Data-driven Mass Estimation in Continuously Variable Transmission Agricultural Tractors

Savaresi, Dario; Formentin, Simone; Savaresi, Sergio M.

doi:10.23919/acc50511.2021.9483368

Cited by 1 publication

(2 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…whereas γ f ri • v is a lumped friction term, an acceptable modeling assumption for our purposes [22]; g is the gravity acceleration; φ is the road inclination angle; M veh is the tractor payload; and v is the time derivative of v, which can be computed through a derivative filter. The time dependence (t), explicit in (1), will be dropped from now on for the sake of simplicity.…”

Section: Eco Indexmentioning

confidence: 99%

“…Furthermore, although being usually extremely well performing, model-free approaches heavily relying on machine learning algorithms are well known to lack in robustness, especially when tested over events not present in the training set: this is a critical issue when coming to agricultural tractors from an implementation-oriented perspective. In fact, vehicle payload might change significantly [22], as different tools or trailers might be connected to the vehicle front or rear part, thus potentially yielding unpredictable results.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Driving Style Assessment System for Agricultural Tractors: Design and Experimental Validation

2022

Self Cite

View full text Add to dashboard Cite

The diffusion of electronics and sensors in agricultural vehicles is enabling a revolution in the field, leading—among the rest—to the introduction of advanced driving-assistance systems (ADAS). From this perspective, the three key performance indicators (KPI) in a tractor are indeed the driving safety, fuel consumption, and operator comfort. Such indexes describe the way the driver interacts with the vehicle, the environment, and other vehicles, respectively. Therefore, such information would be particularly valuable if promptly provided to the driver, e.g., on a dashboard visualizer, so as to adapt the driving style accordingly. Within this context, we propose an algorithmic solution for the on-line estimation of such KPIs. More specifically, by using an off-the-shelf smart-sensor equipped with an Electronic Control Unit (ECU), the chassis accelerations are first processed to extract physics-inspired features and then used to assess the safety and comfort levels; similarly, the speed profile is used to evaluate the economicity of the driving style. The developed method is based upon a cheap setup, and thus it is industrially amenable for its simplicity and robustness. A sensitivity analysis to establish the best sensor placement is finally carried out, together with an extensive experimental campaign considering offroad, urban, and circuit paths.

show abstract

Section: Eco Indexmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%