This paper presents a method for autonomous terrain traversal based on prior identified vehicle parameters. Parameters are identified real-time using an extended Kalman filter. This is important on an autonomous vehicle because tire-road friction and other vehicle properties can change with varying soil properties. Carsim, a high-fidelity vehicle simulation software, is used to test the parameter identification algorithms. Some important parameters that will be identified include: lateral tire stiffness, tire-road friction, and weight split. These parameters are used with a vehicle model to look ahead to predict how fast the autonomous vehicle should approach a predefined route, assuming prior knowledge of the route. The downside of this technique is the tires must become saturated or excited to estimate the max lateral force the tire can allow before sliding out. The effectiveness of this method is also discussed and validation plots are shown.
展开▼
