Predictive Machine Control With Policy Maps for Future Field Conditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current control systems for mobile machines, such as agricultural vehicles, operate reactively and are unable to dynamically adjust settings based on future sensor observations, limiting their productivity, efficiency, and wear characteristics.
Innovation Solution
A predictive control system that uses a policy map generated by a plant model to optimize machine settings and trajectory, incorporating forward-looking sensors and a predictive vehicle model to adjust control signals for improved performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If reactive control systems are used to control mobile machines, then the control system can respond to current sensor measurements, but the system cannot anticipate future conditions and adjust settings proactively
Solution Approach 1:
The system performs preliminary actions by using forward-looking sensors to detect future conditions before the machine encounters them. The predictive controller pre-adjusts machine settings based on anticipated conditions, allowing the system to react proactively rather than merely responding to current states. This resolves the contradiction by enabling future-oriented adaptation without requiring complete redesign of the control architecture.
Solution Approach 2:
A predictive vehicle model acts as an intermediary between forward-looking sensors and the controller. This model simulates future vehicle responses to different control actions, enabling the system to evaluate potential outcomes without direct physical experimentation. The intermediary allows complex predictive capabilities while maintaining manageable system architecture.
2Productivity
If forward-looking sensors and predictive modeling are implemented, then productivity and efficiency are enhanced, but the device complexity increases
Solution Approach 1:
The system performs preliminary computations and adjustments by predicting future vehicle responses before actual operations occur. The predictive controller calculates optimal control signals in advance based on anticipated conditions, reducing real-time computational burden and enabling faster execution. This approach enhances productivity by pre-optimizing machine performance while managing complexity through time-shifted processing.
Solution Approach 2:
The predictive vehicle model creates a virtual copy of the actual vehicle system to simulate future behaviors and test control strategies. This digital twin allows the system to evaluate multiple scenarios without affecting the real machine, enabling sophisticated predictive capabilities while isolating complexity to the simulation environment. The copying principle resolves the contradiction by transferring computational complexity from the physical control system to the virtual model.
Data Source
AI summary
A policy map is obtained that includes a first set of machine settings values for controlling a machine to perform an operation at a site, at different locations in the field. An adjustment component receives sensor signals indicating conditions that will be encountered by the machine in the future, as it moves through the site. A cost determination component determines whether an adjustment is to be made to the first set of settings values, based upon the sensor signals. A predictive vehicle model provides an expected vehicle response, based upon the adjusted settings values that are selected. Control signals are generated based upon adjusted setting values that are generated from the policy map, the expected vehicle response, and the future condition sensor signals. The control signal generator generates the control signals to control a set of controllable subsystems.