Near-Optimal Reactive Synthesis Incorporating Runtime Information.

Area of Science:

• Robotics and Control Systems
• Artificial Intelligence
• Operations Research

Background:

• Optimal reactive synthesis aims to create strategies for dynamic environments that optimize performance metrics.
• Incorporating real-time, task-critical information into strategy synthesis is challenging due to computational demands of online re-synthesis.

Purpose of the Study:

• To develop a computationally feasible method for optimal reactive synthesis that utilizes runtime information.
• To improve strategy performance by integrating dynamic, task-critical data without requiring real-time re-synthesis.

Main Methods:

• Pre-synthesizing a set of strategies based on candidate instantiations (representative information scenarios).
• Proposing a novel switching mechanism for dynamic, safe, and live-goal-guaranteed strategy transitions at runtime.
• Characterizing bounds on performance suboptimality.

Main Results:

• Demonstrated a novel approach for optimal reactive synthesis in dynamic environments.
• Successfully integrated runtime information into strategy synthesis, improving performance.
• Guaranteed safety and liveness during dynamic strategy switching.

Conclusions:

• The proposed method enables efficient, real-time adaptation of strategies in dynamic environments.
• This approach offers a practical solution for problems requiring optimal reactive synthesis with time-varying information.
• The technique is applicable to complex scenarios like robotic motion planning and air traffic management.