Robotic Grasping Control With Neural Feedback to Handle Perturbations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robots face challenges in successfully grasping objects due to variability in environmental conditions and inaccurate robotic actuation, which existing end effector technologies struggle to address effectively.
Innovation Solution
A deep neural network, such as a convolutional neural network (CNN), is trained to predict the probability of successful grasping based on candidate motion data and environmental images, enabling iterative updates in motion control commands for actuators to improve grasping performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional end effector technologies are used for grasping objects, then the robot can perform basic grasping operations, but the system fails to provide fast feedback to perturbations and is vulnerable to inaccurate actuation
Solution Approach 1:
The patent implements a feedback mechanism where the deep neural network continuously predicts grasp success probability based on current sensor data and motion commands. This feedback loop allows the system to detect perturbations and actuation inaccuracies in real-time, adjusting subsequent commands to maintain reliable grasping operations despite environmental variability and robotic imperfections.
Solution Approach 2:
The patent replaces traditional mechanical control systems with a data-driven deep neural network approach. Instead of relying on complex mechanical feedback mechanisms and precise actuation hardware, the system uses software-based prediction and adaptation, substituting mechanical complexity with computational intelligence to achieve robust grasping.
2Reliability
If the robot uses iterative updates in motion control commands to improve grasping performance, then grasping reliability improves, but computational processing time increases
Solution Approach 1:
The deep neural network is trained offline in advance on large datasets of grasp scenarios, pre-learning the complex mappings between sensor inputs, motion commands, and grasp outcomes. This preliminary training allows the network to make rapid predictions during actual grasping operations without requiring extensive real-time computation, thus maintaining both reliability and speed.
3Measurement precision
If the deep neural network is trained on comprehensive training examples including sensor outputs and grasp outcomes, then prediction accuracy improves, but training data requirements and processing complexity increase
Solution Approach 1:
The patent employs a universal deep neural network architecture that can process multiple types of sensor inputs (images, depth data, force-torque measurements) and predict multiple outcomes (grasp success probability, contact point predictions, force predictions) using the same model. This multi-functionality reduces the need for separate specialized models for each sensor type or outcome, thereby reducing overall training data requirements while maintaining high prediction accuracy across diverse grasp scenarios.
Data Source
AI summary
Deep machine learning methods and apparatus related to manipulation of an object by an end effector of a robot. Some implementations relate to training a deep neural network to predict a measure that candidate motion data for an end effector of a robot will result in a successful grasp of one or more objects by the end effector. Some implementations are directed to utilization of the trained deep neural network to servo a grasping end effector of a robot to achieve a successful grasp of an object by the grasping end effector. For example, the trained deep neural network may be utilized in the iterative updating of motion control commands for one or more actuators of a robot that control the pose of a grasping end effector of the robot, and to determine when to generate grasping control commands to effectuate an attempted grasp by the grasping end effector.