Model-based chassis control system for an over-actuated planetary exploration rover

Abstract

In planetary exploration, wheeled mobile robots (rovers) are popular for extending action range compared to a lander. Despite their success, they continue to struggle with soft grounds which shows in high sinkage and can lead to an immobilization in the worst case. Rovers usually are over-actuated due to individual wheel drives and steering, which is rarely made use of in current missions. Some work optimizing the resulting degrees of freedom exists but often does not use all available model knowledge. In this work, the rover is consequently modeled with the subsystems rigid body dynamics, kinematics and wheel/ground dynamics. Feedback linearization is used for the rigid body and the underlying wheel/ground controllers on individual wheel level. The control allocation of the forces is done via the pseudo-inverse and a base of the null-space to extract the available degrees of freedom. A verification of the approach is shown in a co-simulation with a high-fidelity model of the ExoMars rover.

Fast Traction Control of Rovers on Prescribed Dynamic Trajectories with Wheel-Fighting Consideration

“…Remark 3. Based on Proposition 1, f cm is independent of the control distribution τ a ; hence, the sign specifier Ξ f does not change the quadraticity of (34) and accordingly the linearity of (31) with respect to τ a . Remark 4.…”

“…According to the determined functionality of tractive and normal forces under the explained assumptions, a cost function r can be formulated with the aim of improving traction. A seemingly proper choice is the quadratic norm of the vector of tractive ratios weighted by the inverse of static friction coefficients [27]- [31]. Assuming convexity of this cost function, it results in a nonlinear necessary condition of optimality in τ a , similar to (19).…”

“…Note that although the function Ξ f is discontinuous, the cost function r f is continuous and differentiable with respect to τ a . Therefore, (19) leads to the linear necessary and sufficient optimality condition in (31), knowing that the positive semidefinite matrix Q T f Q f is the Hessian. Remark 3.…”

“…e., all of its singular values are greater than a tolerance, the optimal solution τ * a is calculated by (31) has infinite number of solutions [38]. To obtain a solution we consider singular value decomposition of Q T f Q f , i.e., this matrix is formatted as…”

“…Another dynamic traction control strategy for redundant autonomous rovers is proposed in [30], where a rigid body dynamical model of the system and a terramechanics model are combined to introduce an optimal control distribution to maximize traction. This approach is enhanced by redefining the optimization criterion to consider the quadratic norm of tractive ratios [31].…”