A method of estimating the state of contact between (1) the wheels of a robotic vehicle equipped with a rocker-bogey suspension and (2) the ground has been devised. The contact-state information is used to estimate the position of the vehicle with an accuracy greater than that achievable by simple odometry and inertial sensing when the vehicle moves over undulating or bumpy terrain under conditions in which processing of visual odometry information is infeasible or undesirable. The contact-state information can also be used to increase understanding of those aspects of design and operation of the vehicle that affect stability and traction, as affected by the configuration of the rocker-bogey and steering mechanisms. The method in its original form is meant to be applied to Rocky-7 (see figure) - a "rover"-type vehicle used in research on robotic-vehicle concepts for the exploration of Mars. The method is also potentially applicable to terrestrial robotic vehicles that could be used in field operations in agriculture, mining, and other industries.

Rocky-7 includes a rocker-bogey suspension with two steerable front wheels. The present method provides for estimation of the state of wheel/ground contact on the basis of sensor readings and the highly nonlinear kinematics of the rocker/bogey suspension.

In this method, the forward kinematics of the vehicle (proceeding from the vehicle frame to the wheel/ground contact points) are embedded within a constraint that is treated as a measurement. The forward kinematic chain velocity for each wheel includes a component defined by the sequence of links that join the vehicle frame to the wheel/ground contact point, plus a component given by the slip between the wheel and the ground. An important element of the method is the notion of a slip measurement or constraint that defines the six-degree-of-freedom (6-dof) motion of the contact frame on the wheel, relative to the ground. The slip is a function of the vehicle configuration, the 6-dof vehicle velocity, the location of the wheel/ ground contact points, and the rates of rotation of the joints along the various kinematic chains.

The slip can be decomposed into a deterministic component and a component that is known in a statistical sense only. The deterministic component of the slip is used to capture the effects of a known steering action; for example, a known rotational slip about the vertical is always present at each wheel to accommodate the yaw motion of the vehicle during a turn. Also, some transverse slip is introduced because of the nature of the nonsteered bogey wheels. These deterministic slips are easily calculable for steered motions on flat ground, and those calculated for flat ground are used as approximations for those on nonflat ground. Other slip constraints can be derived from experiments. Slips that are modeled statistically include those attributable to wheel-ground interactions and to curvature of the terrain.

The state-estimation algorithm of this method utilizes an extended Kalman filter to fuse data from multiple sensors aboard the vehicle (e.g., a gyroscope, a Sun sensor, and accelerometers) while taking account of the kinematics of the rocker-bogey suspension and steering. In addition to the kinematic elements described above, the extended Kalman filter incorporates process models of attitude, translation, gyroscope bias, plus observation models for the gyroscope, Sun sensor, and accelerometer. The highly nonlinear kinematics of the rocker-bogey suspension and the wheel/ ground contact points are incorporated into the filter via the slip constraint. The algorithm exploits the ability of the Kalman filter to perform the appropriate least-squares averaging of the action of each kinematic chain in the vehicle.

This work was done by J. Balaram of Caltech for NASA's Jet Propulsion Laboratory.


This Brief includes a Technical Support Package (TSP).
Estimation of Wheel-Contact State of a Robotic Vehicle

(reference NPO20960) is currently available for download from the TSP library.

Don't have an account? Sign up here.