Loading...

Autonomous Vehicles Network Design To Improve Network Performance

Abedi, Mohammad Ali | 2020

421 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53511 (09)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Zokaei Aashtiani, Hedayat
  7. Abstract:
  8. Nowadays, with the advent of Autonomous Vehicles(AV), Researchers in different areas focused more on this type of vehicle. Research shows that autonomous vehicles are faster and more precise than Human-driven vehicles(HDV). So if all the users in a road use autonomous Vehicles, the road’s capacity increases; this Thesis develops a mathematical approach for designing AV exclusive lanes for a time horizon in which there are two types of AVs and HDVs on the road. Considering each type of vehicle’s relative performance for an OD pair in the given year, the demand for that OD pair for the next year calculates with a diffusion model. Designing AV exclusive lanes problem is formulated in a bi-level model. For the upper level, we determine design variables (e.g., when, where, and how many AV lanes should be deployed), and the lower level is a multi-class traffic assignment problem. This problem was first introduced by Chen et al., they used the Active-set algorithm for solving the problem. This algorithm needs a good initial solution, and results show that the final solution is not reliable enough and optimum, and the algorithm is not suitable for big problems. So this Thesis develops a new approach for this problem; in which, for the lower level problem, we rewrite the model as a complementary problem “based on paths,” and we use the Linearization algorithm introduced by Aashtiani for solving the lower level problem. For the upper-level problem, we introduce a new heuristic algorithm. Results in the South-Florida network show that this approach, without an initial solution, finds a better solution in a short time, and so can be used for big problems. For instance, we used this algorithm for solving the South-Florida network in the same horizon time, with all 232 links of the network as candidate links instead of 44 candidate links by Chen et al. This algorithm solved this problem in 52 minutes runtime and improved the total social cost by 15%
  9. Keywords:
  10. Autonomous Vehicles (AVs) ; Network Design ; Two-Class Traffic Assignment ; Ashtiani Linearization Algorithm ; Autonomous Vehicle Lane ; Performance Improvement ; Transportation Network

 Digital Object List

 Bookmark

No TOC