Fixed

Typically, an evaluation plan is drawn up, including identification of the various influencing factors, evaluation criteria definition, and test scenarios. Using task-appropriate metrics, the overall product performance is assessed, and underperforming scenarios can be identified. It is also necessary to have various data from the product, which must be qualified and annotated in order to evaluate the product as accurately as possible. This facility is based on MATICS software and high-performance computing resources.

This modular platform covers several fields of robotics. In particular, it features a climate chamber to validate the robot's operation at different temperatures and tracking cameras to measure its movements. The modules designed for movement tests (rough terrain, mazes, etc.) and manipulation tests (handles, valves, 6D positioning) are based on NIST (National Institute of Standards and Technology) standards.

Evaluation on this bench consists of placing a device at the heart of a simulation projected onto a 300° screen while a motion capture system and an instrumented conveyor belt measure its movements, if any. These data are integrated into the simulator in real time so that the device's digital twin follows the same movements and the projected environment is updated accordingly. The simulator also incorporates a physics engine and advanced sensor models, enabling a virtual sensor output to be substituted for the robot's sensors in real time.

The bench offers a realistic representation of a physical environment. This includes high-quality graphics with adjustable precision details for textures, lighting, and shadows, as well as advanced physical models to simulate realistic interactions between objects (collision, thermography). The various environments cover a wide range of scenarios to test AI systems in varied and unpredictable situations. The platform also enables new scenarios to be created to meet specific customer requirements.

This is a setup for testing the transverse spray uniformity of liquid using nozzles mounted on a spray boom. It serves to verify the correct operation of the nozzles and examines the degree of proper liquid spraying on the surface. The degree of uniformity of spraying and quality of the nozzles are determined by the coefficient of spray irregularity.​ Parameters: Table width: 3000 mm; collection groove width: 50 mm​.

The stand is used for verification of the correct operation of algorithms controlling the spectrum of sprayed working liquid in agricultural and orchard sprayers as a function of pressure and temperature. It is used for measuring the size of particles sprayed in the air​. - recording rapid process changes in real time​.

The laboratory specialises in multi-species analysis, including soft wheat, durum wheat, barley, corn, forage, triticale, sorghum, and feed ingredients like rapeseed and soy. With approximately 45 analysis methods, the lab handles 10,000 samples per year and performs 65,000 analyses across 1,000 m² of facilities, including 150 m² of cold storage. For over 10 years, it has maintained a comprehensive database of reference samples. Key areas of expertise include moisture, protein, starch, fat content, hardness, and soil composition.

The facility is designed for efficient and controlled grain storage and processing. It features 350 tonnes of storage capacity with ventilated bins and a silo thermometry system to ensure optimal temperature monitoring. The Sec-LIS® system controls 8 fans automatically to maintain ideal storage conditions, while 4 automatic samplers facilitate continuous quality checks. Two independent circuits with adjustable flow rates (up to 50 t/h) handle grain transport, and a dust extraction system reduces emissions.

There are experimental buildings with an area of 5,430 square meters that can accommodate about two hundred cows. There is a milking parlour with a six-point tandem type. The cows stand one behind the other, are milked, and move forward. The operator's access to the udder is lateral. Milking is much slower, as it allows fewer animals per shift, but the inspection of each cow is more thorough.

This infrastructure is supported by multi-layered connectivity, ranging from low-power networks to 5G, and an advanced data platform. At the core of its capabilities is a supercomputer with 192 cores and 12TB of memory, enabling massive data processing. The infrastructure also features a parallel computing cluster with 960 cores and an NVIDIA DGX2 server, offering 2 petaFLOPS of performance and 512 GB of GPU memory, ideal for handling complex AI models and large datasets.