Machine learning based control and re-routing
Embodiments include technologies that use machine learning to update a route plan to re-route a mobile machine in a field. In some examples, a method includes using a mobile machine to perform work in a field and recording travelled path information. The recorded path information includes information about one or more paths followed by the machine when performing the work in the field. The method also includes using one or more computing devices on the machine to input the path information into a trained deep learning model and to receive new machine travel paths from the trained model. Also, the method includes using the computing device(s) on the mobile machine to control the machine to follow the new travel paths generated by the trained model.
1 . A method, comprising:
using a mobile machine to perform work in a field and recording travelled path information, the travelled path information comprising information about one or more paths followed by the mobile machine when performing the work in the field;
using one or more computing devices on the mobile machine to input the travelled path information into a trained deep learning model and to receive new machine travel paths from the deep learning model;
using the one or more computing devices on the mobile machine to control the machine to follow the new machine travel paths generated by the trained deep learning model;
using one or more computing devices on the mobile machine to input the travelled path information and real-time field attributes into the trained model and to receive new machine travel paths from the trained model, wherein the real-time field attributes are sensed by the mobile machine, or by nearby sensors in the field, or by remote sensors capturing field attributes near the machine; and
using, by the computing system, the trained model to generate the new machine travel paths according to the travelled path information and the real-time field attributes.
2 . The method of claim 1 , further comprising:
receiving, by a computing system, initial routing information, the initial routing information defining routes followed or to be followed by one or more mobile machines in one or more fields; and
prior to using the deep learning model to generate new routing information, training, by the computing system, the deep learning model using the initial routing information.
3 . The method of claim 2 , wherein the initial routing information comprises initial waylines and the new machine travel paths comprise new waylines, wherein the new waylines do not include any of the initial waylines since the initial way lines have already been followed by the machine and none of the initial waylines need to be repeated to complete the route.
4 . The method of claim 1 , further comprising using, by the computing system, the trained model to generate new routing information specific to the field and the mobile machine, and wherein the use of the mobile machine to perform the work is based on the new routing information and the new routing information comprises the new machine travel paths.
5 . The method of claim 1 , further comprising storing the trained model on the mobile machine.
6 . The method of claim 1 , further comprising: receiving, by the computing system, initial mobile machine information pertaining to the mobile machine; and further training, by the computing system, the deep learning model according to the initial mobile machine information.
7 . The method of claim 6 , wherein the initial mobile machine information comprises one or more of machine model information, machine type information, machine size information, machine shape information, machine ground footprint information, machine turn radius information, and energy usage information.
8 . The method of claim 1 , further comprising: receiving, by the computing system, initial field information pertaining to the field; and further training, by the computing system, the deep learning model according to the initial field information.
9 . The method of claim 8 , wherein the initial field information comprises one or more of field size information, field shape information, field elevation information, field topology information, soil type information, soil condition information, crop type information, crop lodging information, soil compaction information, weed density information, and weed location information.
10 . The method of claim 1 , further comprising: receiving, by the computing system, weather data; and further training, by the computing system, the deep learning model according to the weather data.
11 . The method of claim 1 , wherein the trained model is configured to generate the new machine travel paths to minimize fuel consumption of the mobile machine when performing a given field operation.
12 . The method of claim 1 , wherein the trained model is configured to generate the new machine travel paths to minimize operation time of the mobile machine when performing a given field operation.
13 . The method of claim 1 , wherein the trained model is configured to generate the new machine travel paths to minimize soil compaction caused by the mobile machine when performing a given field operation.
14 . A system, comprising: a processing device; and memory in communication with the processing device and storing instructions that, when executed by the processing device, cause the processing device to:
use a mobile machine to perform work in a field and recording travelled path information, the travelled path information comprising information about one or more paths followed by the mobile machine when performing the work in the field;
use one or more computing devices on the mobile machine to input the travelled path information into a trained deep learning model and to receive new machine travel paths from the deep learning model;
use the one or more computing devices on the mobile machine to control the machine to follow the new machine travel paths generated by the trained deep learning model;
use one or more computing devices on the mobile machine to input the travelled path information and real-time machine attributes into the trained model and to receive new machine travel paths from the trained model, wherein the real-time machine attributes are measured or sensed by the mobile machine, or by nearby sensors in the field, or by remote sensors capturing attributes of the machine; and
use the trained model to generate the new machine travel paths according to the travelled path information and the real-time machine attributes.
15 . The system of claim 14 , wherein when the instructions are executed by the processing device, cause the processing device to:
receive initial routing information, the initial routing information defining routes followed or to be followed by one or more mobile machines in one or more fields;
prior to using the deep learning model to generate new routing information, train the deep learning model using the initial routing information; and
store the trained model on the mobile machine.
16 . The system of claim 14 , wherein when the instructions are executed by the processing device, cause the processing device to:
use one or more computing devices on the mobile machine to input the travelled path information and real-time field attributes into the trained model and to receive new machine travel paths from the trained model, wherein the real-time field attributes are sensed by the mobile machine, or by nearby sensors in the field, or by remote sensors capturing field attributes near the machine; and
use the trained model to generate the new machine travel paths according to the travelled path information and the real-time field attributes.
17 . The system of claim 14 , wherein when the instructions are executed by the processing device, cause the processing device to:
use one or more computing devices on the mobile machine to input the travelled path information and real-time machine attributes into the trained model and to receive new machine travel paths from the trained model, wherein the real-time machine attributes are measure or sensed by the mobile machine, or by nearby sensors in the field, or by remote sensors capturing attributes of the machine; and
use the trained model to generate the new machine travel paths according to the travelled path information and the real-time machine attributes.
18 . A method, comprising:
using a mobile machine to perform work in a field and recording travelled path information, the travelled path information comprising information about one or more paths followed by the mobile machine when performing the work in the field;
using one or more computing devices on the mobile machine to input the travelled path information into a trained deep learning model and to receive new machine travel paths from the deep learning model;
using the one or more computing devices on the mobile machine to control the machine to follow the new machine travel paths generated by the trained deep learning model;
using one or more computing devices on the mobile machine to input the travelled path information and real-time machine attributes into the trained model and to receive new machine travel paths from the trained model, wherein the real-time machine attributes are measured or sensed by the mobile machine, or by nearby sensors in the field, or by remote sensors capturing attributes of the machine; and
using, by the computing system, the trained model to generate the new machine travel paths according to the travelled path information and the real-time machine attributes.