Jockusch, Ján: Exploration based on neural networks with applications in manipulator control. 2000
Inhalt
- Introduction
- Robotics Lab Setup
- Human-Machine Interfacing
- Exploration and Control
- Previous Work on Robotic Control
- Introduction to a Novel Approach
- The Controller Architecture
- Manipulator and Sensor Hardware
- A Layered Controller Architecture
- The Controller Layer
- The Controller Structure
- Mechanisms for Safety and Reliability
- Interfacing with Other Processes
- Implementation Details
- A Visual Controller Interface
- Performance Evaluation
- The State Machine Layer
- Exploration with Vector Quantization Networks
- Approaching Intertwined Tasks
- Introduction
- Literal Interpretation of Topological Maps
- Evaluation of Existing Models
- Critical Aspects in Robotics
- Preparation of Input Data
- Adaptive Metrics for Input Rescaling
- Expansive Adaptation
- The Instantaneous Topological Map
- Improving the GNG for Correlated Stimuli
- The Instantaneous Topological Map (ITM)
- Results
- Statistical Distributions
- Architectural Comparison
- Dimensionality of Input Data
- Conclusions
- Path Finding and Obstacle Avoidance with the ITM
- An Active Exploration Engine
- Discussion and Outlook