UNIVERSITA DEGLI STUDI DI CATANIA FACOLTA DI INGEGNERIA PhD course in Electronics, Automation and Control of Complex Systems - XXV Cycle DIPARTIMENTO DI INGEGNERIA ELETTRICA ELETTRONICA E INFORMATICA XXII BrainStorming Day Tutor : Prof. P. Arena Coordinator : Prof. L. Fortuna Alessandra Vitanza PhD Student International PhD - XXV Cycle INGV Sez. Catania 20 th May 2010
XXII BrainStorming Day Activities Overview: SPARKRS4CS IBB Experiments - Adaptive Termination - Integration (Simulated/real Environment) Cooperation - Threshold adaptation with STDP Learning method Task 1: Implementation and test of Insect Brain Architecture Experiments Task 2: Investigation & combination of learning methods for cooperative approaches SPARKLiN 2 SPARK Library for Neural Networks Task 3: Investigation and implementation of supporting tools.
Description: wild-type flies are attracted by an object presented beyond a water-filled moat, stopped walking at the edge of the moat and turned into a new direction. MB-defective flies kept walking toward the attractive but inaccessible object. Experiment: The robot is placed into an arena whith attractive objects. Some obstacles can make unreachable the objects. When an object has been detected, the robot should switch the behaviour from Exploration to Object Approach. Adaptive Termination Involved Blocks: - MB (the MB(2) block): Evaluation of on-going behaviour - BSN: Choice of on-going behaviour - CX: Visual targeting input
Adaptive Termination
Wild type Adaptive Termination MB-defective mutant
Adaptive Termination
Description: A robot controlled through the insect brain architecture behaves in an environment that allows the concurrent evolution of different blocks Involved Blocks: Complete architecture IBB Experiments Integration Arena: The arena includes visual targets, odor plumes, obstacles
Evaluation of sensory State Obstacle recognition Landmark detection setting Mushroom Bodies selection execution Approach Homing Sleep Exploration EB
Evaluation of sensory State IBB Experiments Obstacle recognition Landmark detection Integration setting Mushroom Bodies selection execution Approach Homing Sleep Exploration EB
Evaluation of sensory State Obstacle recognition Landmark detection Integration setting Mushroom Bodies selection execution Approach Homing Sleep Exploration EB
Evaluation of sensory State Obstacle recognition Landmark detection Integration setting Mushroom Bodies selection execution Approach Homing Sleep Exploration EB
Evaluation of sensory State IBB Experiments Obstacle recognition Landmark detection Integration setting Mushroom Bodies selection execution Approach Homing Exploration EB Sleep/No Activity
Integration
Integration Simulation results MB Odor Learning Escaping behaviour
Integration Simulation results
Integration Robot Experiment
XXII BrainStorming Day Activities Overview: SPARKRS4CS IBB Experiments - Adaptive Termination - Integration (Simulated/real Environment ) Cooperation - Threshold adaptation with STDP Learning method Task 1: Implementation and test of Insect Brain Architecture Experiments Task 2: Investigation & combination of learning methods for cooperative approaches SPARKLiN 2 SPARK Library for Neural Networks Task 3: Investigation and implementation of supporting tools.
Cooperation with SNN specialization in a SNN for obstacle avoidance and visual target reaching by using: - an environmentally mediated global Reward - Threshold adaptation New IDEA: Combination with unsupervised learning paradigm: STDP Learning
Cooperation with SNN Correlation based navigation algorithm With STDP Learning Obstacle avoidance Network Object approaching Network Threshold adaptation for role specialization
Cooperation with SNN Before Learning After Learning ~30s ~130s ~60s Fusion of synaptic and threshold plasticity can be useful to investigate: - interplay between Specialization mechanism and STDP Learning - improvement in speed-up of the Specialization - improvement of the learning convergence
Papers & Courses Published papers: G.Interdonato, V.Pappalardo, M.P.Russo, A.Vitanza: An efficient motion system for autonomous robots. Paper submitted to the IEEE ISA Student Contest 2007. Publicized in Proc. of 3rd ISA European Student Paper Competition, St. Petersburg, Russia, 25-26 May (First italian classified). P. Arena, S. De Fiore, L. Patanè and A. Vitanza: A new bio-inspired perceptual control architecture applied to solving navigation tasks, Proc. of SPIE, Vol. 7365, 736507 (2009); doi:10.1117/12.821432 P. Arena, M. Cosentino, L. Patané, A. Vitanza: SPARKRS4CS: a software/hardware framework for cognitive architectures (Invited Paper), Proc. of SPIE, Vol. 8068 (2011) [8068A-18]; doi: 10.1117/12.887388 Attended/Attending Courses: Computer Vision (prof. Battiato DMI) Robotica Industriale (prof. Muscato DIEES) Informatica musicale (prof. M. Salfi DMI) Summer/Professional Schools: Grid School: Course organized by Consorzio Cometa & Cea. University of Catania 21-30 Sept. 2009 Professional School:3rd Euro-Mediterranean UNIversity Summer Semester Catania, 30 Aug.-10 Sept. 2010 PhD School: Electronic, Automation and Control of Complex Systems.4-28 Ott.2010 Conferences: SPIE Microtechnologies (April 18-20 April, 2011- Prague Congress Centre, Czech Republic)