Research
ADDRESSING ACCIDENT-IMMINENT SITUATIONS INVOLVING AUTONOMOUS VEHICLES IN MIXED TRAFFIC: A COOPERATION-BASED METHODOLOGY
Keywords: Autonomous vehicle, Vehicle-to-vehicle connectivity, collision avoidance, model predictive control (MPC), optimal control, cooperative control, safety
Perspective-II (when HDV at fault) human error is considered in the mixed traffic environment
Cooperative control framework is generated using MPC to control the longitudinal acceleration/deceleration of CHDV and CAV
Lane-changing HDV's trajectory is modeled by cubic polynomial function
Two crash avoidance maneuvers are generated based on two calssic collision patterns: rear-end collision and side-impact collision
The success rate of combining two maneuvers reaches exceed 90%
Connectivity enhances the safety benefits in crash avoidance espeacially in imminent situations
URBAN TRAFFIC DYNAMIC REROUTING FRAMEWORK: A DRL-BASED MODEL WITH FOG-CLOUD ARCHITECTURE
Keywords: Dynamic Rerouting, Deep Q Learning, Graph Attention, Fog-Cloud Architecture, Multi-agent System
Fog layer is applied to the urban transportation network to collect information with lower latency
Double-layered rerouting framework is proposed:
DRL: Q network + Graph Attention Mechanism (GAT) generate road index to calculate the final road weights
EBkSP to assign different route to rerouting vehicles (RV) based on their priority and route popularity
