Robot Backup Control for Continuous AI Training Under Task Uncertainty
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robotic training methods face challenges in handling uncertain environments and failures in artificial intelligence (AI) tasks, particularly in generating scalable training data and providing flexible solutions for complex tasks, and existing augmented reality (AR) solutions are limited by computational requirements and internet connectivity.
Innovation Solution
A robot system equipped with sensors and a processor that uses a trained neural network to assess task completion probability, requests operator assistance via communication, and generates training data for AI model updates, allowing for remote control and offline data recording and synchronization, utilizing a cloud-based simulation environment and neural network framework for smart picking tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learning driven program generation is used to solve individual tasks with dynamic motions, then robot flexibility in handling unknown objects is improved, but training data availability and scalability deteriorate
Solution Approach 1:
The patent creates virtual copies of real-world scenarios through simulation environments. Instead of collecting extensive real training data, the system generates synthetic training data by copying and replicating task scenarios in virtual reality, allowing unlimited scalable training data generation without physical resource constraints
Solution Approach 2:
The system performs preliminary training actions in virtual simulation environments before deploying robots to real tasks. By pre-training AI models in simulated settings with generated training data, the robots acquire skills and adaptability without requiring extensive real-world trial and error
2Ease of operation
If AR solutions with real-time object detection and 3D hologram overlay are implemented, then technician guidance capability is improved, but computational performance requirements and tracking reliability deteriorate
Solution Approach 1:
The patent extracts and separates the computationally intensive AR processing functions from the wearable device. Instead of performing all 3D hologram rendering and object tracking computations in the smart glasses, the system offloads these tasks to external computing resources, reducing the computational burden on the wearable device
Solution Approach 2:
The system introduces an intermediary computing layer between the AR display and the robot/system being guided. This intermediary handles the heavy computational tasks of 3D rendering and tracking, acting as a mediator that protects the wearable device from excessive computational requirements while maintaining high-quality AR guidance
3Ease of operation
If remote guidance solutions requiring stable Internet connection are used, then real-time operator assistance is improved, but offline capability and synchronization deteriorate
Solution Approach 1:
The system prepares and buffers guidance data and communication protocols in advance for offline operation. By pre-loading necessary information and establishing communication protocols before disconnection occurs, the system maintains operational continuity and reliability during periods without Internet connectivity
Solution Approach 2:
The system dynamically adapts its operation mode based on connectivity status. It seamlessly transitions between real-time connected mode and offline autonomous mode, adjusting its functionality and data processing strategies according to the available network conditions, ensuring continuous operational reliability
4Manufacturing precision
If native robot programming with teaching pendants is used to write programs directly on robot controller, then programming precision is improved, but task flexibility and scalability deteriorate
Solution Approach 1:
The patent segments the programming system into multiple layers: high-level task planning separated from low-level execution commands. This segmentation allows precise control at the execution level while maintaining flexibility at the task planning level, enabling both precision and adaptability simultaneously
Solution Approach 2:
The system introduces an additional dimension of AI-driven autonomous decision-making alongside traditional programmed control. Instead of relying solely on pre-programmed sequences, the robot operates in a multi-dimensional control space that combines deterministic programming with probabilistic AI reasoning, enabling flexibility without sacrificing precision
Data Source
AI summary
Provided are systems and methods for training a robot. The method commences with collecting, by the robot, sensor data from a plurality of sensors of the robot. The sensor data may be related to a task being performed by the robot based on an artificial intelligence (AI) model. The method may further include determining, based on the sensor data and the AI model, that a probability of completing the task is below a threshold. The method may continue with sending a request for operator assistance to a remote computing device and receiving, in response to sending the request, teleoperation data from the remote computing device. The method may further include causing the robot to execute the task based on the teleoperation data. The method may continue with generating training data based on the sensor data and results of execution of the task for updating the AI model.