Preview only show first 10 pages with watermark. For full document please download

Biology Inspired Robot Control Solutions

   EMBED


Share

Transcript

SysCon 2014 - Ottawa, ON, Canada Biology Inspired Robot Control Solutions Emil M. Petriu University of Ottawa April 2014 For many centuries, engineers were building upon mathematics and natural science principles from mechanics, electricity, and chemistry in order to develop an ever growing variety of more efficient and smarter industrial artefacts and machines. The time has now arrived to add biology and more specifically, anatomy, physiology and psychology to the scientific sources of knowledge to develop more efficient computational, control, sensing and perception technologies for robotics. Robot Perception Mechanisms that Emulate Those of the Humans. Model of the real world perceived by the human brain through sensory organs Real/Material World Bio-inspired Sensing & Perception The sensory cortex: an oblique strip, on the side of each hemisphere, receives sensations from parts on the opposite side of the body and head: foot (A), leg (B, C, hip (D), trunk (E), shoulder (F), arm (G, H), hand (I, J, K, L, M, N), neck (O), cranium (P), eye (Q), temple (R), lips (S), cheek (T), tongue (U), and larynx (V). Highly sensitive parts of the body, such as the hand, lips, and tongue have proportionally large mapping areas, the foot, leg, hip, shoulder, arm, eye, cheek, and larynx have intermediate sized mapping areas, while the trunk, neck, cranium, and temple have smaller mapping areas. (from [H. Chandler Elliott, The Shape of Intelligence - The Evolution of the Human Brain, Drawings by A. Ravielli, Charles Scribner’s Sons, NY, 1969]) NEURAL NETWORKS Looking for a model to prove that algebraic operations with analog variables can be performed by logic gates, Professor J. von Neuman advanced in 1956 the idea of representing analog variables by the mean rate of random-pulse streams [J. von Neuman, “Probabilistic logics and the synthesis of reliable organisms from unreliable components,” in Automata Studies, (C.E. Shannon, Ed.), Princeton, NJ, Princeton University Press, 1956]. Biological Neurons Dendrites Body Axon Synapse Dendrites carry electrical signals in into the neuron body. The neuron body integrates and thresholds the incoming signals.The axon is a single long nerve fiber that carries the signal from the neuron body to other neurons. A synapse is the connection between dendrites of two neurons. Memories are formed by the modification of the synaptic strengths which can change during the entire life of the neural systems. Neurons are rather slow (10-3 s) when compared with the modern electronic circuits. ==> The brain is faster than an electronic computer because of its massively parallel structure. The brain has approximately 1011 highly connected neurons (approx. 104 connections per neuron). Analog/Random-Pulse Conversion VR + V X + XQ -FS 1 0 R p(R) VRQ CLK CLOCK XQ R -FS 0 -1 +FS 1-BIT QUANTIZER ANALOG RANDOM SIGNAL GENERATOR 1 2 FS XQ X FS FS +FS p.d.f. of VR -FS FS-V 1 2 .FS FS+V 1 X 0 V -1 +FS VRP Analog/Random-Pulse and Random-Pulse/Digital Conversion • E.M. Petriu, K. Watanabe, T. Yeap, "Applications of Random-Pulse Machine Concept to Neural Network Design," IEEE Trans. Instrum. Meas., Vol. 45, No.2, pp.665-669, 1996, • E. Pop, E.M. Petriu, "Influence of Reference Domain Instability Upon the Precision of Random Reference Quantizer with Uniformly Distributed Auxiliary Source," Signal Processing (EURASIP), North Holland, Vol. 5, pp.87-96, 1983 Stochastic Data Representation XQ D /2 p.d.f. of VR VR V + X XQ b -BIT QUANTIZER + R ANALOG RANDOM SIGNAL p(R) GENERATOR 1/D R -D/20 +D/2 VRQ CLOCK D /2 CLK k+1 1/ D b. D (1-b) . D b. D k k-1 VRP 0 . (k-0.5)D k. D V= (k-b). D . (k+0.5)D Generalized b-bit analog/random-data conversion E.M. Petriu, L. Zhao, S.R. Das, V.Z. Groza, A. Cornell, “Instrumentation Applications of Multibit Random-Data Representation,” IEEE Trans. Instrum. Meas., Vol. 52, No. 1, pp. 175- 181, 2003. X Quantizat ion levels Relative mean square error 2 72.23 3 5.75 4 2.75 0.18 ... ... 8 1.23 ... ... analog 1 Mean square error 0.16 0.14 0.12 0.1 0.08 1-Bit 0.06 0.04 2-Bit 0.02 0 10 20 30 40 50 Moving average window size 60 70 Neural Network Architectures Using Stochastic Data Representation SYNAPSE SYNAPSE w 1j Xi ... w ij Xm SYNAPSE ... X1 w mj Σ F m Yj = F [ Σ j=1 wij . X i ] Neural Network for Pattern Recognition Auto-associative memory NN architecture a P 30x1 n 30x1 W 30x1 30x30 a = Hardlim ( 30 P1, t1 P2, t2 Training set W *P ) P3, t3 Recovery of 30% occluded patterns Neural Network vs. Analog Computer Modelling Both the Analog Computers and Neural Networks are continuous modelling devices. Neural Networks don’t require a prior mathematical models. A learning algorithm is used to adjust by trial and error during the learning phase the synaptic weights of the neurons. Discreet vs. Continuous Modelling of Physical Objects and Processes y y B A y(j) = ? x(j) DISCREET MODEL • sampling => INTERPOLATION COST y(j) = y(A) + [ x(j)-x(B)] . [ y(B)-x(A)] / [x(A)-x(B)] x x CONTINUOUS MODEL • NO sampling => NO INTEPPOLATION COST NN Modelling of 3D Object Shapes Compare the performance of three NN architectures used for 3D object shape modelling: • • • Multilayer Feedforward (MLFF ) Self-Organizing Map (SOM ) Neural Gas Network A.-M. Cretu, E.M. Petriu, G.G. Patry, “Neural-Network-Based Models of 3-D Objects for Virtualized Reality: A Comparative Study,” IEEE Trans. Instrum. Meas.," Vol. 55, No. 1, pp.99-111, 2006. MLFF Representation - Results 19000 points, 14-7-1, 4 extra surfaces, d=0.055, 1100 epochs, 3.3 hrs 51096 points, 20-10-1, 5 extra surfaces, d=0.055, 2000 epochs, 5.2 hrs. 2500 points, 12-6-1, 2 extra surfaces, d=0.06, 1020 epochs, 45 min. SOM and Neural Gas Modelling - Results Initial pointcloud Neural Gas 19080 points 1125 points, 42 min. 14914 points 875 points, 24.5 min. 13759 points 875 points, 22 min. er= 0.0098 SOM er= 0.0125 1125 points, 26 min. 875 points, 11 min. 875 points, 10 min. MLFF, SOM, and Natural Gas Modelling Performance Comparison: Compactness 1200 Storage space (Kb) 1000 800 600 400 Point Cloud MLFFNN SOM Neural Gas 200 0 hand face pliers statue Models MLFF, SOM, and Natural Gas Modelling of 3D Objects - conclusions * The use of neural network modeling is advantageous from the point of view simplicity and compactness. * MLFNN – provide continuous models, information on the entire object space, convenient for many applications, however they are time consuming. * SOM and Neural Gas – provide compressed models while maintaining the properties of the objects, have very good accuracy, and they are less time consuming * The use of any specific techniques depends on the application requirements. Elastic ball used for experimentation. Sampling points selected with the neural gas network for the ball. (from A.M. Cretu, E.M. Petriu, P.Payeur “Neural Network Mapping and Clustering of Elastic Behavior from Tactile and Range Imaging for Virtualized Reality Applications,” submitted to IEEE Tr. Instr. Meas., Nov. 2006 ). 3D pointcloud of data F profile(f0) profile(f1) profile(f2) profile(f3) Sample points Neural gas network Force/Torque sensor Range finder Deformation profiles Force Measurements Feedforward Neural Network f0 f1 f2 f3 (a) (b) Real and modeled deformation curves using neural network for rubber under forces applied at different angles: a) F=65N, α1=10° and F=65N, α2=170°, b) F=36N, α1=25°, and F=36N, α2=155 (from .A.M. Cretu, E.M. Petriu, P.Payeur “Neural Network Mapping and Clustering of Elastic Behavior from Tactile and Range Imaging for Virtualized Reality Applications,” submitted to IEEE Tr. Instr. Meas., Nov. 2006). FUZZY LOGIC Pioneered by Zadeh in the mid ‘60s fuzzy logic provides the formalism for modeling the approximate reasoning mechanisms specific to the human brain. “In more specific terms, what is central about fuzzy logic is that, unlike classical logical systems, it aims at modeling the imprecise modes of reasoning that play an essential role in the remarkable human ability to make rational decisions in an environment of uncertainty and imprecision. This ability depends, in turn, on our ability to infer an approximate answer to a question based on a store of knowledge that is inexact, incomplete, or not totally reliable.” [ “Fuzzy Logic,” IEEE Computer Mag, April 1988, pp. 83-93: ] The basic idea of “fuzzy logic control” (FLC) was suggested by L.A. Zadeh, “A rationale for fuzzy control,” J. Dynamic Syst. Meas. Control, vol.94, series G, pp.3-4,1972. FLC provides a non analytic alternative to the classical analytic control theory. ==> “But what is striking is that its most important and visible application today is in a realm not anticipated when fuzzy logic was conceived, namely, the realm of fuzzylogic-based process control,” [L.A. Zadeh, “Fuzzy logic,” IEEE Computer Mag., pp. 83-93, Apr. 1988]. CONTROLLED SYSTEM PROCESS ACTUATORS SENSORS ANALOG (CRISP) -TO-FUZZY INTERFACE FUZZY-TOANALOG (CRISP) INTERFACE FUZZIFICATION DEFUZZIFICATION INFERENCE MECHANISM (RULE EVALUATION) FUZZY RULE BASE Early FLCs were reported by Mamdani and Assilian in 1974, and Sugeno in 1985. DESIRED SYSTEM FUNCTION Fuzzy Logic Control OUTPUT OUTPUT y* Defuzzification y* x* INPUT Fuzzification INPUT x* Classic control is based on a detailed I/O function OUTPUT= F (INPUT) which maps each high-resolution quantization interval of the input domain into a high-resolution quantization interval of the output domain. => Finding a mathematical expression for this detailed mapping relationship F may be difficult, if not impossible, in many applications. Fuzzy logic control is based on an I/O function that maps each very low-resolution quantization interval of the input domain into a very low-low resolution quantization interval of the output domain. As there are only 7 or 9 fuzzy quantization intervals covering the input and output domains the mapping relationship can be very easily expressed using the“if-then” formalism. (In many applications, this leads to a simpler solution in less design time.) The overlapping of these fuzzy domains and their linear membership functions will eventually allow to achieve a rather high-resolution I/O function between crisp input and output variables. The key benefit of FLC is that the desired system behavior can be described with simple “if-then” relations based on very lowresolution models able to incorporate empirical engineering knowledge. FLCs have found many practical applications in the context of complex ill-defined processes that can be controlled by skilled human operators: water quality control, automatic train operation control, elevator control, etc., “FUZZY UNCERTAINTY” – WHAT ACTUALLY IS “FUZZY” IN A FUZZY CONTROLLER ?? There is tenet of common wisdom that FLCs are meant to successfully deal with uncertain data. According to this, FLCs are supposed to make do with “uncertain” data coming from (cheap) low-resolution and imprecise sensors. However, using a truck backing-up Fuzzy Logic Controller (FLC) as test bed, experiments show that the low resolution of the sensor data results in rough quantization of the controller's I/O characteristic: Experiments have shown also show that it is possible to smooth the I/O characteristic of a digital FLC by dithering the sensor data before quantization E.M. Petriu, J. Mao, "Fuzzy Sensing and Control for a Truck," Proc. VIMS-2000, IEEE Workshop on Virtual and Intelligent Measurement Systems, pp. 27-32, Annapolis, MD, April 2000. The truck backing-up ϕ θ Design a Fuzzy Logic Controller (FLC) able to back up a truck into a docking station from any initial position that has enough clearance from the docking station. ( x, y) Front Wheel Back Wheel d y (0,0) Loading Dock x LE LC CE RC RI 1.0 0.0 -50 -20 -15 -5 -4 0 4 5 15 20 50 x-position RB RU RV VE LV LU LB 1.0 0.0 00 -90 300 600 800 900 1000 1200 1500 1800 2700 truck angle ϕ NL Membership functions for the truck backerupper FLC -450 NM -350 NS -250 ZE PS 00 steering angle 250 θ PM 350 PL 450 x LE LC CE RC RI ϕ 1 3 4 NM NM NS NL NM NS PS NL NM NS PS PM VE NM NM ZE PM PM LV NM NS PS PM PL RL NL RU NL RV 2 NL 6 5 7 18 The FLC is based on the Sugeno-style fuzzy inference. The fuzzy rule base consists of 35 rules. 30 LU NS LL PS 31 PS PM 32 PM PL 33 PM PL 34 PL 35 PL θ [deg] θ [deg] 30 40 20 30 20 10 10 0 0 -10 -10 -20 -20 -30 -30 -40 -40 0 10 20 30 40 Time (s) 50 60 70 -50 0 10 20 30 Time (s) 40 Time diagram of digital FLC's output q during a docking experiment when the input variables, j and x are analog and respectively quantized with a 4-bit bit resolution 50 60 Dither Analog Input ∑ Dithered Analog Input A/D Low-Resolution Dithered Digital Input ∑ Dithered Analog Input Filter Digital FLC Dither Analog Input High Resolution Low-Pass Digital Output High Resolution Low-Pass Digital Output A/D Low-Resolution Dithered Digital Input Filter Dithered digital FLC architecture with low-pass filters placed at the FLC's outputs to smooth the non-linearity caused by the min-max composition rules of the FLC. θ [deg] θ [deg] 40 30 30 20 20 10 10 0 0 -10 -10 -20 -20 -30 -30 -40 -40 0 10 20 30 40 Time (s) 50 60 70 -50 0 10 20 30 Time (s) 40 50 60 Θ [deg] 30 20 Time diagram of digital FLC's output q during a docking experiment when the input variables, j and x are: (upper left) analog, (upper right) quantized with a 4-bit bit resolution, and (left) dithered before being 4-bit bit quantized and then a low-pass filter is placed at the FLC's output 10 0 -10 -20 -30 -40 -50 0 10 20 30 40 Time (s) 50 60 70 Y 50 40 30 initial position (-30,25) (a) 20 10 0 -50 (c) (b) [dock] 0 50 X Truck trails for different FLC architectures: (a) analog ; (b) digital without dithering; (c) digital with uniform dithering and 20-unit moving average filter Analog FLC Digital FLC Dithered FLC BIO-INSPIRED ROBOT SENSING AND ACTUATION Human Tactile Sensing The skin of a human finger contains four types of cutaneous sensing elements distributed within the skin: Meissner’s corpuscles for sensing velocity and movement across the skin; Merkel’s disks for sensing sustained pressure and shapes; Pacinian corpuscles for sensing pressure changes and vibrations of about 250 Hz; and Ruffini corpuscles for sensing skin stretch and slip. (from R. Sekuler and R. Balke, Perception, McGraw-Hill, 1990) Robot arm with tendon driven compliant joint (E.M. Petriu, D.C. Petriu, V. Cretu, "Control System for an Interactive Programmable Robot," Proc. CNETAC Nat. Conf. Electronics, Telecommunications, Control, and Computers, pp. 227-235, Bucharest, Nov. 1982, and E.M. Petriu, D. Petriu, V. Cretu, "Multi-Microprocessor Control System for an Experimental Robot with Elastic Joints," Proc. Nat. Conf. Cybernetics, (in Romanian), Bucharest, Romania, 1981). Tactile Sensor The tabs of the elastic overlay are arranged in a 16-by-16 array having a tab on top of each node of Merkel’s disk-like matrix of FSR elements sensing sustained pressure and shapes. This tab configuration provides a de facto spatial sampling, which reduces the elastic overlay's blurring effect on the high 2D sampling resolution of the FSR sensing matrix. • P. Payeur, C. Pasca, A.-M.Cretu, E.M. Petriu, “Intelligent Haptic Sensor System for Robotic Manipulation,” IEEE Trans. Instrum. Meas., Vol. 54, No. 4, pp. 1583 – 1592, 2005, • W.S. McMath, S.K.S. Yeung, E.M. Petriu, "Tactile Sensing for Space Robotics," Proc. IMTC/89, IEEE Instrum. Meas. Technol. Conf., pp.128-131, Washington, DC., 1989. Example of GUI window (from [C. Pasca, Smart Tactile Sensor, M.A.Sc. Thesis, University of Ottawa, 2004]) Bio-inspired robot passive-compliant wrist allowing the tactile probe to accommodate the constraints of the touched object surface and thus to increase the local cutaneous information extracted during the active exploration process under the force provided by the robot. Feeling the temperature and thermal conductivity of the touched object surface. Rufini corpuscles-like thermistors and a blood-vessel like source of heat (the white coloured tube) distributed within the tactile sensor’s elastic skin. Avatar Face 3D generic face deformed using muscle-based control Neutral Sad Happy Surprised Combining muscle actions it becomes possible to obtain a variety of facial expressions of Marius’ avatar: M.D. Cordea, E.M. Petriu, “A 3-D Anthropometric-Muscle-Based Active Appearance Model,” IEEE Trans. Instrum. Meas., Vol. 55, No. 1, pp. 91 - 98, 2006. Android Face • • • Plastic skull Latex rubber Proof-of-concept design P. Santos, E. de Castro Maia Jr., M, Goubran, E.M. Petriu, “Facial Expression Communication for Healthcare Androids,” Proc. MeMeA2013, 8th IEEE Int. Symp. on Medical Measurement and Applications, pp. 44-48, Ottawa, ON, Canada, May 2013 Avatar-Android Face Expressions Mapping From left to right: neutral, happiness, sadness, surprise, anger, fear, disgust P. Santos, E. de Castro Maia Jr., M, Goubran, E.M. Petriu, “Facial Expression Communication for Healthcare Androids,” Proc. MeMeA2013, 8th IEEE Int. Symp. on Medical Measurement and Applications, pp. 44-48, Ottawa, ON, Canada, May 2013 Thank you!