2021
DOI: 10.48550/arxiv.2109.02323
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

Safe Reinforcement Learning using Formal Verification for Tissue Retraction in Autonomous Robotic-Assisted Surgery

Abstract: Deep Reinforcement Learning (DRL) is a viable solution for automating repetitive surgical subtasks due to its ability to learn complex behaviours in a dynamic environment. This task automation could lead to reduced surgeon's cognitive workload, increased precision in critical aspects of the surgery, and fewer patient-related complications. However, current DRL methods do not guarantee any safety criteria as they maximise cumulative rewards without considering the risks associated with the actions performed. Du… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

0
4
0

Year Published

2021
2021
2022
2022

Publication Types

Select...
2

Relationship

0
2

Authors

Journals

citations
Cited by 2 publications
(4 citation statements)
references
References 16 publications
0
4
0
Order By: Relevance
“…We use proximal policy optimization (PPO) [65] (as in [53]) with hindsight experience replay (HER) [66] for model-free RL. We use our DeformerNet architecture for the actor and critic network except for the critic output being set to a single scalar to encode the value function.…”
Section: ) Comparison With Other Planning Methodsmentioning
confidence: 99%
See 1 more Smart Citation
“…We use proximal policy optimization (PPO) [65] (as in [53]) with hindsight experience replay (HER) [66] for model-free RL. We use our DeformerNet architecture for the actor and critic network except for the critic output being set to a single scalar to encode the value function.…”
Section: ) Comparison With Other Planning Methodsmentioning
confidence: 99%
“…Nagy et al [52] propose the use of stereo vison accompanied by multiple control methods, however the method assumes a thin tissue layer and a clear view of two tissue layers. Pore et al [53] introduce a model-free reinforcement learning method which learns safe motions for a robot's end effector during retraction, however it does not explicitly reason over the deformation of the tissue. We compare against a similar approach, using a model-free reinforcement learning algorithm, but adapted to our task to explicitly reason over the tissue state.…”
Section: Related Workmentioning
confidence: 99%
“…Nagy et al [36] propose the use of stereo vison accompanied by multiple control methods, however the method assumes a thin tissue layer and a clear view of two tissue layers. Pore et al [37] introduce a model-free reinforcement learning method which learns safe motions for a robot's end effector during retraction, however it does not explicitly reason over the deformation of the tissue. We compare against a similar approach, using the same model-free reinforcement learning algorithm, but adapted to our task to explicitly reason over the tissue state.…”
Section: Related Workmentioning
confidence: 99%
“…We use proximal policy optimization (PPO) [46] (as in [37]) with hindsight experience replay (HER) [47] for model-free RL. We use our DeformerNet architecture for the actor and critic network except for the critic output being set to single scalar to encode the value function.…”
Section: A Goal-oriented Shape Servoingmentioning
confidence: 99%