Crossroads: Autonomous Intersection Management
Partially funded by National Science Foundation grants CCF 1055094 (CAREER), CNS 1525855, NIST 60NANB16D305, & NIST 60NANB15D322
The introduction and adoption of autonomous vehicles will require intelligent autonomous intersection management. Intersection management includes maintaining a safety buffer around the vehicle while factoring variable network and computational time delays. Modeling these network and computational delays can degrade the throughput of the intersection even when processed by a state-of-the-art intersection manager.
To combat these limitations, ASU researchers have developed Crossroads, a time-sensitive method to program the interface of a vehicle and intersection that does not require an additional buffer to account for network and computational delays.
Using a small-scale intersection model with radio control cars, the Crossroads approach demonstrated a reduction in average vehicle wait time at a single-lane intersection of 24 percent. Crossroads obviates the need for large buffers to accommodate network and computational delays.
Through extensive MATLAB simulations, Crossroads has shown increased throughput compared to existing approaches: 1.62X higher throughput (on average) than Velocity Transaction Intersection Management (VT-IM) with extra safety buffers, and 1.36X higher throughput (on average) than autonomous intersection management (AIM).
For more information, including publication in the 54th-annual Design Automation Conference (DAC), see http://aviral.lab.asu.edu/.