Multimodal intelligent logistics robot combining 3D CNN, LSTM, and visual SLAM for path planning and control

Area of Science:

• Robotics and Automation
• Artificial Intelligence
• Logistics and Supply Chain Management

Background:

• Intelligent robots are vital for efficiency, cost reduction, and safety in dynamic logistics.
• Traditional path planning methods fail in complex, changing environments, causing collisions and conflicts.
• Enhancing robot autonomy and reliability in logistics is a significant challenge.

Purpose of the Study:

• To develop an advanced path planning and control method for logistics robots in complex environments.
• To improve obstacle avoidance capabilities by integrating multi-source perception information.
• To increase the autonomy and reliability of logistics robots.

Main Methods:

• Utilized a 3D Convolutional Neural Network (CNN) for object recognition and feature learning.
• Employed Long Short-Term Memory (LSTM) models for spatio-temporal feature extraction and dynamic obstacle prediction.
• Applied the Dijkstra algorithm for optimal path planning and control decisions.

Main Results:

• The proposed method demonstrated superior path planning accuracy compared to traditional approaches.
• Significant improvements in obstacle avoidance performance were observed.
• The intelligent path planning and control scheme proved highly effective.

Conclusions:

• The research enhances the practicality of logistics robots in complex operational settings.
• The integrated approach of object recognition, spatio-temporal modeling, and optimized path planning boosts robot precision and reliability.
• This advancement contributes to increased efficiency and safety in the logistics industry and autonomous operations.