Autonomous Mobile Device Sensor Fusion for Load Carrier Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated industrial trucks and robots face challenges in accurately recognizing and orienting themselves relative to load carriers like pallets or wire mesh boxes for efficient handling and transport, particularly in dynamic warehouse environments.
Innovation Solution
A mobile device equipped with movement actuators, image capture devices, LiDAR sensors, and a processor implementing reinforcement learning neural networks to autonomously navigate and orient itself relative to load carriers, using a combination of image and LiDAR data to determine optimal control actions for movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image recognition methods are used for load carrier recognition and orientation, then the system structure is simple, but the recognition accuracy and orientation precision are insufficient in dynamic warehouse environments
Solution Approach 1:
The patent combines multiple sensing modalities (image capture device and LiDAR sensor) into an integrated perception system. The image capture device provides visual information about load carriers while the LiDAR sensor provides depth and spatial information, and their data are processed together by the processor to achieve accurate recognition and orientation in dynamic warehouse environments.
Solution Approach 2:
The patent replaces traditional mechanical or simple optical recognition systems with an advanced sensor fusion system comprising image capture devices, LiDAR sensors, and a processor that implements complex algorithms for perception and navigation, enabling superior recognition accuracy despite increased system complexity.
2Productivity
If the mobile device navigates autonomously in dynamic warehouse environments, then the operational efficiency is improved, but the risk of collisions with obstacles and load carriers increases
Solution Approach 1:
The processor plans a navigation path for the mobile device to move from its current location to the target location. This path planning is performed in advance before the mobile device begins movement, allowing the system to anticipate and avoid obstacles and load carriers, thereby reducing collision risk while maintaining operational efficiency.
Solution Approach 2:
The system continuously captures images and LiDAR data during navigation, processes this sensory information in real-time, and adjusts the navigation path accordingly. This feedback loop enables the mobile device to respond to dynamic changes in the warehouse environment, avoiding obstacles and load carriers while maintaining efficient operation.
3Measurement precision
If multiple sensors and neural networks are integrated for autonomous navigation, then the navigation accuracy is improved, but the computational load and processing time increase
Solution Approach 1:
The processor plans the navigation path in advance before the mobile device begins movement. This preliminary path planning allows the system to process complex sensor data and compute optimal routes ahead of time, reducing real-time processing requirements and enabling accurate navigation without excessive delays.
Solution Approach 2:
The system captures images and LiDAR data at discrete time intervals during navigation, processes this periodic sensor input, and updates the navigation path accordingly. This periodic processing approach balances navigation accuracy with computational efficiency by not requiring continuous processing.
Data Source
Figure 1
Figure 2a
Figure 2b
AI summary
The invention relates to a mobile device (110), in particular a robot for the automated transport of a load carrier (170). The mobile device (110) comprises at least one motion actuator (160) configured to be controlled by means of control data in order to move the mobile device (110), an image acquisition device (130) for acquiring a plurality of images of a section of the environment of the mobile device (110) at different times, and at least one LiDAR sensor (140) for acquiring LiDAR data in the environment of the mobile device (110).Furthermore, the mobile device (110) comprises a processor (120) which is configured to implement a reinforcement learning (RL)-trained neural network, wherein the RL-based neural network is configured to determine, on the basis of the multitude of images, the LiDAR data, a multitude of past control data and a multitude of past rewards, the control data for controlling the at least one motion actuator (160) in order to move the mobile device (110).