Under-Canopy Robot Navigation Using Multi-Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Agricultural robots face challenges in navigating through fields with crop canopies that interfere with GPS signals, leading to inaccurate navigation and potential damage to crops and robots, as well as issues with traction that can cause robots to get stuck.
Innovation Solution
A system that determines navigation modes based on sensor data and location using a supervision model, switching between under-canopy, out-row, and recovery modes, and generates reference maps for path navigation, utilizing different sensors such as GNSS, RTK, LiDAR, and kinetic data to ensure accurate and safe robot movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPS navigation is used for agricultural robots, then the robot can navigate in open areas, but the navigation accuracy deteriorates under crop canopies due to signal interference
Solution Approach 1:
The navigation system is segmented into multiple independent subsystems: GPS/GNSS for open-area navigation, LiDAR for under-canopy navigation, and visual odometry for supplementary positioning. Each subsystem operates independently in its optimal environment, with the system switching between them based on canopy density detection, thereby maintaining navigation reliability while avoiding positioning accuracy deterioration in any single condition
Solution Approach 2:
The robot employs a universal navigation framework that integrates multiple positioning methods (GPS, LiDAR, visual odometry) into a single system. This multi-functional approach allows the robot to navigate effectively in both open areas and under crop canopies, eliminating the trade-off between navigation reliability and positioning accuracy by having each method serve different operational contexts
2Productivity
If the robot navigates under crop canopies, then it can access rows for agricultural tasks, but the risk of collision with crops and obstacles increases
Solution Approach 1:
The LiDAR system performs preliminary scanning of the environment ahead of the robot's path, creating a forward-looking collision detection zone. This preliminary action identifies potential obstacles and crop rows before the robot reaches them, allowing the navigation system to adjust the trajectory in advance and prevent collisions while maintaining productivity in under-canopy environments
Solution Approach 2:
The system implements continuous feedback loops where LiDAR and camera sensors constantly monitor the robot's surroundings, compare detected features with the reference map, and provide real-time feedback to the path tracking controller. This feedback mechanism enables dynamic collision avoidance by adjusting navigation commands based on actual environmental conditions, reducing collision risk while allowing efficient row access
3Measurement precision
If the robot uses multiple sensors for navigation, then navigation accuracy improves, but the system complexity increases
Solution Approach 1:
The system dynamically adjusts sensor activation and data processing based on operational context. In open areas, only GPS is active with minimal processing. Under canopies, LiDAR and cameras are activated with increased processing frequency. This dynamic adaptation maintains positioning accuracy where needed while reducing computational complexity and power consumption in simpler environments
Solution Approach 2:
The navigation system extracts and processes only the necessary sensor data for each operational mode. In under-canopy navigation, it extracts geometric features from LiDAR point clouds and visual features from camera images, discarding redundant information. This selective extraction maintains positioning accuracy while minimizing the computational burden and system complexity
4Measurement precision
If the robot follows a strict navigation path, then navigation accuracy is maintained, but the time to traverse the field increases due to course corrections
Solution Approach 1:
The system performs preliminary path planning that anticipates upcoming turns and adjustments based on the reference map, smoothing the navigation path before the robot reaches problem areas. This preliminary action reduces the need for reactive course corrections, maintaining path following accuracy while minimizing time loss during navigation
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system prevents damage to crops and robots by ensuring accurate navigation, reduces time for field traversal, and enables efficient recovery from navigation failures, ensuring consistent task performance even in areas with poor GPS connectivity.
Implementation Method 1
analyzing Light Detection And Ranging (LiDAR) perception data when the robot may be in the under-canopy mode
Data Source
AI summary
A system and a method for navigating a robot. The system receives a target location, a field map, sensor data from a robot, and a robot location. The sensor data may include image data and kinetic data. Further, the system may determine a navigation mode based on the sensor data and the robot location. The navigation mode may be one of an under-canopy mode, an out-row mode, and a recovery mode. Further, the system may generate a reference map comprising a path for the robot to follow to reach the target location based on the navigation mode, the field map, and the target location. Finally, the system may be configured to navigate the robot based on the reference map.