Mason Chen (Black Belt) Morrill Learning Center, San Jose, CA

Transcription

1 Poster ID 12 Google Robot Mason Chen (Black Belt) Morrill Learning Center, San Jose, CA

2 D1 Observations and Research Google Cars stop at the red light and speed up at green light how & why Google Car can adjust its speed at traffic lights?

3 D2 Design Objectives and Scope Design Objectives : In green light, robot goes full speed. In yellow light, robot slows down by 50%. In red light, robot stops for 5 seconds then goes full speed Project Scope defined by the SIPOC Analysis:

4 D3 Project Team Building G4-G Weekly ASA Meetings Dec ~ Mar Team went through several huge Storming events Mentors facilitated team building progress through Statistical and Objective data-driven approach

5 D4 ASA G4-G6 Poster Team Project-Oriented Use IBM SPSS Statistical Software Apply AP Statistics

6 D5 Project Challenges v Design SUV fields to test Robotics Self-Driving Capability (S curve: many sharp turns; U Curve: two ~90 o turns; V Curve: one inner tip turn) How to maximize speed at sharp turn locations

7 M1 Methods and Materials Added the Black Line to separate the background area and the field G4-G Need 2 color sensors: 1. Color Intensity (Red LED) to follow the black line 2. Color Number Mode (Green Yellow Blue LED) to adjust the robot speed Change the back wheel to Ball Design Two Color Sensors placed at a distance close to the Black Line width

8 M2 Software Design Principles G4-G Define Turn and Speed Variables to Slow Down the Robot Movement at Sharper Turns (Defined as Threshold) Hard Turn and Gentle Turn Threshold Variables Hard Turn Threshold Gentle Turn Threshold Slow Down Hard Speed Gentle Speed Straight Speed

9 M3 Turn & Speed Control 1. Right Y is Color Number (3= Green, 4= Yellow, 5= Red); Left Y is Turn Threshold (+/- 80= Hard, +/-25= Gentle); X is Time 2. Observed 2 major Hard Turn (at -85 or at 85) area:

10 M4 Baseline Capability Analysis After implemented the new SUV field, Back Ball Wheel, Two Color Sensors, Turn and Speed Variables, conducted baseline capability analysis Due to resource constraint, set up sample size =7 (to avoid sample-t test confidence interval penalty) Box-Plot was conducted to minimize any Outlier impact on the Central Tendency (Median over Mean) and Spread (IQR over Standard Deviation) G4-G6 2578

11 M5 Box Plot Analysis Summary S Curve observed worse performance than U and V curves. Team will focus on S Curve improvement G4-G Set up the Project Target to achieve 1.25s-1.75s Reduction S Curve Median under 20 seconds U Curve Median under 19 Seconds V Curve Median under 18.5 seconds

12 A1 Histogram Analysis Objective: analyze why S curve observed a longer cycle time At Hard Turns, Robot will change to negative Hard Speed to make such a Hard Turn Observed significant portion of hard turns at -80 and +80 turn levels which will slow down the Robot Cycle time Improvement Strategy: Minimize the Hard Turn Proportion Optimize Robot s Starting Position (better start to avoid IEOM Society International early Hard Turns) Hard Turn Hard Turn

13 A2 Baseline Contingency Objective: analyze Hard Turn proportions across Colors Total 43.8% cycle time belonged to Hard Turns Reduce cycle time by minimizing Hard Turn portions Observed two major Hard Turn area in S field 2 1

14 A3 Identify Key Factors Identified four key factors: (1) Hard Turn Threshold, (2) Gentle Turn Threshold, (3) Gentle Speed, (4) Straight Speed All sample distributions passed Normality Test (P-Value> 0.05) Turn Variable is calculated by Improvement Directions both Shape & Robotics Speed Optimize Hard Turn Threshold & Gentle Speed Robot can not move too Optimize Gentle Turn fast at Sharper Turns Threshold & Straight Speed

15 A4 Validate Improvement Baseline Setting vs. Optimization Setting: S < 20s U < 19.5s V < 18.5s 1-sided 2-sample t test, reject Ho, cycle time reduction N Mean StDev Baseline Optimize T-Statistic = 5.87 P-Value = 0.00 DF = 40 ( )

16 A5 Optimization Summary Cycle Time Comparison (Y: Color Number, X: Time) Baseline (Yellow Line) vs. Optimization (Red) Cycle Time Reduction % (Exclude 10s Stop Time) Major cycle time reduction at: 1. Green Zone in middle session 2. Yellow Zone in end session S curve has more cycle time reduction than U & V S: more cycle time reduction in the middle and end

17 C1 Minimize Hard Turns? Turn Count Histogram Comparison: Baseline (Left Chart) vs. Optimization (Right Chart) YES, completely eliminated the Hard Turns at -80, which was happened at the Yellow Zone in the end session

18 C2 Optimization Contingency Hard Turn Count: > 50% reduction (from 273 to 126) Hard Turn %: 43.8% (Baseline) to 31.7% (Optimize) Straight %: 13.8% (Baseline) to 15.6% (Optimize) Minimize Hard Turn Optimization Strategy: Successful!

19 C3 Improve Average Power Compare Power Distribution (Baseline vs. Optimize) Significantly reduced Hard Turn Speed at the Yellow Increase Straight Speed from 30 to 100 If excluding the Red Stop rest time, the average power has been improved from 9.10 (Baseline) to (Optimize), more than 200% improvement Average Power= 9.10 Average Power= 28.82

20 C4 Conclusions and Learning Achieved Cycle Time Reduction on all SUV curves Demonstrate Google Car concept using Lego Robot EV3 Overcome the Hard Turns (Green/Middle; Yellow/End) Minimize the Hard Turn proportion > 50% Improve Average Power > 200% Learning Opportunities from Mentors and Team Experts Project Team Building Process, and SIPOC Practice SPSS: Histogram, Box-Plot, Contingency, t test, Scatter Integrate Lego Robotics Hardware and EV3 Software Conduct statistical and objective root cause analysis We love and believe statistics in our real life! Future Opportunities Study Environmental Color Intensity Impact Apply more than 3 Turn Regions (Hard, Gentle, Straight) Will continue 2017 STEM Project