Robot Path Navigation With Dynamic Sensor Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing navigation systems for vehicles, such as robotic lawn mowers, require expensive training and intermediate information to navigate precisely along predetermined paths, leading to inefficient path coverage and significant overlapping.
Innovation Solution
Dynamic weighting of odometry and distance measurement data based on the geometry of the path within a state observer's kinematic model, allowing for improved navigation precision by adjusting data reliability according to path characteristics, thereby minimizing errors and optimizing path length and coverage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If training and intermediate information are used for navigation, then navigation precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent extracts and eliminates the complex training phase and intermediate information generation steps from the navigation system. By using a simplified probabilistic model that directly processes sensor data without requiring pre-trained maps or intermediate representations, the system achieves comparable navigation precision with significantly reduced complexity
Solution Approach 2:
The patent replaces expensive, complex training procedures and intermediate information structures with a lightweight probabilistic model that can be quickly adapted. The system uses simple sensor fusion with dynamically adjusted weights rather than costly pre-processing and mapping infrastructure
2Productivity
If chaotic back and forth movement is used to cover the area, then the entire area is eventually covered, but path length and time increase
Solution Approach 1:
The patent implements dynamic path planning that adapts the robot's movement strategy in real-time based on its current position, orientation, and the geometry of the work area. The probabilistic model continuously updates the optimal path by adjusting the weighting of sensor data according to the local environment, enabling the robot to follow efficient trajectories rather than chaotic patterns
Solution Approach 2:
The system uses continuous feedback from sensors (odometry, distance measurements) to update the probabilistic model of the robot's position and the work area geometry. This feedback loop enables real-time optimization of the navigation path, allowing the robot to correct deviations and adjust its course to minimize total path length while ensuring complete area coverage
3Measurement precision
If fixed weighting of sensor data is used in navigation, then system simplicity is maintained, but navigation precision deteriorates
Solution Approach 1:
The patent dynamically changes the parameters (weighting factors) of the sensor data fusion based on the geometric properties of the work area and the robot's current state. The system adjusts the relative importance of different sensors (odometry, distance measurements) according to the local environment, improving position estimation accuracy without requiring complex adaptive algorithms
Solution Approach 2:
The patent applies different weighting strategies to different sensors based on their local reliability in specific geometric contexts. Rather than using a uniform weighting scheme, the system adapts the quality assessment of each sensor input according to the local work area geometry and the robot's position, achieving higher precision with simple local adjustments
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for moving a vehicle (1) along a predefined path (3), which comprises straight sections (4, 5) and curved sections (6), in which method a current location on the predefined path (3) and a driving direction (θ) of the vehicle (1) present at said location are determined cyclically by the fusion of odometer data from the robot (1) and data concerning the robot location from UWB distance measurement with at least two anchor points (7 to 10) of known position by means of a Kalman filter, and then a course correction is determined for the vehicle (1) from the location obtained together with the driving direction obtained in a trajectory controller, and predefined basic values for the variables for a process noise Q of the odometer data and a measurement uncertainty R of the data from the distance measurement are used in the Kalman filter, characterized in that at least one of the values for the variables for the process noise Q and the measurement uncertainty R is modified relative to the predefined basic values according to whether the current location determined is located on a straight section or on a curved section of the predefined path.