Towards Controlling Bucket Fill Factor in Robotic Excavation by Learning Admittance Control Setpoints

Abstract

This paper investigates the extension of an admittance control scheme toward learning and adaptation of its setpoints to achieve controllable bucket fill factor for robotic excavation of fragmented rock. A previously developed Dig Admittance Controller (DAC) is deployed on a 14-tonne capacity robotic load-haul-dump (LHD) machine, and full-scale excavation experiments are conducted with a rock pile at an underground mine to determine how varying DAC setpoints affect bucket fill factor. Results show that increasing the throttle setpoint increases the bucket fill factor and increasing the bucket's reference velocity setpoint decreases the bucket fill factor. Further, the bucket fill factor is consistent for different setpoint values. Based on these findings, a learning framework is postulated to learn DAC setpoint values for a desired bucket fill factor over successive excavation iterations. Practical implementation problems such as bucket stall and wheel-slip are also addressed, and improvements to the DAC design are suggested to mitigate these problems.

Iterative Learning-Based Admittance Control for Autonomous Excavation

“…One promising approach to autonomous excavation is an admittance-based interaction control strategy that has been developed through extensive field experiments [16,7,8]. In contrast to pure motion control, which rejects disturbance forces to track a given motion reference trajectory, admittance control attempts to comply robot motion with environment interaction and react quickly to measured interaction forces by rapidly modifying the robot's reference velocity [17].…”

“…In this study, an admittance control strategy for autonomous excavation [16,7,8] is augmented with iterative learning to enhance performance when excavating material with unknown or changing characteristics. This section provides an overview of admittance control and of Iterative Learning Control (ILC), which are used to formulate an iterative learning-based admittance control algorithm for autonomous excavation.…”

“…Effective digging can be achieved by prescribing a constant admittance H(s) = K A > 0, a constant reference velocity v d =v d , and a target force f d obtained by filtering the measured force signalf r [8]. In this implementation, the dump cylinder extends at a constant nominal ratev d , which increases by a rate v f whenf r > f d (e.g., if a subsurface obstacle is encountered), resulting in faster bucket curl; and v d decreases by a rate v f whenf r < f d (e.g., if machine loses contact with the pile) resulting in slower bucket curl.…”

“…The nominal level of interaction forces is difficult to predetermine because it is affected by several factors, such as the characteristics of the media, amount of material in the bucket, orientation of the bucket, and throttle. In our earlier work [8], we showed that the target force can be obtained by filtering the measured forcesf r with a windowed median filter. Given a discrete-time implementation of the control system, at a time step k, the target force filter is defined as…”

“…Recent research [7,8] has shown that an admittance-based control strategy for autonomous excavation, which utilizes force feedback to regulate velocity commands to a loader's bucket actuator, achieves consistent bucket filling performance; however, certain control parameters require tuning when the excavated media changes characteristics (e.g., pile fragmentation, water content, etc.). In practice, material characteristics are unknown and changing throughout an excavation operation.…”