Design of computational environments for digital testing
Summary
Design of computational environments for digital testing
S00176
Version: 1.11
- arable farming
- horticulture
- viticulture
- test design
- design/documentation; other (including: nothing)
- Location: remote
- Offered by: POLIMI; UNIMI
Description
Any test activity involves three main components, i.e.: environment (where the tests take place), protocol (defining what tests are executed and how) and evaluation metrics (used to assess the results of the tests).
This service covers the design of a digital testing environment. Indeed, many agri-food technologies need to be tested and validated in digital environments: e.g., i) in simulation, before the system is physically deployed, or ii) on specific datasets, to test the performance of algorithms in controlled scenarios.
With this service, we assist customers in the design of a computational environment that is suitable for running digital tests. Specifically, we will:
define the requirements on data (and metadata) requirements to support the test,
identify the software needed to support the test,
select simulator and define simulated environment (if needed)
select remotisation tools (if needed)
identifying hardware requirements (e.g., memory specifics, GPU and TPU acceleration) for the digital infrastructure supporting the tests
Based on the identified requirements, we will provide the design of a test setup that is compliant with the AgrifoodTEF digital infrastructure. The chosen solution will take full advantage of the resources and expertise made available through the consortium but will also be applicable by the customer on their own. Finally, the service will provide customers with alternatives, should more customised options be needed.
Example service: To avoid damaging plants during weeding operations, the customer wants to test an intra-row navigation solution in simulation before deploying the system in field. To enable such tests, we reconstruct a subset of the customer’s deployment environment within the Gazebo simulator, including three cultivated plant rows. To take into account the kinematics of the customer’s weeding robot, we rely on a virtual replica of a skid-steering mobile robot tailored to closely mimic the customer’s solution. The robot to be tested is equipped with a LiDAR sensor and a camera system, so corresponding virtual sensors are added to the Gazebo model of the machine and proper interaction between them and the simulated environment is verified. Leveraging the possibilities offered by simulation, we design different simulation environments by changing the row length (3, 5, up to 10 metres) and by making the corridor narrower (intra-row width from 3 down to 1 metre), to enable tests of the navigation system in increasingly difficult scenarios.