lander was declared officially lost on 6 February 2004 [11]. ESA first announced in January 2015 that MRO imagery showed the lander’s wreckage, ahead of the formal identification published in 2017 [11]. A 2004 recalculation cut the original 3-sigma landing ellipse from a 174 km major axis to 57 km, and High Resolution Imaging Science Experiment coverage of that smaller, 57 by 7.6 km ellipse at 30 cm per pixel led to a candidate identification in 2014, with further targeted imaging and super-resolution processing supporting it [5]. The identification rests on an object about 1.5 m across with a distinctive multilobed shape, high reflectivity against the local terrain, specular reflections, and a position close to the center of the planned ellipse. HiRISE has 25 cm per pixel resolution, so the lander is a few pixels across and its configuration is partly masked by those specular reflections [5]. What can be resolved is consistent with deployment of the lid and then some or all of the solar panels. Each panel is 0.57 m in diameter and the fully deployed lander is 1.7 m across, with the panels at 20 degrees to the horizontal when fully deployed [5].

Failure to deploy all four panels is a sufficient explanation for the silence, because the UHF antenna is exposed only when the last panel opens [5]. Entry, descent and landing therefore appear to have worked, with only the final panel deployment failing. Everything downstream of that, including the ARM, the PAW and PLUTO, was never exercised on Mars.

Technologies developed

PLUTO is the program’s transferable product: a self-penetrating tethered mole requiring no reaction force from the lander after initial penetration, capable of both vertical burial and horizontal crawling on the same hammer mechanism, and carrying its own temperature sensing and soil-mechanics inference [9]. The PAW itself is an architectural argument that did not recur: rather than an arm carrying interchangeable tools to instruments mounted on the body, Beagle 2 fixed the entire in situ instrument suite to the wrist, which removes every sample transfer step between instruments at the cost of making arm placement accuracy the limiting factor for all of them. Contemporary JPL landers took the opposite path: the Mars Polar Lander’s MVACS Robotic Arm and the later Phoenix Robotic Arm each carried a single scoop end effector plus a camera, moving samples to a fixed deck instrument rather than carrying the instrument suite itself, with Phoenix’s version a 4-degree-of-freedom, 2.4 m arm [2] [3]. The MVACS arm also positioned its own camera for close-up imaging of samples still in the scoop before delivery, the same camera-on-arm inspection role Beagle 2 gave the stereo pair mounted on the PAW [2]. Phoenix’s arm delivered dry and icy soil samples to two deck instruments in turn and also placed a probe against the regolith for humidity readings, another way a scoop-and-deck arm covers what Beagle 2 did instead by fixing every instrument to the wrist [3]. The contact strategy that follows from Beagle 2’s choice, a piezoelectric charge detector used as an event trigger with a fixed post-contact drive interval set by joint stiffness, is the minimum-mass alternative to a force-torque sensor. The mass discipline is the other legacy, accepted with minimal or zero redundancy [9].