Adaptively reallocating resources of resource-constrained devices
Implementations are disclosed for adaptively reallocating computing resources of resource-constrained devices between tasks performed in situ by those resource-constrained devices. In various implementations, while the resource-constrained device is transported through an agricultural area, computing resource usage of the resource-constrained device ma may be monitored. Additionally, phenotypic output generated by one or more phenotypic tasks performed onboard the resource-constrained device may be monitored. Based on the monitored computing resource usage and the monitored phenotypic output, a state may be generated and processed based on a policy model to generate a probability distribution over a plurality of candidate reallocation actions. Based on the probability distribution, candidate reallocation action(s) may be selected and performed to reallocate at least some computing resources between a first phenotypic task of the one or more phenotypic tasks and a different task while the resource-constrained device is transported through the agricultural area.
1. A method implemented by a resource-constrained device while the resource-constrained device is transported through an agricultural area, the method comprising:
monitoring computing resource usage of the resource-constrained device;
monitoring phenotypic output generated by one or more phenotypic tasks performed onboard the resource-constrained device;
generating a state based on the monitored computing resource usage and the monitored phenotypic output;
processing the state based on a policy model to generate a probability distribution over a plurality of candidate reallocation actions; and
based on the probability distribution, selecting and performing one or more of the candidate reallocation actions to reallocate at least some computing resources between a first phenotypic task of the one or more phenotypic tasks and a different task while the resource-constrained device is transported through the agricultural area.
2. The method of claim 1 , wherein the one or more phenotypic tasks comprise a plurality of phenotypic tasks, and the different task comprises a second phenotypic task of the plurality of phenotypic tasks.
3. The method of claim 1 , wherein the different task comprises a new phenotypic task outside of the one or more phenotypic tasks.
4. The method of claim 1 , wherein the state is further generated based on an amount of battery power required for the resource-constrained device to achieve one or more goals.
5. The method of claim 4 , wherein the one or more goals include performance of the one or more phenotypic tasks over at least a threshold amount of the agricultural area.
6. The method of claim 1 , wherein the state is further generated based on one or more environmental conditions of the agricultural area.
7. The method of claim 1 , wherein the monitored phenotypic output signals potential presence of a phenotypic condition, and the different task includes application of a machine learning model trained to detect presence of the phenotypic condition.
8. The method of claim 7 , wherein the phenotypic condition comprises plant disease infection or pest infestation.
9. The method of claim 7 , wherein the phenotypic condition comprises plant dehydration or over-hydration.
10. The method of claim 1 , further comprising causing an output device to render output conveying an effect the selected one or more candidate reallocation actions will have on the first phenotypic task.
11. The method of claim 10 , wherein the output further comprises a prompt seeking approval to perform the selected one or more candidate reallocation actions.
12. The method of claim 1 , wherein the selected one or more candidate reallocation actions include decreasing processor cycles assigned to the first phenotypic task and increasing processor cycles assigned to the different task.
13. The method of claim 1 , wherein the selected one or more candidate reallocation actions include decreasing memory allocated to the first phenotypic task and increasing memory allocated to the different task.
14. The method of claim 1 , wherein the selected one or more candidate reallocation actions include decreasing network bandwidth allocated to the first phenotypic task and increasing network bandwidth allocated to the different task.
15. The method of claim 1 , wherein the selected one or more candidate reallocation actions include decreasing inter-core bandwidth allocated to the first phenotypic task and increasing inter-core bandwidth allocated to the different task.
16. A resource-constrained device configured for transportation through an agricultural area, the resource-constrained device comprising one or more processors and memory storing instructions that, in response to execution by the one or more processors, cause the one or more processors to:
monitor computing resource usage of the resource-constrained device;
monitor phenotypic output generated by one or more phenotypic tasks performed onboard the resource-constrained device;
generate a state based on the monitored computing resource usage and the monitored phenotypic output;
process the state based on a policy model to generate a probability distribution over a plurality of candidate reallocation actions; and
based on the probability distribution, select and perform one or more of the candidate reallocation actions to reallocate at least some computing resources between a first phenotypic task of the one or more phenotypic tasks and a different task while the resource-constrained device is transported through the agricultural area.
17. The resource-constrained device of claim 16 , wherein the one or more phenotypic tasks comprise a plurality of phenotypic tasks, and the different task comprises a second phenotypic task of the plurality of phenotypic tasks.
18. The resource-constrained device of claim 16 , wherein the different task comprises a new phenotypic task outside of the one or more phenotypic tasks.
19. The resource-constrained device of claim 16 , wherein the state is further generated based on an amount of battery power required for the resource-constrained device to achieve one or more goals.
20. A non-transitory computer-readable medium comprising instructions that, in response to execution of the instructions by one or more processors of a resource-constrained device while the resource-constrained device is transported through an agricultural area, cause the one or more processors to:
monitor computing resource usage of the resource-constrained device;
monitor phenotypic output generated by one or more phenotypic tasks performed onboard the resource-constrained device;
generate a state based on the monitored computing resource usage and the monitored phenotypic output;
process the state based on a policy model to generate a probability distribution over a plurality of candidate reallocation actions; and
based on the probability distribution, select and perform one or more of the candidate reallocation actions to reallocate at least some computing resources between a first phenotypic task of the one or more phenotypic tasks and a different task while the resource-constrained device is transported through the agricultural area.