Global Control for Robotic Excavation Using Fuzzy Logic and Statecharts

Abstract

This paper presents global strategies for the computer control of autonomous backhoe-type excavators. The control structure is divided into low and high levels. The low level control utilises fuzzy logic to encapsulate expert experience for capturing soil in many excavation scenarios. UML statecharts are used at the higher level for mapping environment and machine sensor data to actuator control signals. This mapping is based on a deep understanding of excavation performed by a skilful operator and is coded into rule sets. Transition between states or behaviours is accomplished via associated task characteristic functions that not only switch among tasks but also enable/disable rules according to digging phases. Typical excavation tasks are decomposed into statecharts and task elements. The control schemes are illustrated by autonomous trench digging. Field test results are provided to verify the validity of the proposed control architecture.

A Control Architecture for Robotic Excavation in Construction

“…In Santos et al (2000), for the task of digging a trench a task decomposition can be proposed as follows:…”

“…A statechart is used to represent a state-machine consisting of states and transitions between states, together with synchronization states and pseudostates. Every state object has entry and exit actions, and executes the particular behavior or activity until a transition is set and the state is exited (Santos et al, 2000). Transition conditions can be refined to have event priorities if more than one condition is true at the same time, or can be seen as if/else or switch/case structures.…”

“…For example, FLC3 and FLC4 respectively implement the task elements "curl the bucket inward" (τ 3 ) and "curl the bucket outward" (τ 4 ) using the bucket cylinder pressures and the rate of change of bucket joint angle as inputs, and the bucket spool valve opening area as the output. Similarly, FLC5 and FLC6 use the arm cylinder pressures and arm angular speed as inputs and the arm spool valve opening area as the output to implement the element tasks "rotate the arm inward" (τ 5 ) and "rotate the arm outward" (τ 6 ) (Santos et al, 2000). The rule sets are, in general, heuristically formulated from observing skillful operators of earthmoving machines.…”