Where on Earth is Aberystwyth??

Transcription

1

2

3 Where on Earth is Aberystwyth?? On the coast of Cardigan Bay, West Wales. We have mountains, sea, sky, sand, rocks, lots of interesting weather, and robots for every environment..

4 Tethered Aerobot Aerobot has highresolution camera and wide-angle lens Suspended from Hefilled envelope Up to ~100m Tethered to rover Altitude to ~100m, with ~100m image footprint on ground (components not to scale)

5 Tethered Aerobot Provides area context for ground-based rover mission Traverse and waypoint planning (area DEM) Identification of possible science targets See targets and hazards not visible from rover location Offers a viewpoint unavailable to the rover Image feature resolution falls between orbital and ground imagery (~5cm or better) Offline or online use Offline: deploy, image, stow, traverse Online: deploy, image, traverse, image, traverse...

6 PRoViScout Planetary Robotics Vision Scout The PRoViScout project aims to demonstrate visually-guided, autonomous science target assessment, operations planning and navigation for a planetary rover class mission. It brings together major European groups currently working on planetary robotic vision, supported by NASA-JPL from the USA. PRoViScout is a collaborative project within the EU FP7-SPACE framework (Grant Agreement No ). Partner Institutions include: Aberystwyth University, Joanneum Research, SciSys Ltd., University of Strathclyde, University of Leicester, DLR, GMV, CSEM, Trasys, Czech Technical University Prague, UCL, and JPL.

7 Use of Aerobot in PRoViScout PRoViScout field test mission will explore a Mars analogue environment with a rover Initial mission plan will be modified autonomously to include opportunistic science targets Aerobot will provide wide-area multispectral imagery to inform the initial plan: area DEM for waypoint & initial science target choice mineralogical overlay to refine target choice Aerobot data stands in for pre-landing orbital image data Nominally offline use but may test online capability

8 Previous Aerobot research at AU Cooperative Aerobots autonomously flying in formation. Each Aerobot has its own behaviourbased controller. The cooperative control algorithm maintained formation even after external disturbances. Project partner was SciSys, Ltd. ILP aerobot undergoing acceptance trials at ESTEC Planetary Testbed Facility. ESA-funded project on autonomous image-based localisation for a future Mars aerobot mission. Project partners included SciSys Ltd, University of Leicester and Joanneum Research.

9 Current development model Aerobot

10 Aerobot left side Embedded Linux PC FitPC-2i, 8-14V, 6-10W WiFi antenna IMU Nylon bar Image courtesy of JR/Bauer Aluminium plate

11 Aerobot right side LiPo battery pack 2200 mah GigE camera 5 MPix Light baffle Image courtesy of JR/Bauer

12 Oblique view showing suspension Picavet suspension GPS receiver IMU Image courtesy of JR/Bauer SparkFun RAZOR 9DOF IMU 50mm x 28mm

13 Picavet suspension Often used in kite aerial photography 4 strings + pulleys form a constant length suspension Platform stays level even if tether angle changes Single tether disconnects platform from balloon rotation Sideways swaying motion still possible

14 Aerobot system diagram Aerobot provides 5 MPix monochrome camera Additional image data: GPS Lat, Long, Alt Magnetic heading Pitch & Roll angles Filter, Timestamp PC controls sensors & camera Access to functions via clientserver protocol over TCP/IP Local or remote control possible Multiplatform client-side C library (CORBA interface soon) Protocol also implemented in Java and Python

15 Aerobot GigE camera specifications Model Sensor Pixel size Resolution Active sensor size Lens focal length Hyperfocal distance Field of view Ground footprint Pixel field of view Ground pixel size AVT Prosilica GC2450 Sony ICX625 monochrome 3.45µm x 3.45µm 2448 x 2050 pixels 8.45mm x 7.07mm 8mm (TBC) 6m (focus 3m - infinity) 55.7 x m x 88m at 100m high 53m x 44m at 50m high (0.431 milliradians) 43mm at 100m high 22mm at 50m high

16 Aerobot image capture (PRoViScout) Capture images at roughly regular distances while traversing area of interest Try to get good overlap (70% ideal, 50% usable) At each waypoint: Auto-expose Image at nominal Image at 0.5, 1.5 times (exposure bracketing) Repeat several times at each waypoint to allow for Aerobot motion

17 Initial test flight At AU robotics workshop, about 110m tether length

18 First light! Note the inclusion of locale-specific scale reference objects...

19 First Field Tests Field test at Clarach Bay beach scheduled as part of PRoViScout Aims: February 2011, in Aberystwyth Test out Aerobot functionality Collect representative data to exercise image pipeline and DEM generation Acquire 3D laser scan of beach as ground truth But the weather forecast for that week was not good...

20 Day 1 Due to worsening weather prospects, an attempt was made to launch the Aerobot early on the first day of the trial. The balloon was transported inflated in a trailer, but suffered helium loss during handling. Lifting power was badly affected, and launch was not possible (no spare helium was available).

21 Day 2 Bright day but getting very windy with sudden gusts. Balloon fully inflated, but lifting power not sufficient to cope with gusts: Failure to launch :-( Some dramatic movie footage was obtained from a tiny witness camera on the Aerobot.. A windy day at Clarach Bay

22 The backup plan... Self-powered autonomous aerobot deployed on cliff-top. Capture some images, but under less controlled conditions. Wireless comms problems but some data was obtained. Many thanks to Dr Stephen Pugh!

23 Cliff-top image data Somewhat challenging to process.. Spot the rover!

24 Day 3 Simulated Aerobot flight at AU robotics workshop Aerobot hung from long ladder over an outside balcony Targets placed on floor White spheres, easily recognised in data Several other obstacles included Ladder moved between image sequences Large overlap between successive images

25 Aerobot image sequence

26 Image processing operations Interest point identification Interest point matching Camera pose recovery Dense DEM generation DEM rendered in 3D

27 Interest point matching Disparities are obtained from matching interest points (automatically identified features) in successive images

28 Camera pose recovery Images Center for Machine Perception, Czech Technical University in Prague. Disparities are combined with attitude data from the Aerobot to recover camera positions and orientations by bundle block adjustment. Interest point coordinates are also refined.

29 Dense DEM generation Final dense DEM generation using disparities and refined camera positions.

30 DEM rendered in 3D The final DEM can be rendered as a 3D textured mesh and displayed with a suitable 3D viewing application.

31 Issues & lessons learned High/gusting wind is a launch issue - need to maintain excess lift Not sure yet exactly how low a wind speed is safe take measurements Consider using larger envelope for more lift / more blimp-like envelope for added stability (beware tradeoffs..) Purge air & top up properly Always take some spare helium for topping up after transport Framework needs to be a bit more robust Could do with losing a bit of weight as well (1.2kg)

32 Next Steps Aerobot has yet to provide truly representative image data.. Multispectral capability Conduct further field test(s) at Clarach Bay! New board camera and custom filter wheel Improve height estimation Barometric altimeter IMU on tether Proper sequencing & remote operation currently uses ssh and shell scripts... implement Corba interface

33 Where we'd like to get to... Example perspective view of a 3D mineralogy map. This map was generated using data from the CRISM instrument on NASA's Mars Reconnaissance Orbiter. The image shows clays (coloured magenta and blue) which are concentrated on the slopes and mesas and along the canyon walls. Mineralogy map of part of Nili Fossae region. Image courtesy NASA/JPL/JHUAPL/U of Arizona/Brown A wide-area mineralogy context map like this could be used to inform decision-making modules on-board a scout rover.

34 Any Questions?