Related Experiment Videos
Modeling and production of robot trajectories using the Temporal Parametrized Self Organizing Maps
Antonio Carlos Padoan Junior1, Guilherme De A Barreto, Aluizio F R Araújo
1Departamento de Engenharia Elétrica, Universidade de São Paulo, Av. Trabalhador Sancarlense, 400, São Carlos, São Paulo, CEP 13566-590, Brasil. padoan@sel.eesc.sc.usp.br
International Journal of Neural Systems
|August 19, 2003
Summary
A new unsupervised neural network, the Temporal Parametrized Self Organizing Map (TEPSOM), learns and reproduces robot trajectories. TEPSOM demonstrates superior accuracy over SONARX for robot trajectory reproduction and interpolation.
Area of Science:
- Robotics
- Artificial Intelligence
- Machine Learning
Background:
- Robotic trajectory learning and reproduction are crucial for automation.
- Existing methods may struggle with accurate interpolation between learned states.
- Modeling complex robot kinematics requires robust neural architectures.
Purpose of the Study:
- To introduce an unsupervised neural architecture, Temporal Parametrized Self Organizing Map (TEPSOM), for robot trajectory learning.
- To enable TEPSOM to learn, reproduce, and interpolate complex robot trajectories.
- To evaluate TEPSOM's performance in modeling robot inverse kinematics.
Main Methods:
- Proposed an unsupervised neural architecture combining Self-Organizing NARX (SONARX) and Parametrized Self-Organizing (PSOM) networks.
- Utilized the TEPSOM network to model the inverse kinematics of the PUMA 560 robot.
- Trained and tested the network on robotic trajectories with repeated states.
Main Results:
- TEPSOM effectively learns and reproduces complex robot trajectories.
- The architecture successfully interpolates new states between learned trajectory points.
- Simulations confirmed TEPSOM's higher accuracy compared to SONARX in trajectory reproduction.
Conclusions:
- TEPSOM offers an advanced unsupervised approach for robot trajectory modeling.
- The combined SONARX and PSOM architecture enhances interpolation capabilities.
- TEPSOM presents a more accurate alternative for learning and reproducing robotic movements.