Online adaptive observer for rollover avoidance of reconfigurable agricultural vehicles

• Rollover risk is assessed from a standard and low-cost vehicle sensing equipment.
• The immediate evolution of the risk is inferred from an online adapted dynamic model.
• Grip conditions and soil inclination are updated by means of nonlinear observers.
• Reconfigurations in vehicle mass/geometry are tracked from optimization techniques.
• Full-scale experiments with a grape harvester demonstrate the algorithm capabilities.

Tractor rollover is one of the main case of severe accident in agriculture. Since such vehicles move on natural ground with varying conditions and different kind of terrain, the risk of rollover is difficult to estimate and predict using classical on-road approaches. This paper proposes an online adaptive observer to assess and avoid rollover risk in agricultural vehicles which move in off-road terrain. In particular, the approach focuses on reconfigurable tractor dedicated to move in terrain with important slope (such as grape harvester). It is based on the coupling between an intermittent measurement and an estimation of a stability metric, namely the Lateral Load Transfer (LLT). Thanks to this adaptive method, terrain and vehicle parameters are updated in order to take into account for the effects of changes in center of gravity height and total vehicle mass. This then allows to monitor the stability of the vehicle whatever the state of the slope correction system, the soil type and the load of the machine. The algorithm capabilities are tested through experiments using a grape harvester equipped with hydraulic actuators.