# BIAUV Design Review # # tue sep 30 2008 # # hobson, mcewen, erickson, sherman, headley # edgington, herlein, maughan, brown # chaffey, smith, rock, schlining, cline # birch, chaffey, paull, graybeal # jensen, barry ================= = Input ================= ================= = Action Items ================= - check out SnagIt ================= = Notes ================= when intervention not needed faster, straighter than ROV, w/ better images # missions basic benthic transect 200 images/run image 1/30s 400 images continuous benth trans 2000 images, 30 gb data target of interest mosaic 10x10 patch e.g. whale fall 10x 10m long transects 70 images /site, mult sites/mission 5km between sites # functional reqmts Dorado + Zephyr color, hi res images (1 mm) even illum (2 sources) image 1/1.5 s store up to 2000 images link images to AUV metadata # challenges is dorado (cruiser) appropriate less stable than Abe, can't fly slow navigation precision can we capture, store and light high res images fire camera and strobes fast enough lot of water between camera and seafloor # Controls/Navigation 3x2 m image field amplitude of motion, how fast it's changing lower bound is 1/750 s (0.75 m/s, 1 mm res) will start flying 1.0 m/s shutter speed trade: slow enough to expose, fast enough not to smear cline: what is smearing: camera shake, blurring chaffey: modulation transfer ratio is between 0 mm and 1 mm movement -> linear transfer ratio we're using 1 mm as a point of reference rock: how do the operate seabed? veh 0.5 m/s, black and white [veh pointing and rate stability] schlining: how big is image patch? 3x2 m roll:amplitude and rate plenty of roll margin (slow shutter speeds reqd) pitch, wobble (+/-9cm) rates, small but adds to fwd speed max shutter speeds ~0.9 ms biggest prob is wobbling herlien: are shutter speeds discrete ? we will use strobe time to set exposure rock: lateral has greater impact than cline: where is this going [scale drawing] approx errors, 1/2 s, veh pitch up # payload camera [Prosilica] [Nikon D3] [test images, Nikon] single strobe 0.8 ms stopped down iris (good depth of field) [test images, prosilica 15 mS, G10, f8] hanu says most important thing is dynamic range -> blow out center vs dark corners prosil has lower dynamic range can do some gamma correction to pull out backgd details Nikon ~OK at 3m range, better at 2.5 m 1mm a little beyond what we can do, right on edge [camera selection decision matrix] herlein: does prosil embed metadata? no. we'll still need post processing maughan: are there processing cards (sensor vs full image processing solution) [payload camera] what standoff is best what FOV is best what cam/lighting settings are best [nautilus sphere] flexible option 90% transmission, could be defects [strobes] Sea and Sea YS-250 (used w/ Rover) AquaFlash (not as fast at high power) - needs reflector - 6km housing - we plan to use this # Mechanical [BIAUV] length = CTD vehicle [FOV] rock: no DVL? just moved forward, we can use beam info to tell angle assuming we can make reflectors strobes adjustable 2-3m standoff smith: abosrbing full vehicle payload? no sci sensors on vehicle vehicle is optimized for imaging, it's valuable to run mapping and imaging together (on same vehicle ideally) need to eliminate backscatter [nose section] a little reserve bouyancy sonardyne iUSBL nav imagenex OAS 675 kHz strobe [ext mid body] need to cut 9.5 in this section for FOV; final camera will be lower add 2 lasers next year rock: why don't want to move sphere down? to protect sphere from damage brown: will hole affect control? historically no, may cause drag, could cause pitch? # Electrical [reqmts and interfaces] [components] camera strobe XPort Nikon WT-4 Ethernet switch [schematics] [block diagram] is there strobe feedback? could do [power budget] 80 W peak power # Guidance and Navigation [iUSBL guidance and nav] will use DVL aided dead reckoning graybeal: is it better to put in this data on the fly? rock: need to be able to update metadata as you improve your estimate of navigation data what you store on the fly may be different from what is used later iusbl 1% slant range, dvl 3% DT may be better to always fly toward the beacon rock: how far off can you fly on parallel track? what is usbl FOV and accuracy must convert slant range error to xy error graybeal: will you be doing these computations on the fly (high comp overhead)? we do it in docking, this is not compute intensive chaffey: could you over sample as well? could run 1/2 m or 1/4 m tracks smith: what is limit of dist? need to be able to run 1-2 km to sample biota. Can we run 2 homers on either end ? yes Jim barry is OK w/ 3 km long transects with dead reckoning. it would be nice to be able to tie it to acoustic image. rock: we'll demo navigation of CTD w/ mapping data next year to 1 m accuracy. smith: can you use passive sonar instead of active? the beacons get 500000 xmissions (5-10 yrs) barry: could we fly along a depth path? we'll use track points and take errors, but this could be done. will record depth/alt as we go. rock: 3% of DT can be knocked out if there is relief where we are flying [mission file] [simulation] [next steps] graybeal: there are several lockin sources w/ diff accuracies; is there something in place to evaluate and select from these sources? we'll be doing very specific sites, don't need to do anything on the flight...in a broader sense, have you thought about blending these sources. kalmann filter. would it make sense to have one common navigation module. rock: this projct has hooks to do that, but bigger than this project. like idea of general nav pckg for use on all vehicles. # payload control software cline:need to be able to view these in google earth smith: deeper you go, more important a transect is vs mosaic; don't need complete image over whole transect barry: can't have google earth, but will have swaths, that we can place them on a map; don't need much more than where they are located on seafloor. provide us w/ a method to get image into arcGIS, we need to be able to plot images smith: want to be able to place with bathy data cline: we should be able to look at these in a better 3d way barry: visualization lab; think about cost vs what else we might do with that investment usually big wall isn't how the important information is gotten. want to see changes, need rock: big diff w/ MPA, overlapping images. don't put off decisions about metadata; brian developing tols for MPA (SIMPA) ; we should coordinate with Brian and Rock.