Human-inspired emotion and attention encoding for autonomous vehicles' decision-making: a lane-change timing...

Area of Science:

• Autonomous Systems
• Cognitive Science
• Neuroscience

Background:

• Autonomous vehicle (AV) lane-changing in mixed traffic is challenging due to coordination requirements.
• Conservative AV strategies can reduce efficiency and cause disruptions.
• Human-like decision-making is needed for flexible and reliable AV lane-changes.

Purpose of the Study:

• To develop an emotion and attention encoding framework for human-like AV lane-changing.
• To enable AVs to adaptively optimize lane-change timing in complex traffic scenarios.

Main Methods:

• Utilized Cognitive Energy Theory, Attenuator Theory, and Prospect Theory.
• Constructed neural encoding processes for driver arousal, experience, attention, and emotional utility.
• Introduced an emotional utility model (EUM) and a human-like lane-changing decision (HLD) method.

Main Results:

• The HLD method achieved a lane-change rate exceeding 99.8% under the 3-sigma rule.
• Lane-change timings closely matched real driver behavior while enhancing safety.
• The EUM adaptively adjusted risk-weighting for improved utility and risk balancing.

Conclusions:

• The proposed framework enables human-like lane-changing for AVs, improving efficiency and safety.
• The EUM is key to balancing risks and optimizing decisions in dynamic environments.
• This approach offers insights for AV decision-making in diverse complex scenarios.