Neural Visuomotor Controller for a Simulated Salamander Robot

1
Neural Visuomotor Controller for a Simulated
Salamander Robot

• Biljana Petreska
• Diploma Thesis March 2004
• Responsible
• Prof. Auke Jan Ijspeert

2
Goals of the Project

• Investigate through simulations tightly coupled
  with neurobiological data, the neural mechanisms
  underlying visually guided behaviour in
  amphibians
• Implement a closed-loop with the environment onto
  the existing neuromechanical simulation developed
  by Ijspeert, by adding biologically inspired
  models for parts of the salamander brain
• Develop a controller that accounts for
  observations in feeding behaviour, including prey
  localization and prey recognition
• Study a model proposed by Ijspeert of structured
  mapping between the optic tectum (primary visual
  processing center) and the brain stem (motor
  centers) as a solution to the visuomotor
  coordination

3
Interests

• Relevant for perceptual robotics
• decoding the brain processes, assigning meaning
  to complex patterns of sensor stimuli may lead to
  the solution of many robotics tasks
• Test bed for probing neurobiological
  contributions
• ideal for the validation or refutation of new
  theories

4
Overview

• Short introduction on relevant topics and
  previous works
• Implemented Models
• Respective Results
• Conclusion and Future Work

5
Everything youve always wanted to know on
Salamanders

• Amphibians
• Great variety of species (3924 indexed so far),
  sizes (from 16mm to 1.5m), aspects and lifestyles
  (terrestrial and/or aquatic).
• A relatively simple neural circuitry that
  presents all main vertebrate features
• Tractable from an experimental point of view an
  important amount of behavioral, biological and
  neurological data exists

6
Visually Guided Behavior

• Vision is by far the most important feeding
  guiding sense. Under good visual conditions the
  other signals such as olfactory are overridden
• Feeding strategies (some species can switch from
  one to another)
• hunter strategy active search for prey.
  Prerequisites are a short massive tongue and poor
  visual capacities.
• ambush strategy wait until prey comes close.
  Prerequisites are a highly specialized projectile
  tongue (up to 80 body length), evolved visual
  system and frontally oriented eyes

7
Visually Guided Behaviour

• Sequence of feeding behavior
• orienting
• approach
• olfaction tests
• gaze stabilization
• snapping
• Prey preferences (in order of importance)
• stimulus size
• stimulus velocity
• stimulus-background contrast
• stimulus shape
• movement pattern
• experience-dependant

8
Morphology of the Salamander Brain

• Functional differentiation of the brain
  structurally different regions accomplish
  different tasks
• Global top to bottom visual information
  processing
• Principal components photoreceptors, retina,
  optic tectum, nucleus isthmi, pretectum,thalamus,
  medulla oblongata and brain stem

9
Retinal Ganglion Cells

• First layer of visual processing, transfers
  visual signals to the brain via the optic nerve
• 3 Types of retinal ganglion cells that project to
  particular layers in the optic tectum

10
Optic Tectum

• Main visual processing center. Integrates also
  multimodal perception, such as ascending
  somatosensory, auditory, olfaction and vestibular
• Stratification in 9 layers, first three are
  retinal afferents
• Six morphological neuron types identified (one
  interneuron)
• Topographic representation of the visual field
• Viewed as a set of partial overlapping maps due
  to the different types of tectal projection
  neurons
• Number of tectal cells in Hydromantes Italicus
  92 000 and 3300 out of 5000 projection neurons
  are descending
• Projection patterns from and to the retina,
  pretectum, thalamus, nucleus isthmi and medulla
  (reaching the spinal cord)

Distribution and Receptive Field Sizes of Tectal
Neurons in H.Italicus
11
Pretectum

• Has been ascribed a role in optokinetic
  nystagmus, figure-background discrimination,
  pupillary reflex, fixation, phototaxis, and
  prey-enemy distinction.
• Properties of pretectal neurons
• Homogenous arborisation
• (no classification was possible)
• Divergent projections
• (including to the tectum and spinal cord)
• Large receptive fields
• Receive direct and indirect (from tectum) retinal
  input
• Direction-sensitive neurons (predominantly in
  temporonasal direction)
• Respond to stimuli in the contralateral visual
  field

12
Lesion Experiments

• Give insight of the function of the destroyed
  brain region
• Lesion of the optic tectum both visual
  prey-catching and predator avoidance fail to
  occur. Local lesions produce scotoma, total
  blindness for a part of the visual field
  corresponding to the size of the lesion.
• Lesion of the pretectum locally facilitates
  feeding and abolishes prey-predator
  discrimination, attack everything that moves
  including their own extremities and threatening
  stimuli
• Lesion of the thalamus unable to avoid collision
  to a vertically stripped barrier, affects the
  binocular field
• Lesion of the medulla oblongata affects
  distance, elevation or horizontal eccentricity
  estimates, overshoots prey or snaps only in
  frontal positions gt different components of the
  stimulus position are handled through different
  pathways
• Difficulty in some cases the animal recovers
  shortly after the lesion and the relative
  precision of lesions may induce errors

13
Previous Works

• Based upon the principle of coarse coding (Eurich
  et al, 1997)
• Motivation the high sensory resolution observed
  in nature seems incompatible with the large size
  of receptive fields of tectal neurons
• Definition population-coding using mapping
  combinatorics of intersecting receptive fields
• A non-firing neuron conveys as much information
  as a firing neuron. All neurons participate at
  the information coding
• Weakness likely to suffer from metamery
  (convergence of information channels)
• Simulander I
• Feedforward network with only 100 neurons,
  trained by an evolution strategy for the specific
  task of head orienting (implies prey
  localization)
• Distribution and sizes of receptive fields of
  tectal neurons are respected and firing rates
  have been adapted
• Unstructured mapping follows the prey with high
  accuracy
• Simulander II
• Similar to Simulander I, but trained for the
  specific task of frontal tongue projection
  (implies depth perception)

14
Addressed Questions

• How can the stimulus location and depth estimates
  be extracted from the tectum maps?
• What sensorimotor transformations occur at the
  level of the optic tectum, the brainstem and the
  pathways between them? Can a structured mapping
  provide an accurate visual tracking?
• Which type of a tectum-brainstem mapping explains
  the typical curved approach in monocularized
  salamanders?
• How is the visual perception influenced by head
  motion during the approach toward a stimulus? Are
  additional mechanisms necessary for dealing with
  the remaining shifts in the visual background?
• Which mechanism implements the release of the
  snapping behavior? And how is the tongue
  controlled?

15
Neural Networks

• Restrictions (performance motivated)
• Uniformly distributed neuron units
• Square receptive fields
• Specification
• Center receptive field (in degrees of visual
  field) determines the size of the neural
  network
• Surround receptive field (in degrees of visual
  field) determines the overlap and redundancy
  feature
• Weights matrix, activation function and
  thresholds
• Features
• Reduction (biologically motivated)
• Visualization (extremely practical)

16
Eyes of the Simulated Salamander

• Virtual cameras extract views using provided
  OpenGL functions
• Correct the view using a spherical projection
• Photoreceptors are equivalent to pixel grey
  values
• Scalable visual field

17
Retinal Ganglion Cells of Type 1

• Properties
• small size excitatory (2-3) and strong
  inhibitory (12-16) receptive fields
• no response to change in light
• involved in local contrast calculation gt edge
  detector
• project to the contralateral spinal cord gt
  obstacle avoidance?
• give rise to a fine grained representation of the
  visual field in the retina
• Modelled with the laplacian of a gaussian filter
• Classic edge detector in computer vision and
  confirmed by the study of a larval tiger
  salamander retina receptive field

18
Retinal Ganglion Cells of Type 2

• Type 2 retinal ganglion cells respond only to
  moving objects gt motion detectors
• Detection of change compare the corresponding
  pixels at different times, using a linear
  difference function
• where t is a threshold, j and k are moments in
  time, x and y are the pixel positions in the
  frame
• Biological inspirations
• Reflects signals with delayed pathways that give
  rise to a simultaneous representation of the same
  object at different times in the brain
• Flat weights the activity sharply increases when
  an object enters the receptive field variation
• Tectal neurons are contrast-sensitive linear
  function

19
Retinal Ganglion Cells of Type 3

• Properties
• large receptive fields (10-20)
• tonic response to change in light intensity
• respond to overall luminosity (dimming detectors)
  
• respond also at low contrast and velocity
• Model
• flat weights
• simple summing network
• Predator detectors among other

20
Optic Tectum Model

• Principal biological inspirations
• Retinotopic map in the optic tectum electrical
  stimulations result in turning movements that
  roughly correspond to this map
• Only two synapses between the retina and the
  brain stem the tectum directly projects onto the
  brain stem
• Input retinal ganglion cells of type 2. Motion
  is a necessary prerequisite for a stimulus to be
  interpreted as prey
• Structured mapping different strengths along the
  rostro-caudal axis, reflects the stimulus
  eccentricity
• Motoneuron activation function (integrating
  weighted tectal activity)
• where x is the change in light intensity of the
  pixel at positions i and j
• Linear weights function
• where a and ß are parameters

21
Optic Tectum Model II

• Version with ipsilateral input
• (contribution from both eyes)

22
Optic Tectum Model III

• Normalizing tectal activity. The modified model
  is robust to changes in the stimulus parameters
  and visual scene
• Biological reference
• TO4 neurons, arborize in RGC2
• TO2 neurons, large receptive fields
• both project to the nucleus isthmi

23
Pretectum Model

• Large stimuli based on RGC3 gt dimming detectors
  with three times larger receptive fields
• Motion compare direct and indirect (via tectum)
  RGC3 responses
• Direction-sensitive neurons
• Why temporonasal sensitivity?
• Based upon separating the ON and OFF channels
• Hypothesis only sensitive to dark objects
  (biologically consistent)

24
Snapping Model

• Relevant for depth estimation
• Tongue mechanism (biologically consistent) 4
  muscles, protraction and retraction times
  modulated by the stimulus position
• We proposed a mechanism for frontal snapping
  based upon divergent projections of the tectal
  neurons

25
Results

• Find optimal a and ß parameters of the linear
  weights function through an exhaustive search of
  the parameter space
• Cost function difference between the stimulus
  direction and the salamander orienting movement
• Experimental conditions
• Ewert experiment the stimulus is moved on a
  semi-circular trajectory with a constant speed in
  front of the animal
• (task of head orienting)
• Body muscles were inhibited (only neck muscles)
• The stimulus parameters (size, speed, distance,
  ...) and network parameters (number of neurons)
  were fixed according to values found in
  literature. Both single stimulus and complex
  background were used

26
Optimal Values

• Many combinations of values give similar results
• Good results are also achieved without
  ipsilateral input

27
a and ß Parameters

• Regular parameter space. With different fixed
  values the aspect is conserved and the minimal
  error area (in black) is shifted
• ß parameters are not essential, the minimal error
  area is centered in point (0,0)

28
Performance Results

• An accuracy of less than 3 (real value) is
  achieved for small stimulus velocity values with
  20000 RGC2 and 2000 (less than 3300) tectal
  neurons.
• Robustness stable reaction to change in
  stimulus parameters and visual scene

29
With Complex Background

• The salamander has difficulties with following
  the prey stimulus as the amount of noise is
  considerable. It discriminates between objects
  with same apparent angular size, however orients
  at "average flies"
• The model should be coupled with a selective
  visual attention mechanism (enhanced retinal
  signals in the area containing the prey stimulus)
  and/or optokinetic or vestibucollic image
  stabilization reflexes (antagonistic head
  movements that compensate for body undulations)
• Integrating approach is trivial with a unique
  prey stimulus

30
Pretectum

• The salamander discriminates between a small prey
  object and a large predator object
• When the pretectum is abolished, escape behavior
  fails to occur
• Delayed response the salamander escapes for a
  longer time than the predator is visible
• Weakness based upon angular size, close prey may
  be interpreted as predator. Therefore the
  threshold is essential (arbitrary as no data
  exists on predation)

31
Snapping

• No additional neurons, based upon divergent
  patterns of tectal neurons projections
• Consistent with biological lesion data
• codes for closeness
• realistic precision (about 30)
• Depends on the movement direction

32
Reproduced Phenomena

• Lesion and stimulation experiments
• Lesion and stimulation of the optic tectum
• Lesion of the pretectum
• Generation of saccadic movements
• Monocularized salamanders
• Prey preferences

33
Saccadic Movements

• Pursuit movements such as the head accelerates
  for a few seconds, until maximum velocity is
  reached, and then is released
• We attribute them to the tectal cells resolution

34
Monocularized Salamander

• With one eye covered, H.Italicus shows a
  conspicuous approach behavior toward a prey
  stimulus. It takes a curved path and bends its
  body toward the side of a seeing eye,
  compensating by turning the head between 60 and
  90

35
Monocularized Salamander
36
Prey preferences

• All preferences are inherent to the network!!

37
Comparison to Previous Works

• Simulander I
• More neurons, but still biologically plausible
  (2000 vs. 100)
• Less accurate, more realistic (2-6 vs. 1)
• Inherent preferences vs. a function reflecting
  the stimulus size and velocity (corresponds to
  the observers knowledge)
• No real distribution of tectal neurons, respected
  in Simulander
• Faster reaction
• No positions in which stationery prey elicit
  orienting behavior
• Simulander II
• Lower precision, but more realistic (90, real
  success rate 40)
• In Simulander far objects elicit more activity,
  double inconsistency (should code for closeness
  and further objects seem smaller)

38
Response to questions

• Extraction of stimulus localization and depth
  estimates can be achieved with a structured
  mapping between the optic tectum and the brain
  stem
• The sensorimotor transformation of the horizontal
  angular distance of the tectum neurons to muscle
  activity can provide an accurate prey
  localization.
• Direct observation of the influence of head
  motion during the approach is provided.
  Additional mechanisms for dealing with the
  self-motion visual shifts are necessary
• The investigated tectum model accounts for the
  typical curved approach in monocularized
  salamanders
• A plausible mechanism that acts as a releaser for
  the snapping behavior is proposed

39
Conclusion

• We have implemented models of the three types of
  retinal ganglion cells, the optic tectum, the
  pretectum and a tongue projection mechanism that
  account for the typical feeding sequence and
  escape behavior
• The optic tectum model reproduces many
  experimental data
• Everything is observable
• Warning data and methodology dependant
• Our salamander resembles a newly born salamander
  thrown in the world

40
Future Work

• Study a tectum model with nonlinear weights
  functions
• Use the real distribution and receptive fields
  sizes of tectal neurons
• Time-dynamics vs. discrete time steps
• Study the effect of overlapping fields
  (redundancy gt error resistant, maybe emerging
  properties)
• Implement a visual attention model
• Implement experience-based models such as
  habituation
• Further development of the pretectum model
• Extend the model to other brain areas such as the
  nucleus isthmi or thalamus (obstacle avoidance)
• Develop a more elaborate model for depth
  estimation (not only frontal)
• Work on an object-background discrimination with
  respect to self-motion shifts of the visual input