Robotic Arm (RA) data sets
PHOENIX MARS ROBOTIC ARM 4 RDR DERIVED V1.0
The Phoenix Robotic Arm Derived Data consists of Robotic Arm (RA) Scoop Tip position data and components of force exerted by the RA. Data are included for both the spacecraft RA and the Payload Interoperability Testbed (PIT) RA. These data are derived from raw RA telemetry data that are not archived in PDS due to ITAR restrictions.
Data Set Overview
The Phoenix Robotic Arm Derived Data consists of Robotic Arm (RA) Scoop Tip position data and components of force exerted by the RA. Data are included for both the spacecraft RA and the Payload Interoperability Testbed (PIT) RA.
These data are derived from raw RA telemetry data that are not archived in PDS due to ITAR (International Traffic in Arms Regulation) restrictions. This dataset is sufficient for most purposes. Users who need more information about the RA and its activities may contact the JPL Office of Export Compliance, Mail Stop 202-204, 4800 Oak Grove Drive, Pasadena, CA, 91109, 818-354-9323.
The spacecraft data comprise the majority of the data in this archive. These are generated from data returned from the Mars Phoenix Lander Robotic Arm. They are found in the DATA directory. The archive also includes test data from the Payload Interoperability Testbed (PIT), a facility at the University of Arizona which housed an engineering model of the Phoenix lander. Test data are found in the TESTDATA directory. Both categories of data consist of comma-separated-value (CSV) text files with position, time, and force data from the RA telemetry. The test data include JPEG image files that correspond to the CSV data files.
The CSV files in this archive have undergone limited calibration, mainly the removal of data points which do not correspond to proper force retrievals (see Processing below).
Also included in this archive is a lookup table ACTIVITY.CSV in the INDEX directory which correlates data files with descriptions of the activities conducted by the RA.
More information on the Phoenix RA data and PIT test data can be found in Bonitz et al. 2008 [BONITZETAL2008 IN REF.CAT], Shaw et al. 2009 [SHAWETAL2009 in REF.CAT] and Arvidson et al. 2009 [ARVIDSONETAL2009 in REF.CAT]. See the latter for images of the RA and its scoop.
The CSV files contain 9 columns. The first column, time(s), in every file gives time stamps (spacecraft clock time marked in seconds from an arbitrary start point). The next four columns give the position of the RA scoop. These include: x_t(m) for which (+/-) values indicate a distance (north/south) of the RA shoulder, y_t(m) for which (+/-) values indicate a distance (east/west) of the RA shoulder, z_t (m) for which (+/-) values indicate the distance (below/above) the RA shoulder, and th_t(rad) which indicates the angle of inclination of the scoop. See Shaw et al. 2009 [SHAWETAL2009] for a diagram depicting scoop inclination values. The last four columns give force values. These include: Fr(N) for horizontal, radial force; Ft(N) for horizontal, tangential force; Fz(N) vertical force; and R(N) resultant of Fr(N) and Fz(N). Forces and positions are derived from motor currents, joint angles, and arm link lengths. See Shaw et al. 2009 [SHAWETAL2009] for more information. Sampling intervals for the data vary, but can be determined by looking at the first column of each file.
For the CSV files, the original source of the data is RA telemetry sent from the Phoenix spacecraft. Processing includes removal of data points which do not correspond to proper force retrieval (for example, if there was too much noise, or if one of joints 2-4 had zero torque). In these cases, data is listed as zero-valued. There were some arm poses for which torque could not be determined. Column headings were added to the top of each data file.
Commands used in directing the arm for each activity and the line numbers corresponding to data retrieved from that activity are also included in the headers at the top of each data file. Every line that is part of the header begins with the # (pound) character or with a double quotation mark. For TestData images, the original source of the data was PIT images taken by digital camera. No processing was performed on these images. They are just meant to provide context for the CSV files and the pixel values should not be used for analysis.
The main data set (Data) consists of CSV files, which can be viewed in a text editor or in a spreadsheet program such as Microsoft Excel. These files contain data from activities conducted on the surface of Mars. Most files contain the RA data collected on a particular sol, however in some instances the data from one sol is split over several files and 'a', 'b', 'c', etc. are appended to the end of each filename to indicate which data was collected first, second, third, etc. respectively. Diverse types of observations are included among the data, so a lookup table ACTIVITY.CSV is included and can be used to correlate filenames with the RA activities to which the data correspond. Because the files represent varying activity types, they encompass varying time ranges, but the first column in every file gives time stamps (spacecraft clock time in seconds).
Images of the martian surface before and after a particular RA activity are kept in other Phoenix archives. These images can be located using the Phoenix Analyst's Notebook (http://an.rsl.wustl.edu/phx) to look for images on the sol given in the RA file name.
The test data files follow the same format as the spacecraft data files described above, but they correspond to activities conducted in the PIT. Test data also include images that provide context for the test measurements (for example, images show the soil simulant after modification by the RA).
RA data are in the Payload Frame coordinate system. This is described in detail in the Phoenix Camera Software Interface Specification, available from the Phoenix Analyst's Notebook and archived with various Phoenix camera data sets in PDS, such as the Phoenix Robotic Arm Camera data set (PHX-M-RAC-2-EDR-V1.0). The origin of the coordinate system is at the shoulder of the RA (i.e. where the RA is attached to the lander deck). The z-axis points downward, the x-axis points north, and the y-axis points east.
Many thanks to Phoenix RA Team members Robert Bonitz, Joseph Carsten, Matthew Robinson, Ashitey Trebi-Ollennu, and Richard Volpe for their contributions to this archive.
Confidence Level Overview
From Shaw et al. 2009 [SHAWETAL2009]: 'When the RA was in contact with the soil, positions returned in the data were less exact than for moves in free air. When the arm was loaded against the surface, it flexed, resulting in errors in the calculated position. In extreme cases, the error was 2-4 cm at the end of the 2.4 m-long arm. [Relative RA positioning is] generally more accurate than absolute positioning, and positions were repeatable to within 2 mm.' See Shaw et al. 2009 [SHAWETAL2009] for sources of error and variations in the force data.
The uncertainty in the data is highly variable, and depends on the type of move and the type of terrain. It is probably reasonable to take the uncertainty to be at least 1N at any given time, but users are also expected to use their judgment when looking at the data. For an example of statistics from a dig: on Sol 22, when the RA conducted dig one in Snow White trench, mean radial force in the negative (towards the lander) direction was 9 N and the standard deviation was 5 N (this value includes soil variability effects). Only the portion of the data corresponding to touching soil was used to obtain these values. Keep in mind that the Robotic Arm was not originally intended to provide force values for soil mechanical analysis, the original reason for collecting motor currents (from which force values were calculated) was to ensure the safety of the instrument.
This data set was archived in the Planetary Data System following the successful completion of the required PDS peer review.
Data Coverage and Quality
Data were collected at different frequencies depending on spacecraft storage and downlink capabilities. Time stamps are included for each data point in the CVS files.
Because of variations in the data (see Shaw et al. 2009 [SHAWETAL2009]) it may be advisable to work with large sample sizes from the data set.