Limited Visibility Aware Motion Planning for Autonomous Valet Parking Using Reachable Set Estimation.

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Vision

Background:

• Autonomous driving offers convenience but parking remains a challenge.
• Current Automated Valet Parking (AVP) systems require driver monitoring due to sensor limitations and occlusions in parking lots.
• Limited fields of view in parking lots hinder obstacle detection from occluded areas.

Purpose of the Study:

• To propose a novel method for risk estimation in Automated Valet Parking (AVP) systems.
• To address the challenge of unobservable obstacles in occluded parking lot environments.
• To enhance the safety and efficiency of AVP systems by proactively considering limited visibility.

Main Methods:

• Developing a risk estimation technique based on worst-case assumptions for unobservable obstacles.
• Integrating occlusion awareness directly into the motion planning process for AVP systems.
• Evaluating the proposed method through extensive simulations under various parking lot scenarios.

Main Results:

• The proposed method effectively incorporates occlusion into the planning process.
• The system can ensure safety without exhibiting excessive caution.
• Simulations demonstrate the proactive nature of the approach in managing limited visibility.

Conclusions:

• The developed risk estimation method offers a proactive solution for AVP systems operating in environments with occlusions.
• This approach enhances the reliability and safety of autonomous parking by addressing sensor limitations.
• The method shows significant potential for improving the practical deployment of AVP technology.