Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Engineering Automated Microrobotic Systems with Machine Learning-Based Control for Biomedical Applications

Behrens, Michael R. (2022) Engineering Automated Microrobotic Systems with Machine Learning-Based Control for Biomedical Applications. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

PDF (behrensmr_etdPitt2021.pdf)
Primary Text

Download (7MB) | Preview
[img] Video (MP4) (Supplementary Movie S1: Training process time course)
Download (15MB)
[img] Video (MP4) (Supplementary Movie S2: HAMR swimming around the track)
Download (28MB)


For many diseases and injuries, the pathological effects are local to particular areas of the body. In such cases, systemic drug delivery is nonideal treatment because drugs used to treat the pathological condition may have a deleterious effect on the remaining healthy tissue and cause damaging side effects. A similar kind of harm is done in many surgical scenarios when healthy tissue must be cut in order to give the surgeon physical access to the damaged tissue underneath. There exists an unmet clinical need to be able to deliver minimally invasive therapeutic interventions precisely within the body, while minimizing harm to healthy tissue. Towards this end, we are developing swimming microrobotic systems to traverse the fluidic pathways within the body, in order to deliver locally targeted therapeutic payloads noninvasively, and with precision, to areas which are difficult to access noninvasively using current methods.
In this dissertation, I present a control approach which uses machine learning in order to develop robust control policies for magnetically actuated microrobots navigating within small scale fluidic environments. We have developed a controller for a helical magnetic hydrogel microrobot that uses the soft actor critic reinforcement learning algorithm to autonomously derive control policies which allow the microrobot to swim through an uncharacterized biomimetic fluidic environment, actuated by time varying magnetic fields generated from a three-axis array of electromagnets. Our results show that this model-free control approach can learn and adapt to the dynamic features of the microrobot, the fluidic environment, and the electromagnetic actuator in order to successfully accomplish navigation tasks. Additionally, we demonstrate that the control policies learned by the reinforcement learning algorithm repeatedly converge to recapitulate the behavior of rationally designed controllers based on physical models of helical swimming microrobots. We also present an approach for developing biohybrid magnetic microrobots with chemical sensing capabilities based on genetically encoded surface chemistry, and explore hydrogel based magnetic microrobots as vehicles for antibiotic delivery. Together, these contributions form a suite of microrobotic technologies for autonomous navigation, sensing, and localized drug delivery within small scale fluidic environments, and are an important step towards clinical applications of microrobotic technology.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Behrens, Michael R.mrb157@pitt.edumrb1570000-0002-8188-034X
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairRuder, Warren
Committee MemberDavidson,
Committee MemberAmbrosio,
Committee MemberWeber,
Date: 16 January 2022
Date Type: Publication
Defense Date: 9 November 2021
Approval Date: 16 January 2022
Submission Date: 26 October 2021
Access Restriction: 1 year -- Restrict access to University of Pittsburgh for a period of 1 year.
Number of Pages: 196
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Bioengineering
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: microrobotics, machine learning, reinforcement learning, synthetic biology
Date Deposited: 16 Jan 2023 06:00
Last Modified: 16 Jan 2023 06:15


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item