Intersecon aucons and reservaon-based control in dynamic traffic assignment Michael W. Levin & Stephen D. Boyles Reservaon-based intersecon control increases capacity and reduces delay for single intersecons (Fajardo et al., 2011) Aucon priority may further reduce delay How are intersecon aucons affected by user equilibrium (UE) behavior on city networks? Movaon AIM4 microsimulator Accepted Rejected Intersecon model of reservaon-based intersecon control compable with general simulaon-based dynamic traffic assignment (SBDTA) Computaonally tractable for city networks Comparison of aucons with first-come-first-serve (FCFS) suggests its benefits are from the randomness of aucons Contribuons Tiles → conflict regions Computaonal results—first come first serve priority 171 zones 546 intersecons 1247 links 62836 trips Downtown Ausn, Texas Link transmission model used for SBDTA Method of successive averages used to solve DTA 922.5 seconds for 50 iteraons Esmated 150 hours for AORTA (Carlino et al., 2012) Convergence of DTA with reservaon-based intersecons For computaonal tractability, les collision checks are simplified to conflict regions—larger intersecon areas with limited capacity Turning movements pass through 1 or more conflict regions Determined by radial division of intersecon—automated method Analysis of aucons Histogram of travel mes in aucons/loery compared with FCFS 1)Vehicles communicate with the intersecon manager and request a space-me reservaon through the intersecon 2)Intersecon manager accepts or rejects reservaon based on le occupancy of other reservaons Inializaon C ij conflict regions in the path from i to j f(v) priority of vehicle v ℓ i number of lanes in i Q c (Q i ) capacity of conflict region c (link i ) S i (t) sending flow of i at me t R j receiving flow of link j at me t V vehicles that can enter the intersecon y ij (t) (y c (t) ) flow between i and j (through c ) at t Vehicle propagaon 1. Set V = Ø 2. For all incoming links i 3. Sort S i (t) by arrival me at I 4. Remove first ℓ i vehicles from S i (t) and add them to V 5. Sort V by f(v) 6. For all v ϵV 7. Let (i, j) be the origin/desnaon links of v 8. If R j (t) - ∑ i’ y i’j (t) ≥ 1 and Q c - y c (t) ≥Q ij /Q c for all c ϵC ij 9. y ij (t) := y ij (t) + 1 10. For cϵC ij : y c (t) := y c (t) + Q ij /Q c 11. Remove first vehicle in S i (t) and add it to V 12. Go to 5 Receiving flows Intersecon algorithm Vehicle propagaon Sending flows Vehicle priority Background Properes Greater use of intersecon—including simultaneous use by conflicng turning movements Flexible priority strategies—FCFS, aucons, etc. Requires microsimulaon of intersecons. Previous work on networks of intersecons was limited in size or used a single le, and did not consider UE behavior Objecves Admit arbitrary priority strategies Retain simultaneous use by vehicles with conflicng paths Independent of specific intersecon characteriscs Sasfy invariance principle (Tampère et al., 2011) Aucon experiment Vehicles bid value of me (VOT) at each intersecon—highest bidder gets priority VOTs based on income distribuon Loery experiment Each vehicle is assigned a random number that is their priority on Sioux Falls network Assumpons Flow is discrezed to assign vehicle priority All vehicles have the same physical characteriscs In the absence of other demand, flow is restricted only by sending and receiving flows (to be independent of geometry) Conclusions Conflict region model for SBDTA of reservaon-based intersecon control for autonomous vehicles Compable with general SBDTA and computaonally tractable for large city networks Builds on characteriscs of general DTA intersecon models (Tampère et al., 2011): First-in-first-out behavior within links Sasfies invariance principle Dependent on intersecon geometry due to conflict regions, but conflict region division is automated Link transmission model (LTM) with conflict regions converges to dynamic user equilibrium Aucons reduce congeson over FCFS, but the effects are due to the randomness of bids: loery has similar results Future work Comparison of traffic signals and reservaon-based control under user equilibrium behavior DTA model of shared roads (human drivers and autonomous vehicles) Opmal priority strategies for reservaon-based control Possibility of Braess paradox-like phenomena due to higher capacity and/or reservaon priority Lile to no benefit for high VOT vehicles from aucons Intersecon delay increased but congeson decreased, leading to a net benefit Comparison of queue lengths indicates that FCFS creates large queues on high demand links Queue length (FCFS) Queue length (aucons) FCFS allows queues to build on high-demand links because priority is independent of queue size The randomness of aucons (and loery) results in a more even distribuon Intersecon algorithm