The mission of the

1
The mission of the Center for Neuromorphic
Systems Engineering is to endow machines with
senses and sensory like behavior
MUSEUM BOJIMAN VAN BEUNINGEN
2
HMI - Human Machine Interfaces
OBJECT CLASS RECOGNITION
Task Recognize objects that come in a variety of
shapes, can be partially occluded and may be in
front of a cluttered background.
3
HMI - Human Machine Interfaces
HUMAN MOTION DETECTION and LEARNING
Prof. P. Perona
Probabilistic model of joint position and
velocity Detection rate 90 in 2 frames (60 ms)
and 98 in 7 frames (200 ms)
Learn models of human motion that can generate
synthetic yet realistic-looking motions and based
on learning probabilistic models of human motion.
4
Collaborative Robotics
design algorithms or distributed
problem-solving devices inspired by the
collective behavior of social insect colonies and
other animal societies
Task Allocation and Distributed Sensing

• algorithms for optimal task allocation and
  probabilistic models that provide accurate
  forecast of the resulting collective behavior.
• simple local rules of distributed sensing,
  communication and action

Collective manipulation

• stigmergy for collective building
• local communications

5
Collaborative Robotics / Electronic Nose
Collective exploration finding a plume source

• local communications
• collaborative vs., non-collaborative approaches
• team aggregation vs. segregation

Swarm Intelligence and Traffic Safety

• framework in which autonomy, emergence, and
  distributed robustness replace centralized
  control
• controller with access to data from on-board
  obstacle sensors and lane sensors to have an
  awareness of the state of the environment

6
VISUAL ATTENTION
Prof. C. Koch
The saliency map topographically encodes for the
local conspicuity in the visual scene, and
controls where the focus of attention is
currently deployed
7
(No Transcript)
8
VISION CHIPS and AWARENESS
Based on optic flow patterns as a source of
navigation information, create small, low-power
visual sensor with integrated analog parallel
processing to extract motion in real-time
Developing design principles for intelligent
artificial systems with the ability to mimic the
ability of biological organisms to interact with
complex and variable environments, using
incomplete information from high-dimensional
inputs.
9
MEMS - Micro Electrical Mechanical Systems
MICROBAT
First battery-powered ornithopter (flapping-wing)
Micro Aerial Vehicle (MAV).
10
MEMS - Micro Electrical Mechanical Systems
Lab-on-a Chip
Flow Control
11
Robotics Hyper- Redundant Robots and others

Hyper -redundant robots are analogous in
morphology and operation to snakes, elephant
trunks, and tentacles. Their high degree of
articulation makes them well suited for
inspection in highly constrained environments.

• Caltech has collaborated with JPL, NEC, and
  Cedars Sinai Hospital to develop snake-like
  robots of various sizes for
• Urban search and rescue demonstrations
• Space-station inspection
• Minimally invasive surgical applications.

Robotic fish - study the motion of carangiform
fish. to understand how they generate and control
their motion
12
Neuroprosthetics
Prof. J. Burdick
13
Holography and electro-optical computing

• Optical holographic correlators are excellent
  systems for target recognition and tracking tasks.

• Optical holography can also be used for
  navigation

14
Holography and electro-optical computing Optically
Programmable Reconfigurable Computers
Prof. D. Psaltis
The OPGA module is very compact and has extremely
short configuration time allowing for dynamic
reconfiguration. The reconfiguration capability
of the OPGA can be applied to solve more
efficiently problems in pattern recognition and
digit classification. 
15
Double Shock Waves

• Fast movies
• Angle multiplexing method
• Wavelength multiplexing
• 4-D imaging
• Dynamic 3-D imaging
• Spectroscopy of 3-D objects

t 12ns
One volume hologram can act as a single imaging
lens with an inherent color filter, due to the
spatial selectivity of the hologram and the
wavelength Bragg matching. However, within the
dynamic range of the holographic medium, multiple
holograms can be recorded into the same medium,
which makes it possible to generate a holographic
imaging element to act as a combination of
multiple imaging lens and color filters Each
hologram in the element will image one 2D slice
with corresponding color of a 4D ( 3D spatial
dimension plus color information) objects into a
photo camera. With the presence of multiple
holograms, multiple 2D-slice images can be
collected from separated camera at the same time,
which gives the full information of the 4D
object.
16
MACHINE LEARNING
Q How can we find the correct function when the
data set is sparse and/or noisy? A Incorporate
Prior Information - Hints
17
MACHINE LEARNING
System is designed to differentiate fourteen WBC
types

• Feature Extraction - color, size, shape,
  texture
• Pattern Recognition - trained on examples for
  different classes
• Contextual Learning - identity of a cell is
  dependent not only its own appearance, but also
  on the identity of the other cells in the sample

• Symmetry hint
• Increased return on investment
• Wide range of applications

18
WIRELESS COMMUNICATIONS
LOW POWER SHORT RANGE COMMUNICATIONS

• Milliwatt power ranges
• Short ranges (10m 100m)

19
(No Transcript)
20
(No Transcript)
21
MEMBER AFFILIATES - SPONSORED PROJECTS
22
WHAT ABOUT.

• Institute holds the patent

• Company negotiates license
• (field of use, royalties, royalty free)

• Publications (retain confidential data)

• Reporting to NSF

23
CONTACTS
Professor Pietro Perona, Center Director
perona_at_caltech.edu Professor Joel Burdick,
Deputy Center Director jwb_at_robotics.caltech.
edu