Inductive learning of answer set programs for autonomous surgical task planning

Daniele Meli,Paolo Fiorini,Mohan Sridharan

doi:10.1007/s10994-021-06013-7

Daniele Meli, Paolo Fiorini + Show 1 more

Open Access

https://doi.org/10.1007/s10994-021-06013-7

Copy DOI

Journal: Machine Learning	Publication Date: Jun 15, 2021
Citations: 7	License type: open-access

Affiliation: University of Verona, University of Birmingham

Abstract

The quality of robot-assisted surgery can be improved and the use of hospital resources can be optimized by enhancing autonomy and reliability in the robot’s operation. Logic programming is a good choice for task planning in robot-assisted surgery because it supports reliable reasoning with domain knowledge and increases transparency in the decision making. However, prior knowledge of the task and the domain is typically incomplete, and it often needs to be refined from executions of the surgical task(s) under consideration to avoid sub-optimal performance. In this paper, we investigate the applicability of inductive logic programming for learning previously unknown axioms governing domain dynamics. We do so under answer set semantics for a benchmark surgical training task, the ring transfer. We extend our previous work on learning the immediate preconditions of actions and constraints, to also learn axioms encoding arbitrary temporal delays between atoms that are effects of actions under the event calculus formalism. We propose a systematic approach for learning the specifications of a generic robotic task under the answer set semantics, allowing easy knowledge refinement with iterative learning. In the context of 1000 simulated scenarios, we demonstrate the significant improvement in performance obtained with the learned axioms compared with the hand-written ones; specifically, the learned axioms address some critical issues related to the plan computation time, which is promising for reliable real-time performance during surgery.

Highlights

In the last few decades, robots have been used in operating rooms to assist surgeons in performing minimally invasive surgery, improving the precision of surgeons and the recovery time of patients (Mack 2001; Vidovszky et al 2006)
We used the length of axioms and the computational time required by ILASP as the evaluation measures; we hypothesized that the learned axioms would closely match the ground truth information
For the experiment that focused on learning action preconditions and executability conditions, the background knowledge of each ILASP learning task included the definitions of sorts and helper axioms describing the difference between two different arms or colors: (, ) ∶ − ( ), ( ), ! =

Summary

Introduction

In the last few decades, robots have been used in operating rooms to assist surgeons in performing minimally invasive surgery, improving the precision of surgeons and the recovery time of patients (Mack 2001; Vidovszky et al 2006). A key limitation of our prior framework was that it assumed comprehensive knowledge of the task and domain in terms of domain attributes (e.g., object properties) and axioms governing domain dynamics (e.g., constraints, and action preconditions and effects). This is not feasible in practical robotics domains, especially in surgical scenarios that are characterized by high variability in the patient’s anatomy. The length of the body of axioms is limited to three atoms using a specific ILASP flag from command line, to reduce the dimension of the problem

Methods

Results

Conclusion