Learning-Embedded Motion Planning for Real-Time Box Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Contemporary robotics systems face challenges in efficiently and accurately planning motion paths for robot box pick and place operations, especially in cluttered environments, due to high computational demands and the need for real-time processing, which existing methods struggle to address.
Innovation Solution
A learning-embedded motion planning technique that separates the process into an intuition learning module and a guided motion planning module, utilizing neural networks to bias sample distributions and generate high-quality paths efficiently, enabling real-time box handling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional motion planning methods are used, then motion planning can be performed, but computational time is excessive and real-time processing cannot be achieved
Solution Approach 1:
The motion planning process is segmented into two distinct modules: an intuition learning module that pre-computes and stores motion intuition data, and a guided motion planning module that uses this pre-computed data to rapidly generate motion plans. This segmentation allows the system to avoid re-computing complex motion intuition from scratch during real-time operations, thereby reducing computational time while maintaining planning speed.
Solution Approach 2:
The system performs preliminary action by pre-computing motion intuition data and storing it in a database during an offline training phase. During real-time operation, the guided motion planning module queries this pre-computed data rather than performing complex computations from scratch. This preliminary action enables the system to achieve real-time processing speeds while maintaining high-quality motion planning.
2Measurement precision
If comprehensive motion planning is performed, then planning accuracy is improved, but computational complexity increases
Solution Approach 1:
The system extracts and separates the complex motion intuition computation from the real-time planning process. The intuition learning module extracts and pre-computes motion intuition data, which is then stored and reused during real-time operations. This extraction allows the guided motion planning module to achieve high accuracy without the computational complexity of re-computing motion intuition from scratch, as it simply queries pre-computed results from the database.
Solution Approach 2:
The system uses copying by creating a database of pre-computed motion intuition data that can be queried and reused. Instead of performing complex computations anew for each planning task, the system copies relevant pre-computed intuition data from the database, thereby maintaining high planning accuracy while significantly reducing computational complexity during real-time operation.
3Speed
If real-time processing is required, then responsiveness is improved, but planning quality may deteriorate
Solution Approach 1:
The system performs preliminary action by pre-computing and storing motion intuition data during an offline training phase. This pre-computed data is then available for immediate query during real-time operations. By separating the heavy computational work into the offline phase and the query/retrieval work into the online phase, the system achieves both real-time processing speed and high planning quality without compromise.
Solution Approach 2:
The system introduces an intermediary mechanism in the form of a database that stores pre-computed motion intuition data. This intermediary allows the guided motion planning module to access high-quality planning information without performing complex computations in real-time. The database acts as a mediator between the offline training process and online planning operations, enabling real-time responsiveness while maintaining planning quality through pre-computed results.
Data Source
AI summary
Method of sequential handling of a plurality of freight units by an end effector of a robotic system comprises training a neural network operable to bias a plurality of samples of a configuration space of the end effector. Deployment of the trained neural network involves sensing an arrangement of freight units in a scene based on an image of the scene from a sensor. A freight unit is chosen based on the at least one trained neural network. A starting pose of the robotic end effector as a means to pick the chosen freight unit is chosen; and a goal pose of the robotic end effector is chosen as a means to place the chosen freight unit. A motion of the robotic end effector is planned from the starting pose to the goal pose based on a biased plurality of samples. Then motion on the end effector is controlled based on the planning.