AI in Non-connected AVs Slows Down Traffic, Study Indicates

KEY POINTS

• Connected vehicles improve travel time at intersections.
• Non-connected automated vehicles (AVs) can slow intersection times.
• Safety programming leads automated vehicles to be conservative.

Recent research indicates that while vehicles that wirelessly share data enhance travel times through intersections, autonomous vehicles that lack this connectivity can result in decreased intersection efficiency. A central reason for this is the safety-focused programming of these vehicles. Ali Hajbabaie, an associate professor at North Carolina State University, and his team utilized computational modeling to derive these insights.

The team employed a computational model simulating various traffic conditions. They considered four vehicle types:

1. Human-driven vehicles (HVs)
2. Connected vehicles (CVs) which are human-driven but share data,
3. Automated vehicles (AVs)
4. Connected automated vehicles (CAVs).

The fundamental behavior of AVs is to operate with caution due to their programming, emphasizing safety. In contrast, CVs and CAVs can anticipate future traffic light states, enabling smoother movement and fewer stops than HVs and AVs.

Through 57 traffic simulations, the team evaluated the influence of different variables on intersection travel time. These simulations examined traffic implications of various combinations of HVs, AVs, CVs, and CAVs. A principal observation was that as the number of CVs and CAVs increased, the capacity of intersections also improved, allowing for a more efficient flow of traffic. However, a rise in the number of AVs, which aren’t connected, led to longer travel times at intersections, attributable to their conservative driving behavior.

Hajbabaie emphasized the preliminary nature of their findings, noting that they were derived from computational models. While such models offer valuable insights, field tests involving real vehicles pose challenges in terms of cost, feasibility, and safety considerations.