traffic based pathway optimization michael legore tjhsst csl

15
Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Upload: august-rich

Post on 03-Jan-2016

215 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Traffic Based Pathway Optimization

Michael LeGoreTJHSST CSL

Page 2: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Purpose

To create an algorithm to find an optimal path using the heuristic of time.

To train the algorithm to avoid traffic hotspots at peak hours.

To use traffic data to estimate travel time on roads.

To optimize algorithm based on results of analysis.

Page 3: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Background

Path finding algorithms principally use distance as a heuristic.

Many consumers are concerned only with travel time.

Traffic patterns in certain area at certain times can be used to accurately predict travel times.

Page 4: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Background: A* Search

Algorithm to find the optimal path. Searches through current node and adds each

segment to search space. Iterates over search space until the minimum

pathway returned is at the destination. Uses Heuristic to estimate distance to

destination.

Page 5: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Traffic Based Algorithm

Instead of dealing with distance, algorithm finds path based on travel time

Travel Time is estimated from data VDOT Estimate can take into account contexts such

as weather, time of day, and other traffic fluctuations.

Page 6: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Implementation

Ruby for programming A* Search with a modified heuristic to find path Use VDOT traffic data to model the average

speed. Feed this data into A* Search Use it to find the optimal path Tweak algorithm based on results.

Page 7: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Development

Search Algorithm Algorithm is an A* search with a modified

heuristic. Interpreting of Traffic Data

Needed to provide a heuristic for travel time Provides data to estimate the speed of cars. Speed used to find the time spent on each

road segment.

Page 8: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

VDOT data

Gives three main factors: AAWT – Average Annual Daily Traffic AAWDT – Average Annual Weekday Traffic K Factor – The ratio of AAWT to design (peak)

hour traffic Ratio of AAWT to AAWDT K value used to factor in Rush-Hour effects

Page 9: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Limitations of VDOT Data

Only provides data over time Does not model based on time of day Gives peak hour traffic concentration but not

time of peak hour(s). Only gives concentrations of traffic Does not correlate with the TIGER data used

for the road shapes

Page 10: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Demonstration

Input Addresses Input time of day Returns optimal path for that time of day.

Page 11: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Original Implementation/Plan

Traffic Simulation Use traffic simulation to feed data back into the

algorithm Keep track of travel times of agents Unfeasible given time and computing

constraints TIGER data not adequate for an agent based

model.

Page 12: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Possible Extensions

Find more complete data for estimating travel time

Integrate more factors into the heuristic Weather Holidays More?

Distributed Approach to data collection Integrated into GPS devices iPhone app?

Page 13: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Things I learned

Data collection as a single person is not possible

Writing a sophisticated traffic simulator to model 1 Million+ agents is not at all feasible in the current state of computing

Data for traffic can be very fragmented and not standardized

Page 14: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Conclusion

Time can be a good heuristic for finding a path An A* search is easily modifiable to

accommodate many types of heuristics with different effects

Proper data is necessary for such an algorithm to work

Page 15: Traffic Based Pathway Optimization Michael LeGore TJHSST CSL

Questions?