Three-dimensional dynamical bulldozer modeling and fuzzy control of blade yaw, pitch, and roll

Automating bulldozer control tasks could reduce operator-training demands, but most existing blade-control research addresses pitch alone, limiting automation of tasks that require simultaneous yaw, pitch, and roll control. This study develops a three-dimensional (3D) bulldozer model and a fuzzy-PI controller for these blade motions. The simulation combines surface-to-track dynamics, including track slip, with surface-to-blade dynamics that account for non-symmetric soil mounds. In this framework, a primary fuzzy-PI controller maps blade yaw, pitch, and roll errors to the associated effective hydraulic-displacement rate command feedback gains, while a secondary fuzzy controller selects the desired cutting depth from the observed soil type. A discrete-time Lyapunov analysis establishes nominal asymptotic stability only for the reduced, constant-reference, noise-free, unsaturated blade-error dynamics under ideal decoupled actuation. The proposed controller was compared with P, PI, and fuzzy-PID controllers in four simulated grading tests. It obtained the lowest roll, depth, and yaw Root Mean Squared Errors (RMSEs) in the noisy loose-soil test (2.832 mrad, 4.477 mm, and 2.832 mrad, respectively), whereas the fuzzy-PID controller produced the lowest errors in the noisy compact and mixed soil tests but with a greater drop-off in performance when tested away from the tuned setpoint as compared to the proposed controller. In the noise-free test, however, the proposed fuzzy-PI controller converged more slowly than the comparison controllers. These simulation results indicate that the computationally simple 3D proposed controller can provide consistently performant tracking error in the tested cases with measurement noise, with further tuning and validation on physical equipment being of interest.