1 / 41

Sensor-less Traffic-condition-based Car Navigation

Sensor-less Traffic-condition-based Car Navigation. Yiyu Shi and Yu Hu Electrical Engineering Department University of California, Los Angeles. Outline. Motivation Problem formulation Algorithms Implementations Simulation results Conclusions. Motivation. GPS.

jessie
Download Presentation

Sensor-less Traffic-condition-based Car Navigation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Sensor-less Traffic-condition-based Car Navigation Yiyu Shi and Yu Hu Electrical Engineering Department University of California, Los Angeles

  2. Outline • Motivation • Problem formulation • Algorithms • Implementations • Simulation results • Conclusions

  3. Motivation GPS

  4. Main Contributions (Efforts) • Formulate the sensor-less traffic-condition-based car navigation problem and propose an algorithm to solve it efficiently. • Study on OMNeT++ and Mobility Framework. • Implement a traffic system based on our algorithm on the top of MF and OMNeT++ • Simulate and analyze the results based on our statistic platform with matlab.

  5. Outline • Motivation • Problem formulation • Algorithms • Implementations • Simulation results • Conclusions

  6. Problem Formulation • Given a map with vertices set V and edge set E; The sources and destinations of each vehicle; Traffic information (i.e., the green light period, and real-time traffic information from sensors) • Find the most efficient way for each vehicle to reach its destination as quick as possible.

  7. Outline • Motivation • Problem formulation • Algorithms • Implementations • Simulation results • Conclusions

  8. Previous Approaches • Static Routing • Assign each edge a fixed weight proportional to the estimated congestion • SPSP (Single-pair-shortest-path) • Maze Algorithms • Heuristics (Greedy approaches) • Advantages • Extremely time efficient (Low complexity) • Stable • Disadvantages • Cannot reflect real time cases thus not accurate

  9. Previous Approaches (cont’d) • Dynamic Routing • Assign each edge a time-variant weight proportional to the estimated congestion • Each router automatically adjust to changes in network topology or traffic. • Advantages • Accuracy. Can reflect time-variance. • Disadvantages • High computation cost especially when the map is large • Require global information (In real situation, vehicles that are too far away cannot build connection)

  10. Previous Approaches (cont’d) • Yet another problem • There is a delay between the change in traffic flow and the corresponding change in path selection • If each vehicle is aimed at its own optimal path, severe fluctuation will happen.

  11. Algorithm I (Introduction) • Basic idea • First analysis and obtain the equation for traffic flow • Based on the equation, predict the congestion • Use SPSP algorithm to find the optimal path • Contribution • Combination of dynamic routing and static routing • Set up a math model for the traffic flow

  12. Algorithm I (Traffic Flow Equation) • Basic Assumptions • First study the traffic flow along one endless road without traffic lights. • Overtake is not allowed. • Basic Notations • flow q(x,t): The number of vehicles passing through point x within time period (t, t+ t). • density p(x,t): The number of vehicles located at (x, x+ x) at time t. • speed v(x,t): The speed of vehicles passing through point x at time instant t.

  13. Algorithm I (Traffic Flow Equation) • After a very complex and detailed derivation, we get the following equation (TFE): • f(x) is the initial density • It is very similar to the wave equation, but the essences are different. • The details of the derivation are provided in the final report and are omitted here due to time limitation.

  14. Algorithm I (Further Analysis) • Now we introduce the traffic lights and accidents to see their effects on the traffic flow equation: • In this case, the density will become discrete • The discrete line should satisfy • Where q and p can be calculated by the limit of q and p solved from the TFE equation

  15. Algorithm I (Further Analysis) • Again, we omit the detail procedure and only provide the most important result here: • The red light (or traffic accident) will cause two discrete line. • It would take for the traffic flow to go back to the normal, where is the initial density, is the maximum density, is the block time • Usually we have

  16. Algorithm I (Routing Scheme) • Based on the characteristics of wireless communication, we propose a dynamic and static combined routing scheme: • For the nearby roads, we use the information from the sensors. • For the faraway roads, we use the pre-stored average data • Use the equations to calculate the time consumed on each road as weight • Then perform SPSP algorithm • This can to some extent get the advantages of both static routing and dynamic routing

  17. Algorithm II • Algorithm 1 still has the fluctuation problem as it aims at the optimization of individual vehicle. • Algorithm 2 tries to improve it by using a global scheme.

  18. Algorithm II (Basic Idea) • We regard each road as a processor and each crossroad as a Weighted Fair Queueing driven by the traffic light. The vehicles running along a road are considered as tasks that are being executed by the processor of that road. Under the assumption that no overtake is allowed, when a car enters a road, all the cars ahead of it should have higher priority. In addition, all the processors are based on a fixed-priority schedule. • The weight for the output of each processor is decided by the green-light-period/red-light-period for the corresponding road. • Use the leaky-bucket regulated connection because the rate r can be used to represent the average traffic flow, and the b can be viewed as fluctuation over average and is provided in advance for each road use previous statistics. • The algorithm is composed of two steps: first use the information from the sensors in other vehicles to get the r value for the processors of interest as well as the task distribution (how many tasks are located in the processors of interest). Then we can use a variant of the Parekh-Gallager Theorem to find the maximum delay for each possible path, and select the minimum one.

  19. Algorithm II (Traffic flow to Processor Network)

  20. Outline • Motivation • Problem formulation • Algorithms • Implementations • Simulation results • Conclusions

  21. Implementations - Outline • Simulation platform – OMNeT++ • 802.11 protocol simulation package – MF • System architecture • TrafficSys architecture • Implementation details • Mobility module • Customized application layer • Traffic map generator • Display platform

  22. Implementations - Outline • Simulation platform – OMNeT++ • 802.11 protocol simulation package – MF • System architecture • TrafficSys architecture • Implementation details • Mobility module • Customized application layer • Traffic map generator • Display platform

  23. Simulation Platform OMNeT++ • An object-oriented modular discrete event simulator • Generate C++ simulation control code • Consists of modules that communicate with message passing • Modules can be nested hierarchically • Simple modules

  24. Models are expressed in terms of a topology description language NED (NEtwork Description) • Module can have parameters to customize module topology, module behavior, and module communication OMNeT++ Model Overview module TokenRingStation parameters: mac_address; gates: in: in; out: out; submodules: mac: TokenRingMAC parameters: THT=0.010, address=mac_address; gen: Generator; sink: Sink; connections: mac.to_network --> out, mac.from_network <-- in, mac.to_higher_layer --> sink.in, mac.from_higher_layer <-- gen.out; endmodule simple TokenRingMAC parameters: THT, address; gates: in: from_higher_layer, from_network; out: to_higher_layer, to_network; endsimple

  25. OMNeT++ Simulation Code Structure

  26. Mobility Framework • Support wireless and mobile simulations within OMNeT++. • The core framework implements the support for node mobility, dynamic connection management and a wireless channel model. • A library of standard protocols (802.11 etc.) is provided.

  27. Structure of a Mobile Host

  28. Our Simulation System Architecture • Built on the top of MF and OMNeT++ • Read the input from Traffic Info Generator • Display and analyze simulation results by Display platform

  29. TrafficSys Architecture

  30. Mobility module • Channel Control update connections based on the position provided by Mobility module. • Application layer decides the host position based on algorithm. • We use blackboard to share information between application layer and mobility module.

  31. An Example Host 1 Channel Control Host 2 Host 3 Host 2

  32. An Example Host 1 Host 2 Channel Control Host 3 Host 2

  33. Customized application layer

  34. System Runtime Interface 10 hosts

  35. Traffic Info Generator

  36. Display and Analysis Platform

  37. Simulation Results • Individual Prospective • Average speed (km/h) • Alg 1. can achieve 12.8% improvement; • Alg 2. can achieve 16.5% improvement.

  38. Simulation Results • Entropy (Global prospective)

  39. Efficiency Average Simulation Time (ms per decision) Simulation Results

  40. Conclusions • Formulate the sensor-less traffic-condition-based car navigation problem and propose an algorithm to solve it efficiently. • Study on OMNeT++ and Mobility Framework. • Implement a traffic system based on our algorithm on the top of MF and OMNeT++ • Simulate and analyze the results based on our statistic platform with matlab.

  41. Thank You!

More Related