Multi-Bin Parcel Loading With RL-Based Robot-Feasible Packing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional 3D bin packing algorithms struggle to efficiently pack objects in containers when container dimensions are unknown in advance, and existing robotic systems often fail to execute packing strategies due to physical constraints, leading to inefficiencies and potential collisions.
Innovation Solution
A processor-based method using reinforcement learning to determine optimal orientations and locations for object placement within containers, considering robotic arm capabilities to ensure feasible and efficient packing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional 3D bin packing algorithms are used, then packing can be performed, but packing efficiency is low (approximately 65%) and the number of containers required is high (16 containers)
Solution Approach 1:
The system dynamically adapts the packing strategy based on real-time container state and object characteristics. The reinforcement learning agent continuously learns and adjusts packing decisions during the packing process, rather than following a fixed predetermined sequence, enabling higher packing efficiency of 82% and reducing container usage from 16 to 13 containers.
Solution Approach 2:
The system implements feedback mechanisms where the state of the container is continuously monitored and used to inform subsequent packing decisions. The reinforcement learning agent receives feedback about the current container configuration and object properties to optimize placement decisions, achieving improved packing efficiency and reduced container requirements.
2Ease of operation
If traditional packing methods are used, then packing decisions can be made, but they fail to account for robotic arm physical constraints leading to infeasible placements and potential collisions
Solution Approach 1:
The system incorporates robotic arm capability parameters directly into the packing decision-making process. The reinforcement learning agent learns to account for physical constraints such as arm reach, orientation limits, and collision boundaries by training with these parameters, ensuring that generated packing sequences are executable and safe.
Solution Approach 2:
The reinforcement learning agent acts as an intermediary between the packing problem and the robotic arm execution. It translates high-level packing objectives into robot-executable actions by considering physical constraints, serving as a mediator that ensures feasibility and safety of the packing operations.
3Adaptability or versatility
If offline bin packing methods are used, then complete packing sequences can be predetermined, but they cannot adapt to dynamic scenarios where container dimensions are not available beforehand
Solution Approach 1:
The system uses self-service through reinforcement learning where the agent learns to handle dynamic container dimensions autonomously without requiring pre-programmed adaptation rules. The agent develops its own strategies for adapting to unknown container sizes through training, reducing the need for complex external adaptation mechanisms.
Solution Approach 2:
The system performs preliminary training actions in a simulated environment where various container dimensions and object configurations are exposed to the reinforcement learning agent. This preliminary learning enables the system to adapt to dynamic real-world scenarios without requiring complex runtime adaptation mechanisms.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
State of the art automated bin packing systems fail to handle dynamic scenarios in which information on dimensions of objects to be loaded is not available in advance. These systems also fail to consider capabilities of robots used for the automated packing of objects/bins. The disclosure herein generally relates to automated bin packing, and, more particularly, to a system and method for autonomous multi-bin parcel loading system. The system handles an online object packing in which information on dimensions of objects to be loaded is not available in advance. The system is also configured to consider capabilities of one or more robots used for loading objects to containers, while generating recommendations for object packing.