AMR Sensor Fusion for Flexible Load Pick and Drop
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous mobile robots (AMRs) are limited to interacting with loads at precisely defined locations, requiring manual specification and precise positioning, restricting their ability to pick or drop loads outside small, well-defined areas.
Innovation Solution
An AMR system utilizing a combination of planar scanners, a paddle sensor, and a camera-based pallet detection system (PDS) to dynamically adjust its behavior for engaging and dropping loads, allowing operation in arbitrarily long straight-line regions by selectively fusing sensor data from multiple sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional AMRs interact with loads at precisely defined locations requiring manual specification, then positioning accuracy is improved, but operational flexibility deteriorates
Solution Approach 1:
The system dynamically switches between low-fidelity and high-fidelity sensing modes based on operational context. During approach, low-fidelity sensors provide fast, coarse localization. When near the load, high-fidelity sensors activate for precise positioning. This dynamic adaptation resolves the contradiction by allowing operational flexibility through mode switching while maintaining positioning accuracy when needed.
Solution Approach 2:
The sensing system is segmented into multiple fidelity levels with different sensors dedicated to different stages of the interaction. Low-fidelity sensors (e.g., planar scanners) handle long-range detection, while high-fidelity sensors (e.g., cameras, paddle sensors) handle close-range precise measurement. This segmentation allows each sensor type to operate in its optimal range, achieving both flexibility and precision.
2Adaptability or versatility
If multiple sensors are fused to enable interactions in regions of arbitrary length, then operational flexibility is improved, but system complexity increases
Solution Approach 1:
The system employs dynamic sensor fusion where not all sensors are actively processed simultaneously. Instead, the system activates specific sensor subsets based on the operational phase (approach vs. interaction). This reduces computational complexity while maintaining the ability to operate in arbitrary regions by selectively fusing relevant sensor data.
Solution Approach 2:
The sensor fusion process is segmented into distinct processing stages. Low-fidelity sensor data is processed first for coarse localization and region identification. High-fidelity sensor data is then processed separately for precise interaction. This staged segmentation reduces overall system complexity by avoiding simultaneous processing of all sensors while still achieving flexible operation across arbitrary regions.
3Measurement precision
If high-fidelity sensing is used continuously for precise load engagement, then positioning accuracy is improved, but energy consumption increases
Solution Approach 1:
The system uses periodic switching between low-fidelity and high-fidelity sensing modes rather than continuous high-fidelity sensing. During the approach phase, low-fidelity sensors operate continuously. High-fidelity sensors are activated periodically only when needed for precise engagement and disengagement. This periodic action significantly reduces energy consumption while maintaining positioning accuracy at critical moments.
Solution Approach 2:
The sensing fidelity is dynamically adjusted based on the operational phase. The system transitions from low-fidelity to high-fidelity sensing only when the robot is in the critical engagement zone. This dynamic adjustment ensures high positioning accuracy is achieved only when necessary, minimizing overall energy consumption while maintaining operational effectiveness.
Data Source
AI summary
An autonomous mobile robot (AMR) is provided, comprising a chassis, a navigation system, and a load engagement portion, a plurality of sensors, including an object detection sensor, a load identification sensor, and a load presence sensor, and a load interaction system. The load detection system is configured to exchange information with the plurality of sensors and configured to operate in a load engagement mode and a load drop mode by selectively fusing data from among the plurality of sensors in each mode.