annotate thesis/cortex.org @ 450:432f2c4646cb

sleepig.
author Robert McIntyre <rlm@mit.edu>
date Wed, 26 Mar 2014 03:18:57 -0400
parents 09b7c8dd4365
children 0a4362d1f138
rev   line source
rlm@425 1 #+title: =CORTEX=
rlm@425 2 #+author: Robert McIntyre
rlm@425 3 #+email: rlm@mit.edu
rlm@425 4 #+description: Using embodied AI to facilitate Artificial Imagination.
rlm@425 5 #+keywords: AI, clojure, embodiment
rlm@422 6
rlm@437 7
rlm@439 8 * Empathy and Embodiment as problem solving strategies
rlm@437 9
rlm@437 10 By the end of this thesis, you will have seen a novel approach to
rlm@437 11 interpreting video using embodiment and empathy. You will have also
rlm@437 12 seen one way to efficiently implement empathy for embodied
rlm@447 13 creatures. Finally, you will become familiar with =CORTEX=, a system
rlm@447 14 for designing and simulating creatures with rich senses, which you
rlm@447 15 may choose to use in your own research.
rlm@437 16
rlm@441 17 This is the core vision of my thesis: That one of the important ways
rlm@441 18 in which we understand others is by imagining ourselves in their
rlm@441 19 position and emphatically feeling experiences relative to our own
rlm@441 20 bodies. By understanding events in terms of our own previous
rlm@441 21 corporeal experience, we greatly constrain the possibilities of what
rlm@441 22 would otherwise be an unwieldy exponential search. This extra
rlm@441 23 constraint can be the difference between easily understanding what
rlm@441 24 is happening in a video and being completely lost in a sea of
rlm@441 25 incomprehensible color and movement.
rlm@435 26
rlm@436 27 ** Recognizing actions in video is extremely difficult
rlm@437 28
rlm@447 29 Consider for example the problem of determining what is happening
rlm@447 30 in a video of which this is one frame:
rlm@437 31
rlm@441 32 #+caption: A cat drinking some water. Identifying this action is
rlm@441 33 #+caption: beyond the state of the art for computers.
rlm@441 34 #+ATTR_LaTeX: :width 7cm
rlm@441 35 [[./images/cat-drinking.jpg]]
rlm@441 36
rlm@441 37 It is currently impossible for any computer program to reliably
rlm@447 38 label such a video as ``drinking''. And rightly so -- it is a very
rlm@441 39 hard problem! What features can you describe in terms of low level
rlm@441 40 functions of pixels that can even begin to describe at a high level
rlm@441 41 what is happening here?
rlm@437 42
rlm@447 43 Or suppose that you are building a program that recognizes chairs.
rlm@448 44 How could you ``see'' the chair in figure \ref{hidden-chair}?
rlm@441 45
rlm@441 46 #+caption: The chair in this image is quite obvious to humans, but I
rlm@448 47 #+caption: doubt that any modern computer vision program can find it.
rlm@441 48 #+name: hidden-chair
rlm@441 49 #+ATTR_LaTeX: :width 10cm
rlm@441 50 [[./images/fat-person-sitting-at-desk.jpg]]
rlm@441 51
rlm@441 52 Finally, how is it that you can easily tell the difference between
rlm@441 53 how the girls /muscles/ are working in figure \ref{girl}?
rlm@441 54
rlm@441 55 #+caption: The mysterious ``common sense'' appears here as you are able
rlm@441 56 #+caption: to discern the difference in how the girl's arm muscles
rlm@441 57 #+caption: are activated between the two images.
rlm@441 58 #+name: girl
rlm@448 59 #+ATTR_LaTeX: :width 7cm
rlm@441 60 [[./images/wall-push.png]]
rlm@437 61
rlm@441 62 Each of these examples tells us something about what might be going
rlm@441 63 on in our minds as we easily solve these recognition problems.
rlm@441 64
rlm@441 65 The hidden chairs show us that we are strongly triggered by cues
rlm@447 66 relating to the position of human bodies, and that we can determine
rlm@447 67 the overall physical configuration of a human body even if much of
rlm@447 68 that body is occluded.
rlm@437 69
rlm@441 70 The picture of the girl pushing against the wall tells us that we
rlm@441 71 have common sense knowledge about the kinetics of our own bodies.
rlm@441 72 We know well how our muscles would have to work to maintain us in
rlm@441 73 most positions, and we can easily project this self-knowledge to
rlm@441 74 imagined positions triggered by images of the human body.
rlm@441 75
rlm@441 76 ** =EMPATH= neatly solves recognition problems
rlm@441 77
rlm@441 78 I propose a system that can express the types of recognition
rlm@441 79 problems above in a form amenable to computation. It is split into
rlm@441 80 four parts:
rlm@441 81
rlm@448 82 - Free/Guided Play :: The creature moves around and experiences the
rlm@448 83 world through its unique perspective. Many otherwise
rlm@448 84 complicated actions are easily described in the language of a
rlm@448 85 full suite of body-centered, rich senses. For example,
rlm@448 86 drinking is the feeling of water sliding down your throat, and
rlm@448 87 cooling your insides. It's often accompanied by bringing your
rlm@448 88 hand close to your face, or bringing your face close to water.
rlm@448 89 Sitting down is the feeling of bending your knees, activating
rlm@448 90 your quadriceps, then feeling a surface with your bottom and
rlm@448 91 relaxing your legs. These body-centered action descriptions
rlm@448 92 can be either learned or hard coded.
rlm@448 93 - Posture Imitation :: When trying to interpret a video or image,
rlm@448 94 the creature takes a model of itself and aligns it with
rlm@448 95 whatever it sees. This alignment can even cross species, as
rlm@448 96 when humans try to align themselves with things like ponies,
rlm@448 97 dogs, or other humans with a different body type.
rlm@448 98 - Empathy :: The alignment triggers associations with
rlm@448 99 sensory data from prior experiences. For example, the
rlm@448 100 alignment itself easily maps to proprioceptive data. Any
rlm@448 101 sounds or obvious skin contact in the video can to a lesser
rlm@448 102 extent trigger previous experience. Segments of previous
rlm@448 103 experiences are stitched together to form a coherent and
rlm@448 104 complete sensory portrait of the scene.
rlm@448 105 - Recognition :: With the scene described in terms of first
rlm@448 106 person sensory events, the creature can now run its
rlm@447 107 action-identification programs on this synthesized sensory
rlm@447 108 data, just as it would if it were actually experiencing the
rlm@447 109 scene first-hand. If previous experience has been accurately
rlm@447 110 retrieved, and if it is analogous enough to the scene, then
rlm@447 111 the creature will correctly identify the action in the scene.
rlm@447 112
rlm@441 113 For example, I think humans are able to label the cat video as
rlm@447 114 ``drinking'' because they imagine /themselves/ as the cat, and
rlm@441 115 imagine putting their face up against a stream of water and
rlm@441 116 sticking out their tongue. In that imagined world, they can feel
rlm@441 117 the cool water hitting their tongue, and feel the water entering
rlm@447 118 their body, and are able to recognize that /feeling/ as drinking.
rlm@447 119 So, the label of the action is not really in the pixels of the
rlm@447 120 image, but is found clearly in a simulation inspired by those
rlm@447 121 pixels. An imaginative system, having been trained on drinking and
rlm@447 122 non-drinking examples and learning that the most important
rlm@447 123 component of drinking is the feeling of water sliding down one's
rlm@447 124 throat, would analyze a video of a cat drinking in the following
rlm@447 125 manner:
rlm@441 126
rlm@447 127 1. Create a physical model of the video by putting a ``fuzzy''
rlm@447 128 model of its own body in place of the cat. Possibly also create
rlm@447 129 a simulation of the stream of water.
rlm@441 130
rlm@441 131 2. Play out this simulated scene and generate imagined sensory
rlm@441 132 experience. This will include relevant muscle contractions, a
rlm@441 133 close up view of the stream from the cat's perspective, and most
rlm@441 134 importantly, the imagined feeling of water entering the
rlm@443 135 mouth. The imagined sensory experience can come from a
rlm@441 136 simulation of the event, but can also be pattern-matched from
rlm@441 137 previous, similar embodied experience.
rlm@441 138
rlm@441 139 3. The action is now easily identified as drinking by the sense of
rlm@441 140 taste alone. The other senses (such as the tongue moving in and
rlm@441 141 out) help to give plausibility to the simulated action. Note that
rlm@441 142 the sense of vision, while critical in creating the simulation,
rlm@441 143 is not critical for identifying the action from the simulation.
rlm@441 144
rlm@441 145 For the chair examples, the process is even easier:
rlm@441 146
rlm@441 147 1. Align a model of your body to the person in the image.
rlm@441 148
rlm@441 149 2. Generate proprioceptive sensory data from this alignment.
rlm@437 150
rlm@441 151 3. Use the imagined proprioceptive data as a key to lookup related
rlm@441 152 sensory experience associated with that particular proproceptive
rlm@441 153 feeling.
rlm@437 154
rlm@443 155 4. Retrieve the feeling of your bottom resting on a surface, your
rlm@443 156 knees bent, and your leg muscles relaxed.
rlm@437 157
rlm@441 158 5. This sensory information is consistent with the =sitting?=
rlm@441 159 sensory predicate, so you (and the entity in the image) must be
rlm@441 160 sitting.
rlm@440 161
rlm@441 162 6. There must be a chair-like object since you are sitting.
rlm@440 163
rlm@441 164 Empathy offers yet another alternative to the age-old AI
rlm@441 165 representation question: ``What is a chair?'' --- A chair is the
rlm@441 166 feeling of sitting.
rlm@441 167
rlm@441 168 My program, =EMPATH= uses this empathic problem solving technique
rlm@441 169 to interpret the actions of a simple, worm-like creature.
rlm@437 170
rlm@441 171 #+caption: The worm performs many actions during free play such as
rlm@441 172 #+caption: curling, wiggling, and resting.
rlm@441 173 #+name: worm-intro
rlm@446 174 #+ATTR_LaTeX: :width 15cm
rlm@445 175 [[./images/worm-intro-white.png]]
rlm@437 176
rlm@447 177 #+caption: =EMPATH= recognized and classified each of these poses by
rlm@447 178 #+caption: inferring the complete sensory experience from
rlm@447 179 #+caption: proprioceptive data.
rlm@441 180 #+name: worm-recognition-intro
rlm@446 181 #+ATTR_LaTeX: :width 15cm
rlm@445 182 [[./images/worm-poses.png]]
rlm@441 183
rlm@441 184 One powerful advantage of empathic problem solving is that it
rlm@441 185 factors the action recognition problem into two easier problems. To
rlm@441 186 use empathy, you need an /aligner/, which takes the video and a
rlm@441 187 model of your body, and aligns the model with the video. Then, you
rlm@441 188 need a /recognizer/, which uses the aligned model to interpret the
rlm@441 189 action. The power in this method lies in the fact that you describe
rlm@448 190 all actions form a body-centered viewpoint. You are less tied to
rlm@447 191 the particulars of any visual representation of the actions. If you
rlm@441 192 teach the system what ``running'' is, and you have a good enough
rlm@441 193 aligner, the system will from then on be able to recognize running
rlm@441 194 from any point of view, even strange points of view like above or
rlm@441 195 underneath the runner. This is in contrast to action recognition
rlm@448 196 schemes that try to identify actions using a non-embodied approach.
rlm@448 197 If these systems learn about running as viewed from the side, they
rlm@448 198 will not automatically be able to recognize running from any other
rlm@448 199 viewpoint.
rlm@441 200
rlm@441 201 Another powerful advantage is that using the language of multiple
rlm@441 202 body-centered rich senses to describe body-centerd actions offers a
rlm@441 203 massive boost in descriptive capability. Consider how difficult it
rlm@441 204 would be to compose a set of HOG filters to describe the action of
rlm@447 205 a simple worm-creature ``curling'' so that its head touches its
rlm@447 206 tail, and then behold the simplicity of describing thus action in a
rlm@441 207 language designed for the task (listing \ref{grand-circle-intro}):
rlm@441 208
rlm@446 209 #+caption: Body-centerd actions are best expressed in a body-centered
rlm@446 210 #+caption: language. This code detects when the worm has curled into a
rlm@446 211 #+caption: full circle. Imagine how you would replicate this functionality
rlm@446 212 #+caption: using low-level pixel features such as HOG filters!
rlm@446 213 #+name: grand-circle-intro
rlm@446 214 #+begin_listing clojure
rlm@446 215 #+begin_src clojure
rlm@446 216 (defn grand-circle?
rlm@446 217 "Does the worm form a majestic circle (one end touching the other)?"
rlm@446 218 [experiences]
rlm@446 219 (and (curled? experiences)
rlm@446 220 (let [worm-touch (:touch (peek experiences))
rlm@446 221 tail-touch (worm-touch 0)
rlm@446 222 head-touch (worm-touch 4)]
rlm@446 223 (and (< 0.55 (contact worm-segment-bottom-tip tail-touch))
rlm@446 224 (< 0.55 (contact worm-segment-top-tip head-touch))))))
rlm@446 225 #+end_src
rlm@446 226 #+end_listing
rlm@446 227
rlm@435 228
rlm@449 229 ** =CORTEX= is a toolkit for building sensate creatures
rlm@435 230
rlm@448 231 I built =CORTEX= to be a general AI research platform for doing
rlm@448 232 experiments involving multiple rich senses and a wide variety and
rlm@448 233 number of creatures. I intend it to be useful as a library for many
rlm@448 234 more projects than just this one. =CORTEX= was necessary to meet a
rlm@448 235 need among AI researchers at CSAIL and beyond, which is that people
rlm@448 236 often will invent neat ideas that are best expressed in the
rlm@448 237 language of creatures and senses, but in order to explore those
rlm@448 238 ideas they must first build a platform in which they can create
rlm@448 239 simulated creatures with rich senses! There are many ideas that
rlm@448 240 would be simple to execute (such as =EMPATH=), but attached to them
rlm@448 241 is the multi-month effort to make a good creature simulator. Often,
rlm@448 242 that initial investment of time proves to be too much, and the
rlm@448 243 project must make do with a lesser environment.
rlm@435 244
rlm@448 245 =CORTEX= is well suited as an environment for embodied AI research
rlm@448 246 for three reasons:
rlm@448 247
rlm@448 248 - You can create new creatures using Blender, a popular 3D modeling
rlm@448 249 program. Each sense can be specified using special blender nodes
rlm@448 250 with biologically inspired paramaters. You need not write any
rlm@448 251 code to create a creature, and can use a wide library of
rlm@448 252 pre-existing blender models as a base for your own creatures.
rlm@448 253
rlm@448 254 - =CORTEX= implements a wide variety of senses, including touch,
rlm@448 255 proprioception, vision, hearing, and muscle tension. Complicated
rlm@448 256 senses like touch, and vision involve multiple sensory elements
rlm@448 257 embedded in a 2D surface. You have complete control over the
rlm@448 258 distribution of these sensor elements through the use of simple
rlm@448 259 png image files. In particular, =CORTEX= implements more
rlm@448 260 comprehensive hearing than any other creature simulation system
rlm@448 261 available.
rlm@448 262
rlm@448 263 - =CORTEX= supports any number of creatures and any number of
rlm@448 264 senses. Time in =CORTEX= dialates so that the simulated creatures
rlm@448 265 always precieve a perfectly smooth flow of time, regardless of
rlm@448 266 the actual computational load.
rlm@448 267
rlm@448 268 =CORTEX= is built on top of =jMonkeyEngine3=, which is a video game
rlm@448 269 engine designed to create cross-platform 3D desktop games. =CORTEX=
rlm@448 270 is mainly written in clojure, a dialect of =LISP= that runs on the
rlm@448 271 java virtual machine (JVM). The API for creating and simulating
rlm@449 272 creatures and senses is entirely expressed in clojure, though many
rlm@449 273 senses are implemented at the layer of jMonkeyEngine or below. For
rlm@449 274 example, for the sense of hearing I use a layer of clojure code on
rlm@449 275 top of a layer of java JNI bindings that drive a layer of =C++=
rlm@449 276 code which implements a modified version of =OpenAL= to support
rlm@449 277 multiple listeners. =CORTEX= is the only simulation environment
rlm@449 278 that I know of that can support multiple entities that can each
rlm@449 279 hear the world from their own perspective. Other senses also
rlm@449 280 require a small layer of Java code. =CORTEX= also uses =bullet=, a
rlm@449 281 physics simulator written in =C=.
rlm@448 282
rlm@448 283 #+caption: Here is the worm from above modeled in Blender, a free
rlm@448 284 #+caption: 3D-modeling program. Senses and joints are described
rlm@448 285 #+caption: using special nodes in Blender.
rlm@448 286 #+name: worm-recognition-intro
rlm@448 287 #+ATTR_LaTeX: :width 12cm
rlm@448 288 [[./images/blender-worm.png]]
rlm@448 289
rlm@449 290 Here are some thing I anticipate that =CORTEX= might be used for:
rlm@449 291
rlm@449 292 - exploring new ideas about sensory integration
rlm@449 293 - distributed communication among swarm creatures
rlm@449 294 - self-learning using free exploration,
rlm@449 295 - evolutionary algorithms involving creature construction
rlm@449 296 - exploration of exoitic senses and effectors that are not possible
rlm@449 297 in the real world (such as telekenisis or a semantic sense)
rlm@449 298 - imagination using subworlds
rlm@449 299
rlm@448 300 During one test with =CORTEX=, I created 3,000 entities each with
rlm@448 301 their own independent senses and ran them all at only 1/80 real
rlm@448 302 time. In another test, I created a detailed model of my own hand,
rlm@448 303 equipped with a realistic distribution of touch (more sensitive at
rlm@448 304 the fingertips), as well as eyes and ears, and it ran at around 1/4
rlm@449 305 real time.
rlm@448 306
rlm@449 307 #+BEGIN_LaTeX
rlm@449 308 \begin{sidewaysfigure}
rlm@449 309 \includegraphics[width=9.5in]{images/full-hand.png}
rlm@449 310 \caption{Here is the worm from above modeled in Blender,
rlm@449 311 a free 3D-modeling program. Senses and joints are described
rlm@449 312 using special nodes in Blender. The senses are displayed on
rlm@449 313 the right, and the simulation is displayed on the left. Notice
rlm@449 314 that the hand is curling its fingers, that it can see its own
rlm@449 315 finger from the eye in its palm, and thta it can feel its own
rlm@449 316 thumb touching its palm.}
rlm@449 317 \end{sidewaysfigure}
rlm@449 318 #+END_LaTeX
rlm@448 319
rlm@437 320 ** Contributions
rlm@435 321
rlm@449 322 I built =CORTEX=, a comprehensive platform for embodied AI
rlm@449 323 experiments. =CORTEX= many new features lacking in other systems,
rlm@449 324 such as sound. It is easy to create new creatures using Blender, a
rlm@449 325 free 3D modeling program.
rlm@449 326
rlm@449 327 I built =EMPATH=, which uses =CORTEX= to identify the actions of a
rlm@449 328 worm-like creature using a computational model of empathy.
rlm@449 329
rlm@436 330 * Building =CORTEX=
rlm@435 331
rlm@436 332 ** To explore embodiment, we need a world, body, and senses
rlm@435 333
rlm@436 334 ** Because of Time, simulation is perferable to reality
rlm@435 335
rlm@436 336 ** Video game engines are a great starting point
rlm@435 337
rlm@436 338 ** Bodies are composed of segments connected by joints
rlm@435 339
rlm@436 340 ** Eyes reuse standard video game components
rlm@436 341
rlm@436 342 ** Hearing is hard; =CORTEX= does it right
rlm@436 343
rlm@436 344 ** Touch uses hundreds of hair-like elements
rlm@436 345
rlm@440 346 ** Proprioception is the sense that makes everything ``real''
rlm@436 347
rlm@436 348 ** Muscles are both effectors and sensors
rlm@436 349
rlm@436 350 ** =CORTEX= brings complex creatures to life!
rlm@436 351
rlm@436 352 ** =CORTEX= enables many possiblities for further research
rlm@435 353
rlm@435 354 * Empathy in a simulated worm
rlm@435 355
rlm@449 356 Here I develop a computational model of empathy, using =CORTEX= as a
rlm@449 357 base. Empathy in this context is the ability to observe another
rlm@449 358 creature and infer what sorts of sensations that creature is
rlm@449 359 feeling. My empathy algorithm involves multiple phases. First is
rlm@449 360 free-play, where the creature moves around and gains sensory
rlm@449 361 experience. From this experience I construct a representation of the
rlm@449 362 creature's sensory state space, which I call \Phi-space. Using
rlm@449 363 \Phi-space, I construct an efficient function which takes the
rlm@449 364 limited data that comes from observing another creature and enriches
rlm@449 365 it full compliment of imagined sensory data. I can then use the
rlm@449 366 imagined sensory data to recognize what the observed creature is
rlm@449 367 doing and feeling, using straightforward embodied action predicates.
rlm@449 368 This is all demonstrated with using a simple worm-like creature, and
rlm@449 369 recognizing worm-actions based on limited data.
rlm@449 370
rlm@449 371 #+caption: Here is the worm with which we will be working.
rlm@449 372 #+caption: It is composed of 5 segments. Each segment has a
rlm@449 373 #+caption: pair of extensor and flexor muscles. Each of the
rlm@449 374 #+caption: worm's four joints is a hinge joint which allows
rlm@449 375 #+caption: 30 degrees of rotation to either side. Each segment
rlm@449 376 #+caption: of the worm is touch-capable and has a uniform
rlm@449 377 #+caption: distribution of touch sensors on each of its faces.
rlm@449 378 #+caption: Each joint has a proprioceptive sense to detect
rlm@449 379 #+caption: relative positions. The worm segments are all the
rlm@449 380 #+caption: same except for the first one, which has a much
rlm@449 381 #+caption: higher weight than the others to allow for easy
rlm@449 382 #+caption: manual motor control.
rlm@449 383 #+name: basic-worm-view
rlm@449 384 #+ATTR_LaTeX: :width 10cm
rlm@449 385 [[./images/basic-worm-view.png]]
rlm@449 386
rlm@449 387 #+caption: Program for reading a worm from a blender file and
rlm@449 388 #+caption: outfitting it with the senses of proprioception,
rlm@449 389 #+caption: touch, and the ability to move, as specified in the
rlm@449 390 #+caption: blender file.
rlm@449 391 #+name: get-worm
rlm@449 392 #+begin_listing clojure
rlm@449 393 #+begin_src clojure
rlm@449 394 (defn worm []
rlm@449 395 (let [model (load-blender-model "Models/worm/worm.blend")]
rlm@449 396 {:body (doto model (body!))
rlm@449 397 :touch (touch! model)
rlm@449 398 :proprioception (proprioception! model)
rlm@449 399 :muscles (movement! model)}))
rlm@449 400 #+end_src
rlm@449 401 #+end_listing
rlm@449 402
rlm@436 403 ** Embodiment factors action recognition into managable parts
rlm@435 404
rlm@449 405 Using empathy, I divide the problem of action recognition into a
rlm@449 406 recognition process expressed in the language of a full compliment
rlm@449 407 of senses, and an imaganitive process that generates full sensory
rlm@449 408 data from partial sensory data. Splitting the action recognition
rlm@449 409 problem in this manner greatly reduces the total amount of work to
rlm@449 410 recognize actions: The imaganitive process is mostly just matching
rlm@449 411 previous experience, and the recognition process gets to use all
rlm@449 412 the senses to directly describe any action.
rlm@449 413
rlm@436 414 ** Action recognition is easy with a full gamut of senses
rlm@435 415
rlm@449 416 Embodied representations using multiple senses such as touch,
rlm@449 417 proprioception, and muscle tension turns out be be exceedingly
rlm@449 418 efficient at describing body-centered actions. It is the ``right
rlm@449 419 language for the job''. For example, it takes only around 5 lines
rlm@449 420 of LISP code to describe the action of ``curling'' using embodied
rlm@449 421 primitives. It takes about 8 lines to describe the seemingly
rlm@449 422 complicated action of wiggling.
rlm@449 423
rlm@449 424 The following action predicates each take a stream of sensory
rlm@449 425 experience, observe however much of it they desire, and decide
rlm@449 426 whether the worm is doing the action they describe. =curled?=
rlm@449 427 relies on proprioception, =resting?= relies on touch, =wiggling?=
rlm@449 428 relies on a fourier analysis of muscle contraction, and
rlm@449 429 =grand-circle?= relies on touch and reuses =curled?= as a gaurd.
rlm@449 430
rlm@449 431 #+caption: Program for detecting whether the worm is curled. This is the
rlm@449 432 #+caption: simplest action predicate, because it only uses the last frame
rlm@449 433 #+caption: of sensory experience, and only uses proprioceptive data. Even
rlm@449 434 #+caption: this simple predicate, however, is automatically frame
rlm@449 435 #+caption: independent and ignores vermopomorphic differences such as
rlm@449 436 #+caption: worm textures and colors.
rlm@449 437 #+name: curled
rlm@449 438 #+begin_listing clojure
rlm@449 439 #+begin_src clojure
rlm@449 440 (defn curled?
rlm@449 441 "Is the worm curled up?"
rlm@449 442 [experiences]
rlm@449 443 (every?
rlm@449 444 (fn [[_ _ bend]]
rlm@449 445 (> (Math/sin bend) 0.64))
rlm@449 446 (:proprioception (peek experiences))))
rlm@449 447 #+end_src
rlm@449 448 #+end_listing
rlm@449 449
rlm@449 450 #+caption: Program for summarizing the touch information in a patch
rlm@449 451 #+caption: of skin.
rlm@449 452 #+name: touch-summary
rlm@449 453 #+begin_listing clojure
rlm@449 454 #+begin_src clojure
rlm@449 455 (defn contact
rlm@449 456 "Determine how much contact a particular worm segment has with
rlm@449 457 other objects. Returns a value between 0 and 1, where 1 is full
rlm@449 458 contact and 0 is no contact."
rlm@449 459 [touch-region [coords contact :as touch]]
rlm@449 460 (-> (zipmap coords contact)
rlm@449 461 (select-keys touch-region)
rlm@449 462 (vals)
rlm@449 463 (#(map first %))
rlm@449 464 (average)
rlm@449 465 (* 10)
rlm@449 466 (- 1)
rlm@449 467 (Math/abs)))
rlm@449 468 #+end_src
rlm@449 469 #+end_listing
rlm@449 470
rlm@449 471
rlm@449 472 #+caption: Program for detecting whether the worm is at rest. This program
rlm@449 473 #+caption: uses a summary of the tactile information from the underbelly
rlm@449 474 #+caption: of the worm, and is only true if every segment is touching the
rlm@449 475 #+caption: floor. Note that this function contains no references to
rlm@449 476 #+caption: proprioction at all.
rlm@449 477 #+name: resting
rlm@449 478 #+begin_listing clojure
rlm@449 479 #+begin_src clojure
rlm@449 480 (def worm-segment-bottom (rect-region [8 15] [14 22]))
rlm@449 481
rlm@449 482 (defn resting?
rlm@449 483 "Is the worm resting on the ground?"
rlm@449 484 [experiences]
rlm@449 485 (every?
rlm@449 486 (fn [touch-data]
rlm@449 487 (< 0.9 (contact worm-segment-bottom touch-data)))
rlm@449 488 (:touch (peek experiences))))
rlm@449 489 #+end_src
rlm@449 490 #+end_listing
rlm@449 491
rlm@449 492 #+caption: Program for detecting whether the worm is curled up into a
rlm@449 493 #+caption: full circle. Here the embodied approach begins to shine, as
rlm@449 494 #+caption: I am able to both use a previous action predicate (=curled?=)
rlm@449 495 #+caption: as well as the direct tactile experience of the head and tail.
rlm@449 496 #+name: grand-circle
rlm@449 497 #+begin_listing clojure
rlm@449 498 #+begin_src clojure
rlm@449 499 (def worm-segment-bottom-tip (rect-region [15 15] [22 22]))
rlm@449 500
rlm@449 501 (def worm-segment-top-tip (rect-region [0 15] [7 22]))
rlm@449 502
rlm@449 503 (defn grand-circle?
rlm@449 504 "Does the worm form a majestic circle (one end touching the other)?"
rlm@449 505 [experiences]
rlm@449 506 (and (curled? experiences)
rlm@449 507 (let [worm-touch (:touch (peek experiences))
rlm@449 508 tail-touch (worm-touch 0)
rlm@449 509 head-touch (worm-touch 4)]
rlm@449 510 (and (< 0.55 (contact worm-segment-bottom-tip tail-touch))
rlm@449 511 (< 0.55 (contact worm-segment-top-tip head-touch))))))
rlm@449 512 #+end_src
rlm@449 513 #+end_listing
rlm@449 514
rlm@449 515
rlm@449 516 #+caption: Program for detecting whether the worm has been wiggling for
rlm@449 517 #+caption: the last few frames. It uses a fourier analysis of the muscle
rlm@449 518 #+caption: contractions of the worm's tail to determine wiggling. This is
rlm@449 519 #+caption: signigicant because there is no particular frame that clearly
rlm@449 520 #+caption: indicates that the worm is wiggling --- only when multiple frames
rlm@449 521 #+caption: are analyzed together is the wiggling revealed. Defining
rlm@449 522 #+caption: wiggling this way also gives the worm an opportunity to learn
rlm@449 523 #+caption: and recognize ``frustrated wiggling'', where the worm tries to
rlm@449 524 #+caption: wiggle but can't. Frustrated wiggling is very visually different
rlm@449 525 #+caption: from actual wiggling, but this definition gives it to us for free.
rlm@449 526 #+name: wiggling
rlm@449 527 #+begin_listing clojure
rlm@449 528 #+begin_src clojure
rlm@449 529 (defn fft [nums]
rlm@449 530 (map
rlm@449 531 #(.getReal %)
rlm@449 532 (.transform
rlm@449 533 (FastFourierTransformer. DftNormalization/STANDARD)
rlm@449 534 (double-array nums) TransformType/FORWARD)))
rlm@449 535
rlm@449 536 (def indexed (partial map-indexed vector))
rlm@449 537
rlm@449 538 (defn max-indexed [s]
rlm@449 539 (first (sort-by (comp - second) (indexed s))))
rlm@449 540
rlm@449 541 (defn wiggling?
rlm@449 542 "Is the worm wiggling?"
rlm@449 543 [experiences]
rlm@449 544 (let [analysis-interval 0x40]
rlm@449 545 (when (> (count experiences) analysis-interval)
rlm@449 546 (let [a-flex 3
rlm@449 547 a-ex 2
rlm@449 548 muscle-activity
rlm@449 549 (map :muscle (vector:last-n experiences analysis-interval))
rlm@449 550 base-activity
rlm@449 551 (map #(- (% a-flex) (% a-ex)) muscle-activity)]
rlm@449 552 (= 2
rlm@449 553 (first
rlm@449 554 (max-indexed
rlm@449 555 (map #(Math/abs %)
rlm@449 556 (take 20 (fft base-activity))))))))))
rlm@449 557 #+end_src
rlm@449 558 #+end_listing
rlm@449 559
rlm@449 560 With these action predicates, I can now recognize the actions of
rlm@449 561 the worm while it is moving under my control and I have access to
rlm@449 562 all the worm's senses.
rlm@449 563
rlm@449 564 #+caption: Use the action predicates defined earlier to report on
rlm@449 565 #+caption: what the worm is doing while in simulation.
rlm@449 566 #+name: report-worm-activity
rlm@449 567 #+begin_listing clojure
rlm@449 568 #+begin_src clojure
rlm@449 569 (defn debug-experience
rlm@449 570 [experiences text]
rlm@449 571 (cond
rlm@449 572 (grand-circle? experiences) (.setText text "Grand Circle")
rlm@449 573 (curled? experiences) (.setText text "Curled")
rlm@449 574 (wiggling? experiences) (.setText text "Wiggling")
rlm@449 575 (resting? experiences) (.setText text "Resting")))
rlm@449 576 #+end_src
rlm@449 577 #+end_listing
rlm@449 578
rlm@449 579 #+caption: Using =debug-experience=, the body-centered predicates
rlm@449 580 #+caption: work together to classify the behaviour of the worm.
rlm@449 581 #+caption: while under manual motor control.
rlm@449 582 #+name: basic-worm-view
rlm@449 583 #+ATTR_LaTeX: :width 10cm
rlm@449 584 [[./images/worm-identify-init.png]]
rlm@449 585
rlm@449 586 These action predicates satisfy the recognition requirement of an
rlm@449 587 empathic recognition system. There is a lot of power in the
rlm@449 588 simplicity of the action predicates. They describe their actions
rlm@449 589 without getting confused in visual details of the worm. Each one is
rlm@449 590 frame independent, but more than that, they are each indepent of
rlm@449 591 irrelevant visual details of the worm and the environment. They
rlm@449 592 will work regardless of whether the worm is a different color or
rlm@449 593 hevaily textured, or of the environment has strange lighting.
rlm@449 594
rlm@449 595 The trick now is to make the action predicates work even when the
rlm@449 596 sensory data on which they depend is absent. If I can do that, then
rlm@449 597 I will have gained much,
rlm@435 598
rlm@436 599 ** \Phi-space describes the worm's experiences
rlm@449 600
rlm@449 601 As a first step towards building empathy, I need to gather all of
rlm@449 602 the worm's experiences during free play. I use a simple vector to
rlm@449 603 store all the experiences.
rlm@449 604
rlm@449 605 #+caption: Program to gather the worm's experiences into a vector for
rlm@449 606 #+caption: further processing. The =motor-control-program= line uses
rlm@449 607 #+caption: a motor control script that causes the worm to execute a series
rlm@449 608 #+caption: of ``exercices'' that include all the action predicates.
rlm@449 609 #+name: generate-phi-space
rlm@449 610 #+begin_listing clojure
rlm@449 611 #+begin_src clojure
rlm@449 612 (defn generate-phi-space []
rlm@449 613 (let [experiences (atom [])]
rlm@449 614 (run-world
rlm@449 615 (apply-map
rlm@449 616 worm-world
rlm@449 617 (merge
rlm@449 618 (worm-world-defaults)
rlm@449 619 {:end-frame 700
rlm@449 620 :motor-control
rlm@449 621 (motor-control-program worm-muscle-labels do-all-the-things)
rlm@449 622 :experiences experiences})))
rlm@449 623 @experiences))
rlm@449 624 #+end_src
rlm@449 625 #+end_listing
rlm@449 626
rlm@449 627 Each element of the experience vector exists in the vast space of
rlm@449 628 all possible worm-experiences. Most of this vast space is actually
rlm@449 629 unreachable due to physical constraints of the worm's body. For
rlm@449 630 example, the worm's segments are connected by hinge joints that put
rlm@449 631 a practical limit on the worm's degrees of freedom. Also, the worm
rlm@449 632 can not be bent into a circle so that its ends are touching and at
rlm@449 633 the same time not also experience the sensation of touching itself.
rlm@449 634
rlm@449 635 As the worm moves around during free play and the vector grows
rlm@449 636 larger, the vector begins to define a subspace which is all the
rlm@449 637 practical experiences the worm can experience during normal
rlm@449 638 operation, which I call \Phi-space, short for physical-space. The
rlm@449 639 vector defines a path through \Phi-space. This path has interesting
rlm@449 640 properties that all derive from embodiment. The proprioceptive
rlm@449 641 components are completely smooth, because in order for the worm to
rlm@449 642 move from one position to another, it must pass through the
rlm@449 643 intermediate positions. The path invariably forms loops as actions
rlm@449 644 are repeated. Finally and most importantly, proprioception actually
rlm@449 645 gives very strong inference about the other senses. For example,
rlm@449 646 when the worm is flat, you can infer that it is touching the ground
rlm@449 647 and that its muscles are not active, because if the muscles were
rlm@449 648 active, the worm would be moving and would not be perfectly flat.
rlm@449 649 In order to stay flat, the worm has to be touching the ground, or
rlm@449 650 it would again be moving out of the flat position due to gravity.
rlm@449 651 If the worm is positioned in such a way that it interacts with
rlm@449 652 itself, then it is very likely to be feeling the same tactile
rlm@449 653 feelings as the last time it was in that position, because it has
rlm@449 654 the same body as then. If you observe multiple frames of
rlm@449 655 proprioceptive data, then you can become increasingly confident
rlm@449 656 about the exact activations of the worm's muscles, because it
rlm@449 657 generally takes a unique combination of muscle contractions to
rlm@449 658 transform the worm's body along a specific path through \Phi-space.
rlm@449 659
rlm@449 660 There is a simple way of taking \Phi-space and the total ordering
rlm@449 661 provided by an experience vector and reliably infering the rest of
rlm@449 662 the senses.
rlm@435 663
rlm@436 664 ** Empathy is the process of tracing though \Phi-space
rlm@449 665
rlm@450 666 Here is the core of a basic empathy algorithm, starting with an
rlm@450 667 experience vector: First, group the experiences into tiered
rlm@450 668 proprioceptive bins. I use powers of 10 and 3 bins, and the
rlm@450 669 smallest bin has and approximate size of 0.001 radians in all
rlm@450 670 proprioceptive dimensions.
rlm@450 671
rlm@450 672 Then, given a sequence of proprioceptive input, generate a set of
rlm@450 673 matching experience records for each input.
rlm@449 674
rlm@450 675 Finally, to infer sensory data, select the longest consective chain
rlm@450 676 of experiences as determined by the indexes into the experience
rlm@450 677 vector.
rlm@449 678
rlm@450 679 This algorithm has three advantages:
rlm@450 680
rlm@450 681 1. It's simple
rlm@450 682
rlm@450 683 3. It's very fast -- both tracing through possibilites and
rlm@450 684 retrieving possible interpretations take essentially constant
rlm@450 685 time.
rlm@450 686
rlm@450 687 2. It protects from wrong interpretations of transient ambiguous
rlm@450 688 proprioceptive data : for example, if the worm is flat for just
rlm@450 689 an instant, this flattness will not be interpreted as implying
rlm@450 690 that the worm has its muscles relaxed, since the flattness is
rlm@450 691 part of a longer chain which includes a distinct pattern of
rlm@450 692 muscle activation. A memoryless statistical model such as a
rlm@450 693 markov model that operates on individual frames may very well
rlm@450 694 make this mistake.
rlm@450 695
rlm@450 696 #+caption: Program to convert an experience vector into a
rlm@450 697 #+caption: proprioceptively binned lookup function.
rlm@450 698 #+name: bin
rlm@450 699 #+begin_listing clojure
rlm@450 700 #+begin_src clojure
rlm@449 701 (defn bin [digits]
rlm@449 702 (fn [angles]
rlm@449 703 (->> angles
rlm@449 704 (flatten)
rlm@449 705 (map (juxt #(Math/sin %) #(Math/cos %)))
rlm@449 706 (flatten)
rlm@449 707 (mapv #(Math/round (* % (Math/pow 10 (dec digits))))))))
rlm@449 708
rlm@449 709 (defn gen-phi-scan
rlm@450 710 "Nearest-neighbors with binning. Only returns a result if
rlm@450 711 the propriceptive data is within 10% of a previously recorded
rlm@450 712 result in all dimensions."
rlm@450 713 [phi-space]
rlm@449 714 (let [bin-keys (map bin [3 2 1])
rlm@449 715 bin-maps
rlm@449 716 (map (fn [bin-key]
rlm@449 717 (group-by
rlm@449 718 (comp bin-key :proprioception phi-space)
rlm@449 719 (range (count phi-space)))) bin-keys)
rlm@449 720 lookups (map (fn [bin-key bin-map]
rlm@450 721 (fn [proprio] (bin-map (bin-key proprio))))
rlm@450 722 bin-keys bin-maps)]
rlm@449 723 (fn lookup [proprio-data]
rlm@449 724 (set (some #(% proprio-data) lookups)))))
rlm@450 725 #+end_src
rlm@450 726 #+end_listing
rlm@449 727
rlm@449 728
rlm@450 729 #+caption: Program to calculate empathy by tracing though \Phi-space
rlm@450 730 #+caption: and finding the longest (ie. most coherent) interpretation
rlm@450 731 #+caption: of the data.
rlm@450 732 #+name: longest-thread
rlm@450 733 #+begin_listing clojure
rlm@450 734 #+begin_src clojure
rlm@449 735 (defn longest-thread
rlm@449 736 "Find the longest thread from phi-index-sets. The index sets should
rlm@449 737 be ordered from most recent to least recent."
rlm@449 738 [phi-index-sets]
rlm@449 739 (loop [result '()
rlm@449 740 [thread-bases & remaining :as phi-index-sets] phi-index-sets]
rlm@449 741 (if (empty? phi-index-sets)
rlm@449 742 (vec result)
rlm@449 743 (let [threads
rlm@449 744 (for [thread-base thread-bases]
rlm@449 745 (loop [thread (list thread-base)
rlm@449 746 remaining remaining]
rlm@449 747 (let [next-index (dec (first thread))]
rlm@449 748 (cond (empty? remaining) thread
rlm@449 749 (contains? (first remaining) next-index)
rlm@449 750 (recur
rlm@449 751 (cons next-index thread) (rest remaining))
rlm@449 752 :else thread))))
rlm@449 753 longest-thread
rlm@449 754 (reduce (fn [thread-a thread-b]
rlm@449 755 (if (> (count thread-a) (count thread-b))
rlm@449 756 thread-a thread-b))
rlm@449 757 '(nil)
rlm@449 758 threads)]
rlm@449 759 (recur (concat longest-thread result)
rlm@449 760 (drop (count longest-thread) phi-index-sets))))))
rlm@450 761 #+end_src
rlm@450 762 #+end_listing
rlm@450 763
rlm@449 764
rlm@449 765 There is one final piece, which is to replace missing sensory data
rlm@449 766 with a best-guess estimate. While I could fill in missing data by
rlm@449 767 using a gradient over the closest known sensory data points, averages
rlm@449 768 can be misleading. It is certainly possible to create an impossible
rlm@449 769 sensory state by averaging two possible sensory states. Therefore, I
rlm@449 770 simply replicate the most recent sensory experience to fill in the
rlm@449 771 gaps.
rlm@449 772
rlm@449 773 #+caption: Fill in blanks in sensory experience by replicating the most
rlm@449 774 #+caption: recent experience.
rlm@449 775 #+name: infer-nils
rlm@449 776 #+begin_listing clojure
rlm@449 777 #+begin_src clojure
rlm@449 778 (defn infer-nils
rlm@449 779 "Replace nils with the next available non-nil element in the
rlm@449 780 sequence, or barring that, 0."
rlm@449 781 [s]
rlm@449 782 (loop [i (dec (count s))
rlm@449 783 v (transient s)]
rlm@449 784 (if (zero? i) (persistent! v)
rlm@449 785 (if-let [cur (v i)]
rlm@449 786 (if (get v (dec i) 0)
rlm@449 787 (recur (dec i) v)
rlm@449 788 (recur (dec i) (assoc! v (dec i) cur)))
rlm@449 789 (recur i (assoc! v i 0))))))
rlm@449 790 #+end_src
rlm@449 791 #+end_listing
rlm@449 792
rlm@435 793
rlm@441 794 ** Efficient action recognition with =EMPATH=
rlm@425 795
rlm@450 796 In my exploration with the worm, I can generally infer actions from
rlm@450 797 proprioceptive data exactly as well as when I have the complete
rlm@450 798 sensory data. To reach this level, I have to train the worm with
rlm@450 799 verious exercices for about 1 minute.
rlm@450 800
rlm@449 801 ** Digression: bootstrapping touch using free exploration
rlm@449 802
rlm@432 803 * Contributions
rlm@447 804
rlm@447 805
rlm@447 806
rlm@447 807
rlm@447 808 # An anatomical joke:
rlm@447 809 # - Training
rlm@447 810 # - Skeletal imitation
rlm@447 811 # - Sensory fleshing-out
rlm@447 812 # - Classification