An Optimization-Oriented Framework to Autonomous Excavation


Student thesis: Doctoral Thesis

View graph of relations


Related Research Unit(s)


Awarding Institution
  • Lidai WANG (Supervisor)
  • Jia Pan (External person) (External Co-Supervisor)
Award date2 Sept 2021


Hydraulic excavators are widely used in construction, mining and many other scenarios. Nowadays, the demands of autonomous excavation emerge due to labor shortage of skillful operators and hazardous working conditions. In industry, it is important to optimize an excavation trajectory for two main purposes: maximizing the time efficiency and minimizing the energy consumption. This thesis aims to establish an optimization-based framework of excavation trajectory generation that allows large space for optimization, various objective functions and flexibility of online replanning during trajectory execution.

For a general excavation task, only the purpose is defined clearly, that is to dig the soil out and try to fill fully the bucket. However, there is no clear-cut mathematical specification of how an excavation should be conducted. Thus, an excavation trajectory generator needs to find a proper trajectory formulation at first and then search trajectories based on this formulation. For the excavation task, it is not trivial because, on one hand, over-simplifying the description would restrict the room for optimization, but on the other hand, complex description might induce heavy computation load which is not suitable for online replanning. The capability of online replanning during excavation is of vital importance, because it is hard to know the underground soil condition in advance of digging. The excavator might get stuck in situations where the resistance force from the soil is too large or the bucket encounters an unpredictable stone.

This thesis solves the aforementioned problems by formulating the excavation task within an optimization-based framework. Trajectories are constrained by a set of nonlinear equations, which guarantees the feasibility of excavation and allows large room of optimization and the flexibility for replanning. The framework supports various objectives, such as the minimum travel time, the minimum joint length and the minimum consumed torque, etc. To handle with local minima issues, we generate multiple initial trajectories by quickly searching simpler parameterized trajectories on a kinematics reachability map.

We conduct experiments on a real robot platform. The results demonstrate that our method is adaptive to different terrain shapes and outperforms other optimal excavation path planners in terms of minimum joint length and minimum travel time. We also conduct experiments in simulation to demonstrate our method could generate trajectories satisfying the torque for different soil conditions. We also demonstrate the capability of online replanning when facing large resistance forces.