Progress_Report-June-28th-2011

Views:
 
     
 

Presentation Description

No description available.

Comments

Presentation Transcript

Meeting June 28th 2011:

Meeting June 28 th 2011 Aidin Foroughi

Visual Servoing Components:

Visual Servoing Components Controller Visual Component Features Camera Parameters Desired Features Gain Interaction Matrix Velocity Controller Parameters Velocity Controller Hybrid/Switching Methods Feature Planning Path Planning

Visual Servoing Constraints:

Visual Servoing Constraints Controller Visual Component Velocity Controller Field of View Image Local Minima Singularities in Image Jacobian Kinematic Const. Dynamic Const. Collision Occlusion

Solutions given so far:

Solutions given so far Partitioned methods Hybrid/Switching methods Path Planning: In image space Global path planning Optimization based planning Potential Field Planners In fact Path planning is proving to be very powerful in dealing with all of the constraints.

Hybrid Systems:

Hybrid Systems Switching based systems can be thought of as simple rule based systems on top of the lower level methods. Platforms for hybrid systems with stable transition states have been implemented. Applications range from fault handling in airplanes to control of large plants. The idea of using a rule based system on top of a single continuous controller to adjust the parameters is also done in the 80s with Expert PID systems, where PID parameters are adjusted based on heuristic rules. Fuzzy expert PID systems came later. Stability is a big issue.

Hybrid Intelligent Controller:

Hybrid Intelligent Controller Controller Visual Component Features Camera Parameters Desired Features Gain Interaction Matrix Velocity Controller Parameters Velocity Controller Hybrid/Switching Methods Feature Planning Path Planning Field of View Image Local Minima Singularities in Image Jacobian Kinematic Const. Dynamic Const. Collision Occlusion Hybrid Intelligent Controller

PbD:

PbD Details of the following will not be discussed. Dynamical systems were discussed. HMM methods were studied. Primitive action methods were studied. We will consider Alex's proposal .for Dynamical Systems

PbD:

PbD Dynamical Systems for Periodic motion learning. There are certain meaningful parameters that can be learned from demonstrations. For instance Main Frequency and Amplitude of the motion Based on the task parameters (the coverage area etc) these parameters can be adjusted or altered. However, while it’s true that some higher level intervention is present with this approach, in the meeting we had with Alex we didn’t find much room for complicated reasoning.

Task Modeling for execution by Visual Servoing:

Task Modeling for execution by Visual Servoing If PbD is going to be implemented and executed using a visual servoing method, the task model should be designed in a way that facilitates this interfacing. PbD Visual Servoing Task Model demonstrations

Task Model:

Task Model It is possible to include a function in between to convert the task models to trajectories to be executed by the visual servoing component, but it may be more efficient to define the tasks so that they can be executed with the visual servoing component more easily. => To define tasks with respect to Visual Features.

Task models:

Task models Studying Industrial Tasks we can group the tasks to three groups. Tool is applied to a point or a series of separate points. Tool is applied to a trajectory. Tool is applied to a surface. This categorization is easy but has very nice properties: There is almost always a visual feature associated with the task model: For example the trajectories in the second category is easily recognizable in the image from edge information. In each category, the task constraints are completely different: for the first class the velocity or even the trajectory between the points are not important, while in the third category the distance to the surface and the velocities are important.

Task models:

Task models The type of the task can be given directly or extracted by classifiers. Task model will be bound to visual features which makes visual servoing easier. A discussion.

Cognitive Vision:

Cognitive Vision A discussion.

authorStream Live Help