home people pubs gallery links

Pictures and videos from the WhiskerBot project

Click on a thumbnail to see the full-size image or play the video

Feb 24th 2004: mascot.jpg [126kB] This is our mascot, called [to be announced], and he (or possibly she) is hanging out with a Celoxica FPGA controller.

Feb 24th 2004: R4513.jpg [566kB] This is R4513, one of the rats who is helping us by behaving at our camera.

Feb 24th 2004: microvibrissal.placements.wmv [591kB] R4513 illustrating "microvibrissal placements" as he explores Pete's hand (infra-red recording).

Feb 24th 2004: unilateral.whisking.wmv [422kB] Some video of R4513 performing what might be "unilateral whisking".

Feb 24th 2004: roborat.gif [84kB] Humour from Steve Bell.

Apr 8th 2004: whisker_long.jpg [83kB] First attempt at attaching strain gauges was to a 900mm length of glass fibre representing a whisker. Actual whiskers will be no longer than 200mm.

Apr 8th 2004: Straingauge.jpg [36kB] Detail of prototype strain gauge attachment.

Apr 21st 2004: searching.behaviour.wmv [933kB] This footage is from our first foray into behavioural experiments. The trained rat (Stripe) passes along a corridor, investigating the mouth of each of three bays for the presence of the stimulus (sandpaper). You can see the stimulus hanging on the wall of bay 4. Entering bays 2 and 3, Stripe finds no stimulus, and moves on. Entering bay 4, Stripe finds the stimulus, and continues down the bay to look in the hopper at the end for a food pellet. Video runs at 1/2 real time.

Apr 21st 2004: true.pos.wmv [735kB] Close-up footage of Stripe encountering a stimulus, and deciding to investigate the bay (true positive). Video runs at 1/10 real time.

Apr 21st 2004: true.neg.wmv [628kB] Close-up footage of Stripe encountering no stimulus, and deciding not to investigate the bay (true negative). Video runs at 1/10 real time.

Apr 28th 2004: moulded_whiskers.jpg [170kB] Scan of the first whiskers out of the mould. Note truncated cone shape, curved and straight versions, and different materials. Two have been despatched to the man to have gauges attached, so we hope to record our first whisker output traces soon. These will form the input to the follicle model presented at Lausanne to generate the primary-afferent-analogue signals.

Apr 29th 2004: L5.pyramidal.jpg [84kB] Photo taken at 100x magnification of barrel cortex, showing a stained layer 5 pyramidal cell. The cell shows substantial dendritic branching within layer 5, but also sends a large dendritic process right up to layer 1, where it branches further (layer 1 and the outer surface of cortex are top right).

May 6th 2004: tom.harper.jpg [68kB] This is Tom Harper, the student who has been running our preliminary behavioural experiment, standing behind the four-arm maze. Standing next to the maze is the close-up digital video camera (not the high-speed one) which captures Stripe's micro-behaviour in the vicinity of one of the stimuli, and hanging from the ceiling is the overview analogue camera which captures Stripe's macro-behaviour.

May 6th 2004: twin.cameras.jpg [87kB] This shot shows the set-up from a top view, with both cameras looking down into the maze.

May 6th 2004: behavioural.arena.jpg [71kB] This is a detail shot of the four-arm maze. The maze has four arms. Well, d'uh. There is a sandpaper stimulus at one side of the entrance to each arm (rough or smooth), and a food pellet waiting in a hopper at the end of one of the arms, associated with the rough stimulus. All Stripe has to do is turn down the arm with the rough stimulus at the top.

May 19th 2004: whisker_in_clamp.jpg [285kB] This is the first one of the proper whiskers to come back with gauges attached. After clamping it in a vice, as shown, Martin pinged it...

May 19th 2004: gauge.data.1.gif [9kB] ...and this is a plot of the output recorded across one of the bridged pairs of strain gauges. The first mode resonance at about 27Hz is the main feature of this plot. This ringing dies away fairly quickly, in around a second. Ideally, we'd like this ringing to be much reduced - Martin has a couple of tricks up his sleeve, more to follow.

May 21st 2004: gauge.data.2.gif [10kB] This is the same as the previous plot, but this time a silicon sheath around the base of the whisker was used to increase the damping. The resonant frequency has dropped to just below 25Hz, and the ringing dies away much faster.

Jul 14th 2004: whisking.module.prototype.1.JPG [352kB] Dizzy's first prototype of the whisking mechanism. The drive is a biowire, which you can see strung temporarily between the pivot and an immovable post. On activating the biowire, which represents the intrinsic muscle, the whisker "protracts". On deactivating the biowire, the whisker "retracts" passively. Alternatively, a solenoid, which represents the extrinsic muscles, can be activated to cause active retraction.

Jul 14th 2004: whisking.module.prototype.2.JPG [194kB] This one shows the biowire clearly.

Jul 14th 2004: whisking.module.prototype.3.JPG [258kB] This one shows the solenoid clearly.

Jul 26th 2004: rat_cakes.jpg [156kB] For three days in Paris we had to survive on nothing but rat cakes and lager. It was fantastic.

Jul 27th 2004: the_tower.jpg [167kB] Dinner in the Eiffel Tower with other attendees of "Towards Artificial Rodents" (Thanks Jean & Agnès & everyone else).

Jul 28th 2004: modern_man.jpg [146kB] Caption competition, answers on a postcard please.

Oct 13th 2004: amouse.jpg [22kB] A couple of randomly chosen items, of similar intelligence level.

Nov 15th 2004: behavioural_arena_2.jpg [36kB] Our second behavioural arena is a custom-built lightbox using high frequency fluorescent illumination to backlight the animal and its whiskers, a Sony MiniDV recording the overview shot, and a DVR Photron high-speed video camera recording a small zone (approximately 240mm x 180mm) at 250fps. On top of the lightbox we can place any behavioural equipment that we like - the picture shows a simple circular arena - and close the top with a sheet of perspex to prevent the animal from rearing.

Nov 15th 2004: behavioural_arena_2_hs_frame_1.jpg [37kB] This frame shows the view through the high-speed camera in `roaming' mode - this is me pointing the camera and tripod at the animal by hand, a la Blair Witch. This gives us very close shots, so long as the animal plays along for a couple of seconds.

Nov 15th 2004: behavioural_arena_2_hs_frame_2.jpg [33kB] This frame shows the view through the high-speed camera when it is set up on a particular zone of the arena. When the animal enters that zone, we can trigger the camera, and hopefully record the behaviour we are looking for. In this frame, we have used a cunning arrangement of smoke and mirrors (actually, just mirrors) to give top and side views of the whiskers using the same camera.

Dec 23rd 2004: conference.jpg [110kB] A typical WhiskerBot video conference. Well, if by typical you mean the only time I remember us all being in the same place at the same time this year.

Mar 10th 2005: coin_whereisit.gif [377kB] A dystrophic rat finding a coin with its macrovibrissae, still from the Photron 1024PCI, and...

Mar 10th 2005: coin_whatisit.gif [286kB] ...subsequent investigation of the coin using the microvibrissae.

Apr 25th 2005: barrelettes.png [187kB] Barrelettes in our model of nucleus principalis responding to a mechanical step deflection of whisker B2. The follicle model (Royal Society Paper) converts the whisker stimulation pattern into ganglion activity, and the 5s model transforms that activity as seen in the animal. In this plot, you can see the principal barrelette (B2) responding vigorously, and adjacent barrelettes responding more weakly.

Apr 27th 2005: whiskerbot_schematic.gif [50kB] Animated short from Dizzy, based on real WhiskerBot schematics, showing the range of whisking and the basic layout. Comments like "where's the fourth wheel" will not be looked upon kindly.

Jul 13th 2005: be_the_rat.jpg [177kB] To know the rat, you must become the rat - see the world through the rat's whiskers. Here's a selection of Zen Whisking by some of our members at our group meeting in July.

Jul 13th 2005: follow_your_nose.avi [5MB] WhiskerBot's motor controller in action: salient stimuli light up superior colliculus, and WhiskerBot moves its nose to investigate more closely.

Jan 13th 2006: pros.jpg [456kB] The boys hard at work, trying to repair the demonstration whisker (after school-child damage), due in Baltimore in 48 hours.

Jan 13th 2006: isitabomb.jpg [471kB] The demonstration whisker, which can, well, whisk. Also houses the complete follicle model, and clicks like a real electrophysiology lab. We could sell these things to kids with scientific aspirations, if they didn't cost £[expletive deleted] each.

Jan 13th 2006: collage.jpg [319kB] WhiskerBot component parts taking shape - see if you can spot (a) the head, (b) a wheel, and (c) the off-road suspension.

Mar 17th 2006: whiskerbot_3D.png [71kB] Rendering of the WhiskerBot mechanical design, showing body with omni-directional wheel modules (two driven), and head with whiskers mounted on "follicle" spindles. That contraption at the front is the cooling duct, which involved more R&D than the rest put together.

Mar 17th 2006: head.jpg [329kB] Detail shot of the head in its current state of development. From bottom up: baseplate, shaft encoders (to drive phase afferents), mezanine floor, whisker spindles with mounted whiskers (these will bear the BioMetal wire), and pre-amp and signal conditioning board at the top.

May 24th 2006: tortoise.jpg [325kB] Ok, ok. So we haven't finished the body yet. And we've only done two of the whiskers (they're tricky, right?). But in the meantime, we've nailed WB's head to this old thing to get us mobile - we reckon it looks like a tortoise...?

Jul 21st 2006: ww-0003-web-demo.avi [11MB] Higher brain functions are beginning to take shape. WhiskerBot now has a Basal Ganglia (GPR model) - in the bottom left you can see the output of this module as inhibition strengths on five channels. In this instantiation, WB has only two actions to choose between: the fixed-action-pattern "orient-and-move-away", and the continuous-action-pattern "explore". See if you can work out which channels represent which actions, kids. This video runs at one quarter real time - the model has rates comparable to a real rat, so it is really fast in real time. The robot is not expected to match this performance!