# Chayatat Ratanasawanya Min He May 13, 2010. Background information The goal Tasks involved in implementation Depth estimation Pitch & yaw correction angle.

## Presentation on theme: "Chayatat Ratanasawanya Min He May 13, 2010. Background information The goal Tasks involved in implementation Depth estimation Pitch & yaw correction angle."— Presentation transcript:

Chayatat Ratanasawanya Min He May 13, 2010

Background information The goal Tasks involved in implementation Depth estimation Pitch & yaw correction angle calculation Image processing LQR controller Experimental setup Results Data analysis Conclusion Questions/comments

Visual servo (VS) control – the use of computer vision data to control the motion of a robot. Relies on techniques from computer vision, image processing, and control theory. Two camera configurations: Eye-in-hand: camera is mounted on a robot manipulator or on a mobile robot. Camera is fixed in the workspace

The goal of visual servoing systems is to minimize the error defined by where s is visual feature vector s* is the desired visual feature vector Design of s: Consists of a set of features that are readily available in the image data (IBVS), or Consists of a set of 3D parameters, which must be estimated from image measurements (PBVS)

Use visual servoing techniques to make the 2DOF helicopter be able to track a constantly moving ping-pong ball. Quanser 2DOF helicopter A typical two-rotor helicopter model on a stand. It pitches and yaws around a pivot point.

Tasks involved: Depth estimation Pitch & yaw correction angles calculation Image processing LQR controller

We did 4 experiments altogether: -2 tests with the camera 25 from the wall -2 tests with the camera moved further back

Use the diameter of the ball in image to estimate the depth Depth, Z Focal length f=268 pixel Center of projection (CoP) Actual ball diameter d b =40mm Ball diameter on image, d

12 u e Z f CoP Ball diameter on image, d 1 2l r Pivot point of the 2DOF helicopter ψ Z u e

Z f CoP Ball diameter on image, d 1 2 l r Pivot point of the 2DOF helicopter ψ inc ueue Z

A controller design technique that works with the state-space representation of a system. with weighting matrices Q and R, calculate Same action as a PD or a PID controller. It is a position-based, joint-level controller. It accepts desired pitch and yaw angles and brings the helicopter to those angles. The desired angles are updated according to image processing result.

Video from on-board camera

Experiment 1 - Visual servoing 2DOF helicopter is at position 1 (25 from background )

Experiment 1 – visual servoing at position1

Horizontal direction (Yaw) overshoot = 9.53% Settling time = 16.19s Steady state error = 1 pixel Vertical direction (Pitch) overshoot = 2.75% Settling time = 14.48s Steady state error = 0.9 pixel

Experiment 2 – tracking at position 1

Experiment 3 – visual servoing at position 2 The helicopter is moved further from the background

Experiment 3 – visual servoing at position 2 Horizontal direction (Yaw) overshoot = 10% Settling time = 17s Steady state error = 1.4 pixel Vertical direction (Pitch) overshoot = 7.09% Settling time = 5.75s Steady state error = 0.4 pixel

Experiment 4 – tracking at position 2

The implemented visual servoing algorithm is simple because the 2DOF helicopter is a very simple system. Data shows that the 2DOF is able to visual servo the ball to the middle of the image frame. The LQR controller works better for the pitch than for the yaw: smaller overshoot and steady-state error.

It is able to track a constantly moving ping pong ball. The limitation is that the ball cannot move faster than 38.1cm/s. The distance of the ball from the camera does not matter as seen in tracking when the ball moves closer to the camera. The location of the 2DOF helicopter does not affect the performance of the system.