Tanker trucks are commonly used for transporting liquid material including chemical and petroleum products. On the one hand, tanker trucks are susceptible to rollover accidents due to the high center of gravity when they are loaded and due to the liquid sloshing effects when the tank is partially filled. On the other hand, tanker truck rollover accidents are among the most dangerous vehicle crashes, frequently resulting in serious to fatal driver injuries and significant property damage, because the liquid cargo is often hazardous and flammable. Therefore, effective schemes for tanker truck rollover avoidance are highly desirable and can bring a considerable amount of societal benefit. Yet, the development of such schemes is challenging, as tanker trucks can operate in various environments and be affected by manufacturing variability, aging, degradation, etc. This paper considers the use of Learning Reference Governor (LRG) for tanker truck rollover avoidance. The LRG is an add-on scheme to a nominal closed-loop system (i.e., the tanker truck in this application). The LRG monitors the maneuver command from the driver (or from a higher-level planning algorithm in the case of an automated tanker truck) and adjusts it when necessary to enforce rollover constraints. One major novelty of this LRG scheme is that its formulation relies on learning, rather than explicit modeling of the system, where learning can be performed on a high-fidelity, black-box model of the truck system or through experimentation with the actual truck. We illustrate the effectiveness of this LRG-based rollover avoidance scheme through simulation studies corresponding to various truck operating conditions.
