About

Tinoosh Mohsenin is an associate professor of electrical and computer engineering in the Whiting School of Engineering at Johns Hopkins University. She is also an affiliate faculty member of the Johns Hopkins Institute for Assured Autonomy (IAA). She serves as the director of the Energy Efficient High Performance Computing Lab and is a member of the Data Science and AI Institute. Prior to joining Johns Hopkins, she spent 11 years at the University of Maryland, Baltimore County’s Department of Computer Science and Electrical Engineering. Her research focuses on energy efficient computing for signal processing and machine learning used in multi-agent aerial and ground autonomous systems, human and machine teaming, wearable smart health monitoring, and cyber physical systems. She has directed more than $10 million in research funding from NSF, LPS, ARL, DARPA, and various health institutes and industrial sponsors. Mohsenin holds a master’s degree from Rice University obtained in 2004 and a PhD in electrical and computer engineering from the University of California, Davis, earned in 2010. She has authored over 150 peer-reviewed journal and conference publications and has received several awards, including the NSF CAREER award in 2017, best paper awards at ACM Great Lakes VLSI conference 2016, and IEEE Circuits and Systems Symposium 2017, among others.

Research topics

• Computer Science
• Artificial Intelligence
• Machine Learning
• Computer architecture
• Embedded system
• Parallel computing
• Software engineering
• Computer engineering
• Distributed computing

Selected publications

• Embodied Foundation Models at the Edge: A Survey of Deployment Constraints and Mitigation Strategies

  ArXiv.org · 2026-03-16

  articleOpen access

  Deploying foundation models in embodied edge systems is fundamentally a systems problem, not just a problem of model compression. Real-time control must operate within strict size, weight, and power constraints, where memory traffic, compute latency, timing variability, and safety margins interact directly. The Deployment Gauntlet organizes these constraints into eight coupled barriers that determine whether embodied foundation models can run reliably in practice. Across representative edge workloads, autoregressive Vision-Language-Action policies are constrained primarily by memory bandwidth, whereas diffusion-based controllers are limited more by compute latency and sustained execution cost. Reliable deployment therefore depends on system-level co-design across memory, scheduling, communication, and model architecture, including decompositions that separate fast control from slower semantic reasoning.

  PublisherOA PDF

• OA-NBV: Occlusion-Aware Next-Best-View Planning for Human-Centered Active Perception on Mobile Robots

  ArXiv.org · 2026-03-10

  articleOpen accessSenior author

  We naturally step sideways or lean to see around the obstacle when our view is blocked, and recover a more informative observation. Enabling robots to make the same kind of viewpoint choice is critical for human-centered operations, including search, triage, and disaster response, where cluttered environments and partial visibility frequently degrade downstream perception. However, many Next-Best-View (NBV) methods primarily optimize generic exploration or long-horizon coverage, and do not explicitly target the immediate goal of obtaining a single usable observation of a partially occluded person under real motion constraints. We present Occlusion-Aware Next-Best-View Planning for Human-Centered Active Perception on Mobile Robots (OA-NBV), an occlusion-aware NBV pipeline that autonomously selects the next traversable viewpoint to obtain a more complete view of an occluded human. OA-NBV integrates perception and motion planning by scoring candidate viewpoints using a target-centric visibility model that accounts for occlusion, target scale, and target completeness, while restricting candidates to feasible robot poses. OA-NBV achieves over 90% success rate in both simulation and real-world trials, while baseline NBV methods degrade sharply under occlusion. Beyond success rate, OA-NBV improves observation quality: compared to the strongest baseline, it increases normalized target area by at least 81% and keypoint visibility by at least 58% across settings, making it a drop-in view-selection module for diverse human-centered downstream tasks.

  PublisherOA PDF

• SatReg: Regression-based Neural Architecture Search for Lightweight Satellite Image Segmentation

  ArXiv.org · 2026-04-11

  articleOpen accessSenior author

  As Earth-observation workloads move toward onboard and edge processing, remote-sensing segmentation models must operate under tight latency and energy constraints. We present SatReg, a regression-based hardware-aware tuning framework for lightweight remote-sensing segmentation on edge platforms. Using CM-UNet as the teacher architecture, we reduce the search space to two dominant width-related variables, profile a small set of student models on an NVIDIA Jetson Orin Nano, and fit low-order surrogate models for mIoU, latency, and power. Knowledge distillation is used to efficiently train the sampled students. The learned surrogates enable fast selection of near-optimal architecture settings for deployment targets without exhaustive search. Results show that the selected variables affect task accuracy and hardware cost differently, making reduced-space regression a practical strategy for adapting hybrid CNN-Mamba segmentation models to future space-edge systems.

  PublisherOA PDF

• Embodied Foundation Models at the Edge: A Survey of Deployment Constraints and Mitigation Strategies

  Preprints.org · 2026-03-17

  preprintOpen access

  Deploying foundation models in embodied edge systems is fundamentally a systems problem, not just a problem of model compression. Real-time control must operate within strict size, weight, and power constraints, where memory traffic, compute latency, timing variability, and safety margins interact directly. The Deployment Gauntlet organizes these constraints into eight coupled barriers that determine whether embodied foundation models can run reliably in practice. Across representative edge workloads, autoregressive Vision-Language-Action policies are constrained primarily by memory bandwidth, whereas diffusion-based controllers are limited more by compute latency and sustained execution cost. Reliable deployment therefore depends on system-level co-design across memory, scheduling, communication, and model architecture, including decompositions that separate fast control from slower semantic reasoning.

  PublisherOA PDFDOI

• OA-NBV: Occlusion-Aware Next-Best-View Planning for Human-Centered Active Perception on Mobile Robots

  arXiv (Cornell University) · 2026-03-10

  preprintOpen accessSenior author

  We naturally step sideways or lean to see around the obstacle when our view is blocked, and recover a more informative observation. Enabling robots to make the same kind of viewpoint choice is critical for human-centered operations, including search, triage, and disaster response, where cluttered environments and partial visibility frequently degrade downstream perception. However, many Next-Best-View (NBV) methods primarily optimize generic exploration or long-horizon coverage, and do not explicitly target the immediate goal of obtaining a single usable observation of a partially occluded person under real motion constraints. We present Occlusion-Aware Next-Best-View Planning for Human-Centered Active Perception on Mobile Robots (OA-NBV), an occlusion-aware NBV pipeline that autonomously selects the next traversable viewpoint to obtain a more complete view of an occluded human. OA-NBV integrates perception and motion planning by scoring candidate viewpoints using a target-centric visibility model that accounts for occlusion, target scale, and target completeness, while restricting candidates to feasible robot poses. OA-NBV achieves over 90% success rate in both simulation and real-world trials, while baseline NBV methods degrade sharply under occlusion. Beyond success rate, OA-NBV improves observation quality: compared to the strongest baseline, it increases normalized target area by at least 81% and keypoint visibility by at least 58% across settings, making it a drop-in view-selection module for diverse human-centered downstream tasks.

  PublisherDOI

• MedMambaLite-v2: Shared Selective Scan for Efficient Edge Medical Mamba

  IEEE Transactions on Biomedical Circuits and Systems · 2026-01-01

  articleSenior author

  AI-powered medical imaging devices are increasingly used in clinical workflows to support real-time, accurate diagnosis and decision-making. Recent advances in State-Space Models (SSMs) such as Mamba have shown remarkable performance in capturing long-range dependencies for medical image classification. However, their computational complexity and sequential data flow make them difficult to deploy on hardware, limiting real-time and energy-efficient applications at the edge. To address this challenge, we propose MedMambaLite-v2, a shared selective scan framework enabling effective acceleration on embedded edge platforms. For this aim, we build upon our earlier MedMambaLite, and further extend it in MedMambaLite-v2 through a channel-only transition mechanism that achieves a 1.7× reduction in operations. We then optimize the Convolution (Conv) branch, and apply knowledge distillation to retain accuracy in a compressed student model. The resulting model is 23× smaller compared to the MedMamba baseline, with only 1.1% reduction in the overall accuracy evaluated across 10 distinct MedMNIST datasets spanning several imaging modalities. The proposed LiteSS2D hardware design also leverages parallelism across scan directions to enable simultaneous state updates, thereby improving memory efficiency, and further incorporates 8-bit quantization to reduce computational overhead. The reconfigurable FPGA hardware prototype demonstrates 9× reduction in latency for a parallel implementation compared with a serial baseline. Moreover, MedMambaLite-v2 is implemented and demonstrated through end-to-end inference on MedMNIST images on CPU and GPU platforms. Performance analysis of the proposed approach on NVIDIA Jetson Orin Nano and Raspberry Pi 5 shows up to 63% and 78% reductions in energy per inference, respectively, compared to the baseline.

  PublisherDOI

• Embodied Foundation Models at the Edge: A Survey of Deployment Constraints and Mitigation Strategies

  arXiv (Cornell University) · 2026-03-16

  preprintOpen access

  Deploying foundation models in embodied edge systems is fundamentally a systems problem, not just a problem of model compression. Real-time control must operate within strict size, weight, and power constraints, where memory traffic, compute latency, timing variability, and safety margins interact directly. The Deployment Gauntlet organizes these constraints into eight coupled barriers that determine whether embodied foundation models can run reliably in practice. Across representative edge workloads, autoregressive Vision-Language-Action policies are constrained primarily by memory bandwidth, whereas diffusion-based controllers are limited more by compute latency and sustained execution cost. Reliable deployment therefore depends on system-level co-design across memory, scheduling, communication, and model architecture, including decompositions that separate fast control from slower semantic reasoning.

  PublisherDOI

• Embodied Foundation Models at the Edge: A Survey of Deployment Constraints and Mitigation Strategies

  2026-01-01

  articleOpen access
  PublisherDOI

• SatReg: Regression-based Neural Architecture Search for Lightweight Satellite Image Segmentation

  arXiv (Cornell University) · 2026-04-11

  preprintOpen accessSenior author

  As Earth-observation workloads move toward onboard and edge processing, remote-sensing segmentation models must operate under tight latency and energy constraints. We present SatReg, a regression-based hardware-aware tuning framework for lightweight remote-sensing segmentation on edge platforms. Using CM-UNet as the teacher architecture, we reduce the search space to two dominant width-related variables, profile a small set of student models on an NVIDIA Jetson Orin Nano, and fit low-order surrogate models for mIoU, latency, and power. Knowledge distillation is used to efficiently train the sampled students. The learned surrogates enable fast selection of near-optimal architecture settings for deployment targets without exhaustive search. Results show that the selected variables affect task accuracy and hardware cost differently, making reduced-space regression a practical strategy for adapting hybrid CNN-Mamba segmentation models to future space-edge systems.

  PublisherDOI

• Metareasoning for Edge-Cloud Collaborative LLM Planning for Efficient Autonomous Navigation

  ACM Transactions on Embedded Computing Systems · 2025-09-05 · 3 citations

  articleOpen accessSenior author

  Cutting-edge Large Language Models (LLMs) play a crucial role in improving autonomous navigation by offering efficient solutions. While LLMs require powerful computers to operate, security concerns and maintaining a stable connection with the cloud can be challenging due to various factors. To address this issue, we propose a metareasoning approach for edge-cloud collaborative LLM planning which leads to an efficient autonomous navigation. The proposed approach allows the system to seamlessly switch between cloud and edge devices to fulfill the mission even in the event of a lost connection or entering a GPS-denied environment. Moreover, we deploy state-of-the-art LLM models on resource-constrained systems like the NVIDIA Jetson Orin Nano 8GB, integrated with ROSMASTER X3. These LLMs have demonstrated exceptional utility in dynamic planning for multi-room or maze environments. A comprehensive LLM profiling of TinyLLM models was performed for five different LLMs. The LLM profiling result shows that while certain models with smaller sizes and lower power consumption were available, their accuracy was insufficient for our application requirements. As a result, LLaMa2-7B is considered the edge LLM model due to its optimal balance of performance and accuracy. The experimental results show that under weak signal conditions (<-50 dB), the metareasoning approach improves energy consumption by up to 4x while the cloud-based implementation exceeds the energy consumption of the onboard LLM implementation. Moreover, with delays of 10-20 seconds, cloud implementation becomes impractical for real-time applications in weak signal environments. This underscores the need for metareasoning, which optimizes energy consumption and response time, providing a balanced solution by adapting to signal strength. A real-world implementation of the proposed approach on ROSMASTER X3 with NVIDIA Jetson Orin Nano board can be found in this video which shows that the mission was completed despite losing the connection with cloud-based LLM.

  PublisherDOI

Recent grants

• CSR: Small:Collaborative Research:Heterogeneous Ultra Low Power Accelerator for Wearable Biomedical Computing

  NSF · $212k · 2015–2019

• CSR: EAGER: Multi-physiological Signal Processing Architectures for Seizure Detection

  NSF · $100k · 2013–2016

• CAREER: DeepMatter: A Scalable and Programmable Embedded Deep Neural Network

  NSF · $475k · 2017–2023

Frequent coauthors

• Houman Homayoun

  University of California, Davis

  39 shared

• Tim Oates

  27 shared

• Nicholas R. Waytowich

  DEVCOM Army Research Laboratory

  25 shared

• Morteza Hosseini

  University of Maryland, Baltimore County

  23 shared

• Arnab Neelim Mazumder

  University of Maryland, Baltimore County

  22 shared

• Avesta Sasan

  University of California, Davis

  21 shared

• Hasib-Al Rashid

  21 shared

• Bevan Baas

  University of California, Davis

  21 shared

Awards & honors

• NSF CAREER award (2017)
• best paper award in the ACM Great Lakes VLSI conference 2016
• best paper honorable award in the IEEE Circuits and Systems…
• ISSCC 2020 Evening Session Award for co-organizing a session…