Human operators of construction and farming equipment exhibit strong situational awareness which enables them to execute tasks while robustly handling unexpected uncertainties in the environment. Automatic control, on the other hand, is generally capable of executing repetitive and well-defined tasks more efficiently and precisely. Collaboratively achieving the tasks by combining the benefits of critical situational awareness and decision making capabilities of human operators and the efficiency and accuracy of automatic control is expected to provide improved performance of these tasks. Development of methods in learning, prediction and human-machine shared control to improve such collaborative task execution and application to hydraulic excavators is the focus of this work. In this paper, we propose a task learning method based on an operator primitives based segmentation (OPbS) and Bayesian non-parametric clustering with temporal ordering (BNPC/TO). Then, we introduce a method for predicting the operator's intent in a dynamical environment by proposing an empirical stochastic transition matrix (ESTM) and a dynamic angle difference exponential (DADE). We then provide a design for blended shared control with conflict-awareness (BSC/CA) extended from the dynamic angle difference. Finally, we evaluate the approach on a scaled hydraulic excavator test-platform for a typical earth-moving task with novice learning operators and a skilled demonstration operator.