Robot Handover Control Using Two-Stage Neural Network Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for controlling robots face challenges in optimizing memory, time, and computing resources, particularly in human-robot collaboration tasks like handovers, where sparse input data and complex human behavior require efficient processing and coordination.
Innovation Solution
A framework utilizing neural networks and a two-stage training process, combining perception and control modules with physics simulation, allows for the development of control policies that can effectively manage robot interactions with humans, leveraging simulation data to improve real-world performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional control systems are used for robot-human interaction, then the system structure is simple, but memory usage and computing resources are excessive
Solution Approach 1:
The control system is segmented into multiple neural networks with specialized functions: a first neural network for detecting human presence and intent, and a second neural network for generating robot control commands. This segmentation allows each network to process only relevant data, reducing overall memory usage while maintaining sophisticated interaction capabilities.
Solution Approach 2:
The patent extracts and separates the human interaction processing functions from the general robot control system. By dedicating specific neural networks to human presence detection and intent recognition, the system eliminates the need for a monolithic control structure, thereby reducing memory requirements for general-purpose processing.
2Measurement precision
If complex human behavior is processed in real-time, then interaction accuracy is improved, but processing time increases
Solution Approach 1:
The first neural network continuously monitors and detects human presence and intent in advance, preparing the system for upcoming interactions. By pre-processing human behavior data and identifying intent before the actual interaction occurs, the system reduces the computational burden during critical interaction moments, thereby maintaining high accuracy without excessive processing delays.
3Productivity
If sparse input data is used for control decisions, then data processing efficiency is improved, but control reliability decreases
Solution Approach 1:
The patent transforms sparse input data into meaningful control decisions by changing the parameters of processing through specialized neural network architectures. The networks are trained to extract critical features from limited data, converting sparse inputs into reliable control signals through learned patterns and relationships, thereby maintaining control reliability without requiring extensive data processing.
Data Source
AI summary
Apparatuses, systems, and techniques to control a real-world and/or virtual device (e.g., a robot). In at least one embodiment, the device is controlled based, at least in part on, for example, one or more neural networks. Parameter values for the neural network(s) may be obtained by training the neural network(s) to control movement of a first agent with respect to at least one first target while avoiding collision with at least one stationary first holder of the at least one first target, and updating the parameter values by training the neural network(s) to control movement of a second agent with respect to at least one second target while avoiding collision with at least one non-stationary second holder of the at least one second target.