KDC Arm Project

John KuaKathryn Rivard

Benjamin StephensKatie Strausser

Outline

• Overview

• Mechanical Design & Fabrication

• Electrical & Software

• Simulation & Control

• Demo

• Future Work

Overview

• Goal: Design and control a novel dynamically balancing manipulator

• Motivation: Assistive robot manipulator to be mounted on friendly household Segway robot

• Typical Task: Pick up objects from cupboard, open doors, take out the garbage

Mechanical

• Started with dynamic analysis to decide system specifications and parameters

Mechanical

• Designed mechanism in Solidworks CAD

• Easy to use aluminum extrusions

• Single linear actuator moves counterweight to balance arm

• Ball bearings support screw drive assembly

Electrical• Sensors

– Joint encoder (500 CPR)– Screw encoder (2048 CPR)– Accelerometer (3-axis)– Rate gyro (2-axis)– Limit switches (SPST)

• Motor control– Serial commands – 24V/10A max

• Gripper control– Standard servo PWM

• Microcontroller– ARM7TDMI-S– 60 MHz

Software

• Microcontroller– 100 Hz main loop

• Motor control input/output• Servo control• Data logging to serial port

– 50 kHz sensor polling– 500 Hz servo PWM

output

• PC Side– Controller (100Hz)– Data logging to file

PC

Microcontroller

Motor Controller

Gripper Servo

Sensors

RS

-232

TT

L S

eria

l

TT

L P

WM

TT

L / A

nalo

g

Software

• Kalman Filter– Estimated states – tilt and gyro bias– Sensor inputs at two speeds

• 2 kHz rate gyro• 50 Hz accelerometer

– Gyro primary sensor• Drifting bias

– Accelerometer to estimate bias• Noisy, captures more than pure rotation

Software – Kalman Filter

• Process update (2 kHz)

• Measurement update (50 Hz)

Software – Kalman Filter

0 1 2 3 4 5 6 7 8 9 10-40

-30

-20

-10

0

10

20

30

Time (s)

Tilt

(D

eg)

Shoulder Position

Encoder Tilt

Accelerometer Tilt

Onboard KF Tilt

Software – Kalman Filter

0 1 2 3 4 5 6 7 8 9 10-300

-250

-200

-150

-100

-50

0

50

100

150

200

Ang

ular

Vel

ocity

(D

eg/s

)

Time (s)

Shoulder Velocity

Encoder Derived Velocity

Compensated Gyro OutputEstimated Gyro Bias

Simulation & Control

• Problem: Move counterweight to balance arm independent of gripper weight

• Key Idea: Counterweight moves slowly -- must plan ahead!

• Solution: – Plan a trajectory from an initial state to the arm set

point– Use acceleration to determine the counterweight set

point.

Simulation Results

Experimental Results

DEMO!

Future Work

• More sophisticated control: Input shaping, MPC, etc.

• Optimize counterweight mass/travel

• Shoulder joint actuation

• Joint shape iteration (yoking joint)

• Increase robustness for object balance

• Interface to Segway

Effect of a yoke: Balanced torques

The End

• Special thanks to Garth Zeglin for helping us design the arm.