Professional Documents
Culture Documents
Slides: A Distributed Neural Controller For Locomotion in Linear Modular Robotic Configurations
Slides: A Distributed Neural Controller For Locomotion in Linear Modular Robotic Configurations
9 Workshop - Robots colabativos e interaccin Humano-Robot - Robocity 2030-II CAR, E.T.S. de Ingenieros Industriales, Madrid. 30/May/2011
Index
Index
1.
Introduction
2. Robotic platform 3. Proposed model 4. Evolution 5. System 1 6. System 2 7. Fault tolerance 8. Conclusion
1. Introduction
1. Introduction
Sinusoidal Oscillator
Sinusoidal Oscillator:
i t=Ai sin 2 t i Oi T
1. Introduction
Objective
1. Distributed 2. Homogeneous 3. Adaptive 4. Fault tolerant
2. Robotic platform
Y1 in OpenRAVE
3. Proposed model
Neural Controller
Input layer:
1-2: Connector information 3: Feedback from self actuator 4-5: Feedback from neighbouring module's actuator One hidden layer with five hidden neurons One output neuron connected to the actuator
Hidden layer:
Output layer:
4. Evolution
Evolution
Genetic Algorithm
Roulette Wheel selection method Intermediate Recombination method Fitness: Displacement measured as euclidean distance from origin Evaluation time: 50 seconds. Parameters Population size Generations Crossover percentage Elite population retained Mutation rate Values 50 100 50% 12.5% 1/Size of genome
8
5. System 1
5. System 1
10
6. System 2
Where, : Output from the ANN : Angle of the actuator at time ti and x are constants.
11
6. System 1
12
6. System 1
13
7. Fault tolerance
14
8. Conclusion
Conclusion
Proposed model validated for locomotory oscillations. Fault tolerance with limit cycle behaviour. Fast convergence to optimal oscillatory pattern.
Future work
Validate this model on the real Y1 modular robots. Adaptive behaviour for locomotion on different surfaces
Evolve the topology to both complexify and to find the minimal neural architecture.
15
9 Workshop - Robots colabativos e interaccin Humano-Robot - Robocity 2030-II CAR, E.T.S. de Ingenieros Industriales, Madrid. 30/May/2011