poster id 11 tesla summon auto-pilot robotics · 2017. 7. 23. · tesla summon auto-pilot robotics...

Poster ID 11

Tesla Summon Auto-Pilot Robotics

Mason Chen

ID-11 2017 IEOM STEM Poster Competition 1 © IEOM Society International

Project Introduction

• What is Tesla Auto-Pilot?

– Self-Parking Technique

– Self-Driving (Labor Saving)

– Less Pollution (Green Energy)

– Save Parking Space (more packed parking)

• Used Lego EV3 Robot to simulate Tesla self-parking Car

– Use Ultrasonic sensors to calculate the drive space

margin and park car accurately

2 ID-11 2017 IEOM STEM Poster Competition © IEOM Society International

Define Tesla Summon Objectives • To self-park quickly and safely.

• To be reliable and accurate in a tiny parking space.

• To be economic and less pollution.

• We conducted project benchmarking by comparing

Tesla self-parking and Lego Robotics self-parking:


Design Robotics Architecture

The ultrasonic sensor was the biggest concern.

• Soft objects absorbed sound waves and weaken the returning signal.

• Round or angled surfaces were hard to return the waves in focus.

• Ultrasonic Sensors can’t sense at 45 degrees (dead angle).

• Objects within 3 cm could not be sensed correctly (dead zone).

We used 3 ultrasonic sensors to resolve the above problems


Design Field & Robot Architecture

• Design U-Shape Parking Field

• Central foam wall flat and hard

• Back wheel to the ball design

• Ball wheel closer to mass center


Design EV3 Parking Algorithm

Left sensor distance < 8cm

Turn Right at 50% & at 50% speed

Yes No

Front sensor distance <

15cm

Turn Right at 40% & at 50% speed

Yes

No

Right sensor distance < 4cm

Turn Left at 30% & at 50% speed

Yes

No

Turn Left at 5% & at 50% speed

6 ID-11 2017 IEOM STEM Poster Competition© IEOM Society International

(1)Zone A: start U turn= 3.4 seconds

(2)Zone B: complete U turn= 3.8 seconds

(3)Zone C: parking= 1.9 seconds

Conduct Baseline Capability Analysis

U Turns


• Reduce 50% Cycle Time from 9.1s to 4.5s

• Won’t trade Safety for Cycle Time Reduction

• Front Sensor at Parking (last 5 points in Zone C)

• 50% reduction from 7.1cm to 3.5+/-0.25cm

• Ultrasonic Dead Zone Constraint ≥ 3cm

• Right Sensor Distance Safety Margin (Zones A & B):

• 25% reduction from 8.6cm to 6.5cm

• Very difficult change due to Ultrasonic Dead Angle limitation

• Right Sensor Parking Safety Margin (last 10 points in Zone C):

• 45% reduction from 10.1cm to 5.5+/-0.5cm

• Avoid slanting parking pattern

• Robot Weight (Energy Saving):

• 30% Robot weight reduction from current 910g - 700g

Set Performance Metrics and Goals


Improve Hardware Design

• Design Two-Stage EV3 Algorithm

– Zone A and Zone B in stage I, and

Zone C in stage II

– Design special parking algorithm in

stage II to improve safety

• Set Left Ultrasonic Sensor at 45 Degrees

to avoid Dead Angle

• Move both Left/Right Ultrasonic Sensors

forward to trigger earlier U turns, and

earlier parking preparation 9 ID-11 2017 IEOM STEM Poster Competition © IEOM Society International

Improve Battery Design

Replaced 6 “AA” Batteries with a Rechargeable Lithium

Battery.

• Trim 71gms to 673gms to meet Weight criteria < 700gms

• Lifetime Cost-Benefit Analysis:

– Extra One-Time $50 charge with Lithium Battery but reduce the

overall Lifetime Cost since Lithium Battery can last longer

– Larger battery charge capacity to improve Reliability

– Reduce Cycle Time from 9.1s to 8.5s

– It’s Worthy to Spend $50 in front

– Payback period is probably < 5K miles


Improve Tire Design

Choose the right tire size to improve Cycle Time

• Large Size won’t work due to too wide 3.5cm tire width (side safety

margin concern, and U-turn challenging)

• Extra-Small Size won’t work because the too slow linear velocity due

to 2cm diameter

• Choice between Small Size and Medium Size

– Linear Velocity determined by tire diameter: 4cm vs. 5.5 cm

– U Turn Side Margin determined by tire width: 2cm vs. 2.5cm

– Small Robot Weight Impact by tire weight: 15.3gm vs. 22.6gm (2% impact)


Improvements:

• Gage Repeatability

& Reproducibility

• Process Capability

(Cp)

• Process Stability

• Avoid Dead Zone

and Dead Angle

Fix Initial Condition

Problem Statements:

• Performance not repeatable

• Unexpected early movement

adjustments


Improved cycle time to 3.4s (meet < 4.5s criteria)

(1) Pattern is similar to baseline but shorter at no safety risk

(2) Design Changes: Lighter Robot, Earlier U-Turn & Medium Tire

Analyze Cycle Time Reduction


Need the following improvements:

(1) Front sensor parking safety margin in (3.25-3.75cm).

(2) Right sensor side safety margin (< 6.5cm).

(3) Right sensor parking safety margin in (5-6cm).

We need Optimize Phase

Optimization Strategy

• Trade longer parking time to parking safety margin

• Minimize any dead angle transition time by optimizing EV3 algorithm

while making smaller and quicker U-turns to improve side margin

• Further optimize the EV3 parking thresholds to improve the parking

safety margin 14 ID-11 2017 IEOM STEM Poster Competition © IEOM Society International

Allow 0.1s parking to adjust the slanted pattern.

Modify the safety margin range while considering the dynamic parking.

Meet 5-6cm safety margin window.

Right Sensor Parking Margin Optimization


We have met all five requirements.

Project Achievements Summary


Project Completion and Summary

Team Achievements:

• Understood Ultrasonic Sensor Dead Angle and Dead Zone

• Optimized three Ultrasonic Sensors

• Integrate and Optimize both EV3 Algorithm and Robot Architecture

• Use Minitab Statistics to discover Robot motion patterns

• Believe in Team Building and Data-Driven Leadership

Continuous Improvement Opportunities:

• Linear and gradient parking algorithm

• Add another sensor (Color or/and Gyro)

• Design slanted Parking Field

• Study Battery Power Reliability Modeling


poster id 11 tesla summon auto-pilot robotics · 2017. 7. 23. · tesla summon auto-pilot robotics...

Documents