Online Trajectory Planning Method for Autonomous Ground Vehicles Confronting Sudden and Moving Obstacles Based on...

Area of Science:

• Robotics
• Artificial Intelligence
• Control Systems

Background:

• Autonomous ground vehicles (AGVs) require robust obstacle avoidance for safe operation.
• Existing methods often struggle with real-time performance and dynamic environments.

Purpose of the Study:

• To develop a novel online trajectory planning method for AGVs to handle sudden and moving obstacles.
• To ensure high real-time performance and optimality in obstacle avoidance maneuvers.
• To enhance the applicability of AGV obstacle avoidance systems.

Main Methods:

• Utilizing long short-term memory-attention (LSTM-Attention) networks for trajectory planning.
• Offline training of the LSTM-Attention network using a generated AGV obstacle avoidance trajectory dataset.
• Implementing a rotation coordinate system to manage obstacles in various directions and moving obstacles.

Main Results:

• The LSTM-Attention network effectively learns the mapping between vehicle-obstacle information and optimal control actions.
• The proposed method achieves real-time performance and optimal trajectory planning for AGVs.
• The rotation coordinate system significantly broadens the method's applicability to diverse scenarios.
• Validation through extensive simulations and physical experiments confirms effectiveness.

Conclusions:

• The novel LSTM-Attention based method provides an effective and efficient solution for AGV obstacle avoidance.
• The approach demonstrates strong real-time capabilities and adaptability to dynamic environments.
• This research contributes to the advancement of safe and reliable autonomous navigation systems.