About

Hamsa Balakrishnan is the William E. Leonhard (1940) Professor of Aeronautics and Astronautics at MIT AeroAstro. She is the principal investigator associated with the research group, focusing on advanced air mobility, AI-assisted optimization of schedules, modeling and control of queuing networks, multi-agent navigation in dynamic environments, network models of air transportation, and trajectory-based operations. Her work involves developing innovative solutions and methodologies to improve air transportation systems and mobility through interdisciplinary research and technological advancements.

Research topics

• Computer Science
• Engineering
• Speech recognition
• Telecommunications
• Automotive engineering

Selected publications

• Vehicle-Based Multi-Services for Future Smart Cities

  Service Science · 2026-03-17 · 1 citations

  article

  Vehicles are crucial for sustaining socioeconomic activity and improving quality of life in modern cities by offering diverse services. These include passenger mobility, goods delivery, information acquisition, and acting as mobile servers such as food trucks and mobile lockers. At the same time, they also contribute to traffic congestion and air pollution. This tension fosters the rise of urban resource-conserving and sustainable service solutions. In this article, we introduce the concept of “Vehicle-Based Multi-Services” (VeMuS), in which a single vehicle offers multiple services simultaneously. Drawing on practical use cases, we examine service classification and integration for vehicles and the potential for the synergy effect and multitasking. We construct a general framework to study VeMuS and propose research questions in three key components: VeMuS design, VeMuS operations, and VeMuS evaluation. This article highlights the potential of VeMuS to create more sustainable, efficient, and adaptable urban services for future smart cities. Funding: This research is supported by the National Research Foundation (NRF), Prime Minister’s Office, Singapore under its Campus for Research Excellence and Technological Enterprise (CREATE) programme. The Mens, Manus, and Machina (M3S) is an interdisciplinary research group (IRG) of the Singapore MIT Alliance for Research and Technology (SMART) centre. H. Sun is in part supported by the National Natural Science Foundation of China [Grant 72301179], the Guangdong Basic and Applied Basic Research Foundation [Grant 2025A1515012815], and the “Pengcheng Peacock Project” of Shenzhen University.

  PublisherDOI

• Orbital Optimization of a Distributed Heliocentric Relay Network for Mars-Earth Communications

  2026-03-07

  articleSenior author

  <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\boldsymbol{A} \boldsymbol{b} \boldsymbol{s} \boldsymbol{t r} \boldsymbol{a} \boldsymbol{c} \boldsymbol{t}$</tex>-Deep-space missions have long depended on Direct-toEarth communications via NASA's Deep Space Network, but the network's limited capacity and the growing data downlink demands of modern missions pose major challenges for future robotic and crewed space exploration. To supplement the Deep Space Network, we propose a Heliocentric Relay Network (HRN): a distributed constellation of Ka-band relay satellites in solar orbit, designed for continuous high-data-rate Mars-Earth communications. We develop a modular hidden-genes genetic algorithm to determine HRN satellite orbits that maximize the average Mars-Earth downlink capacity over a 6.4 -year simulation period. We show that with 25-100 Starship launches, average Mars-Earth data rates through the HRN range from 146 Mbps to 852 Mbps, more than 100 times the throughput of current Direct-to-Earth systems. These results demonstrate that a large-scale HRN using technologically-mature communication systems could enable high-data-rate, continuous Mars-Earth communications, supporting future exploration across the inner Solar System.

  PublisherDOI

• Designing Dense Satellite Clusters for Distributed Space-based Datacenters

  ArXiv.org · 2026-05-14

  articleOpen accessSenior author

  Recent proposals for datacenters in sun-synchronous Low Earth Orbit rely on a large number of compute satellites formation-flying in dense clusters. Designing such satellite clusters requires optimizing the satellites' orbital geometry under several safety and operational constraints applied throughout the cluster's entire orbit. These constraints include guaranteeing a minimum inter-satellite spacing, obstruction-less solar power for every satellite, and that each satellite have a stable set of nearest neighbors with which it can maintain inter-satellite links (ISLs). In this work, we propose two main cluster orbital designs, parametrized by the minimum inter-satellite spacing $R_{min}$ and the cluster radius $R_{max}$: a planar cluster, and a 3D cluster. We show by construction and numerical analysis that both cluster orbital designs are consistent with the inter-satellite spacing, unobstructed sun-vector, and inter-satellite line of sight constraints. The proposed planar architecture is the most efficient packing of satellites in a plane for given $R_{min}$ and $R_{max}$ values, and our 3D architecture allows for the number of datacenter satellites to scale proportional to $(R_{max}/R_{min})^3$, an improvement over all previous LEO datacenter cluster designs. Finally, for a given satellite cluster, we formulate and solve an integer optimization problem that maps a VL2-like Clos network datacenter switching fabric onto the satellites and their corresponding set of feasible ISLs. We confirm that for both the planar and 3D architectures, there are sufficiently many permanently unobstructed ISLs within the cluster to replicate the switching fabric of terrestrial datacenters. We also examine the tradeoff between the number of ISLs each satellite can simultaneously sustain, and the corresponding number of cluster satellites that must be dedicated as aggregation and intermediate switches.

  PublisherOA PDF

• Designing Dense Satellite Clusters for Distributed Space-based Datacenters

  arXiv (Cornell University) · 2026-05-14

  preprintOpen accessSenior author

  Recent proposals for datacenters in sun-synchronous Low Earth Orbit rely on a large number of compute satellites formation-flying in dense clusters. Designing such satellite clusters requires optimizing the satellites' orbital geometry under several safety and operational constraints applied throughout the cluster's entire orbit. These constraints include guaranteeing a minimum inter-satellite spacing, obstruction-less solar power for every satellite, and that each satellite have a stable set of nearest neighbors with which it can maintain inter-satellite links (ISLs). In this work, we propose two main cluster orbital designs, parametrized by the minimum inter-satellite spacing $R_{min}$ and the cluster radius $R_{max}$: a planar cluster, and a 3D cluster. We show by construction and numerical analysis that both cluster orbital designs are consistent with the inter-satellite spacing, unobstructed sun-vector, and inter-satellite line of sight constraints. The proposed planar architecture is the most efficient packing of satellites in a plane for given $R_{min}$ and $R_{max}$ values, and our 3D architecture allows for the number of datacenter satellites to scale proportional to $(R_{max}/R_{min})^3$, an improvement over all previous LEO datacenter cluster designs. Finally, for a given satellite cluster, we formulate and solve an integer optimization problem that maps a VL2-like Clos network datacenter switching fabric onto the satellites and their corresponding set of feasible ISLs. We confirm that for both the planar and 3D architectures, there are sufficiently many permanently unobstructed ISLs within the cluster to replicate the switching fabric of terrestrial datacenters. We also examine the tradeoff between the number of ISLs each satellite can simultaneously sustain, and the corresponding number of cluster satellites that must be dedicated as aggregation and intermediate switches.

  PublisherDOI

• Decentralized Coordination of Autonomous Traffic Through Advanced Air Mobility Corridors

  2026-01-08

  articleSenior author

  The use of dedicated corridors for Advanced Air Mobility (AAM) traffic is one of the most commonly proposed pathways to integrating them into existing airspace operations. Most prior research has focused on the design of networks of AAM corridors and conflict resolution for aircraft within corridors. It is also generally believed that while attractive from an implementation perspective, corridor-based operations may be inefficient, especially in the absence of centralized traffic management. In this paper, we show that contrary to this belief, it is possible for autonomous aircraft to learn to self-organize into corridor flows in decentralized settings. We illustrate our approach using scenarios in which fixed-wing aircraft need to safely and efficiently traverse (1) a single corridor with metering after the exit, (2) a sequence of two consecutive corridors, and (3) a corridor that splits into two. We find that in decentralized settings with only local information, the aircraft are able to conform to the corridor boundaries more than 94% of the time and reach their goal in a relatively efficient manner. Furthermore, tactical interventions to handle violations of the separation minimum are needed only infrequently in low- and medium-density settings. However, such tactical interventions become more frequently necessary only when traffic density is high.

  PublisherDOI

• Privacy-Aware Routing for Drone Delivery as a Service

  Journal of Air Transportation · 2026-02-22

  article

  Uncrewed aerial vehicles (UAVs), or drones, are increasingly being used to deliver goods from vendors to customers. In an emerging business model, a drone operator partners with multiple businesses to offer drone delivery as a service. However, this business model poses a privacy risk due to regulations requiring drones to broadcast location information. Third-party observers may leverage broadcast trajectories to link customers to vendors, resulting in potential privacy issues. A probabilistic definition of privacy risk is proposed based on the likelihood that a third-party observer can correctly infer which vendor a customer ordered from. Next, these risks are quantified, and the impacts of order count, drone capacity, decoy stops, and delivery time requirements on privacy are evaluated. Privacy risk mitigation is then formulated as an optimization problem and integrated into the vehicle routing problem. Doing so allows optimization of efficiency while satisfying a privacy constraint. A comparison of the tradeoffs between privacy and efficiency based on the geographical arrangement of vendors and customers is presented. Finally, limitations, including complexity, assumptions, and areas for future work, are discussed.

  PublisherDOI

• Resolving Conflicting Constraints in Multi-Agent Reinforcement Learning with Layered Safety

  2025-06-21 · 2 citations

  preprintOpen access

  Preventing collisions in multi-robot navigation is crucial for deployment. This requirement hinders the use of learning-based approaches, such as multi-agent reinforcement learning (MARL), on their own due to their lack of safety guarantees. Traditional control methods, such as reachability and control barrier functions, can provide rigorous safety guarantees when interactions are limited only to a small number of robots. However, conflicts between the constraints faced by different agents pose a challenge to safe multi-agent coordination. To overcome this challenge, we propose a method that integrates multiple layers of safety by combining MARL with safety filters. First, MARL is used to learn strategies that minimize multiple agent interactions, where multiple indicates more than two. Particularly, we focus on interactions likely to result in conflicting constraints within the engagement distance. Next, for agents that enter the engagement distance, we prioritize pairs requiring the most urgent corrective actions. Finally, a dedicated safety filter provides tactical corrective actions to resolve these conflicts. Crucially, the design decisions for all layers of this framework are grounded in reachability analysis and a control barrier-value function-based filtering mechanism. We validate our Layered Safe MARL framework in 1) hardware experiments using Crazyflie drones and 2) high-density advanced aerial mobility (AAM) operation scenarios, where agents navigate to designated waypoints while avoiding collisions. The results show that our method significantly reduces conflict while maintaining safety without sacrificing much efficiency (i.e., shorter travel time and distance) compared to baselines that do not incorporate layered safety. The project website is available at https://dinamo-mit.github.io/Layered-Safe-MARL/

  PublisherDOI

• Asynchronous Cooperative Multi-Agent Reinforcement Learning with Limited Communication

  ArXiv.org · 2025-02-01 · 1 citations

  preprintOpen accessSenior author

  We consider the problem setting in which multiple autonomous agents must cooperatively navigate and perform tasks in an unknown, communication-constrained environment. Traditional multi-agent reinforcement learning (MARL) approaches assume synchronous communications and perform poorly in such environments. We propose AsynCoMARL, an asynchronous MARL approach that uses graph transformers to learn communication protocols from dynamic graphs. AsynCoMARL can accommodate infrequent and asynchronous communications between agents, with edges of the graph only forming when agents communicate with each other. We show that AsynCoMARL achieves similar success and collision rates as leading baselines, despite 26\% fewer messages being passed between agents.

  PublisherOA PDFDOI

• An Important Suspect: Fungal Pneumonia in Intensive Care: A Case Series

  American Journal of Respiratory and Critical Care Medicine · 2025-05-01

  articleSenior author

  Abstract Introduction: Fungal pneumonia poses a significant challenge in intensive care. In immunocompromised patients, diagnosis is challenging due to non-specific presentations. Challenge is harder, in patients without classical risk factors, as most guidelines for fungal pneumonia in intensive care focus on high risk patients. This case series aims to highlight fungi as an important suspect in pneumonia in intensive care and the need for a standardised guideline, irrespective of host factors. Case Description:Case 1-A case of Severe influenza pneumonia, initially on mechanical ventilation and weaned, developed respiratory distress and worsening radiographic abnormalities. Endobronchial biopsy grew Aspergillus fumigatus. Injection voriconazole was started. Patient succumbed due to multiple comorbidities, including heart failure and cerebrovascular accident. Case 2-Patient presented with breathlessness, productive cough and fever for one month. He is a chronic obstructive pulmonary disease(COPD) patient on intermittent oral corticosteroids. Despite antibiotic therapy, his serial Computed tomography(CT) chest showed increasing consolidation and multiple cavities in right upper lobe and lower lobe. Sputum culture grew Aspergillus terreus and mucormycosis. He was treated with liposomal amphotericin B. Case 3-Patient presented with cough with expectoration, breathlessness and fever for 20 days. CT chest showed bilateral multiple thick walled cavities with intracavitary strands and surrounding consolidation suggestive of fungal pneumonia. On further evaluation, he was diagnosed with diabetes mellitus. Bronchial wash culture grew Rhizopus. Liposomal amphotericin B was started. Surgical debridement deferred due to bilateral involvement. Case 4-A case of chronic glomerulonephritis, post-renal transplant, was admitted with complaints of dry cough and low grade fever for one week. His CT chest showed left lower lobe thick walled cavity. Bronchial wash culture grew Rhizopus. Patient was started on antifungal and underwent left lower lobectomy, which showed necrosis and angioinvasion by fungal hyphae. He was given liposomal amphotericin B followed by posaconazole. Discussion: Diagnosis of fungal pneumonia poses many challenges. Cases 1-2 signifies that, in intensive care, evaluation for fungal pneumonia is to be considered, even when typical risk factors are absent, especially in cases of severe viral pneumonia, prolonged hospitalization, non resolving pneumonia and structural lung disease. Case 3 signifies if typical radiographic features are present, fungal pneumonia to be suspected and evaluation for risk factors to be done. Case 4 signifies that even when clinical features are non-specific, evaluation is required if patient is immunocompromised. This case series also calls for the need for a comprehensive diagnostic algorithm for fungal pneumonia in intensive care units, irrespective of host factors.

  PublisherDOI

• Vehicle Routing Problem Formulation for Efficient Tracking of Objects in Low Earth Orbit

  2025-01-03

  articleOpen accessSenior author

  The increasing number of resident space objects (RSOs) in low Earth orbit (LEO) endangers the sustainable use of space and necessitates continuous surveillance to prevent collisions. The U.S. Space Surveillance Network (SSN) tracks tens of thousands of LEO RSOs using a suite of ground-based sensors; however, the algorithms that task and schedule these sensors have not improved significantly in the last twenty years. In that time, the number of catalogued LEO RSOs has more than doubled, calling for more efficient tasking algorithms. Prior research has primarily focused on improving the tasking of ground-based sensors for tracking RSOs in geosynchronous Earth orbit (GEO). In this paper, we extend recent work on a vehicle routing problem (VRP) formulation for optimal tasking and scheduling of ground-based radars for tracking GEO RSOs and apply it to tracking LEO RSOs. We introduce a modified VRP formulation, which features discrete time indexing and leverages sparse, binary feasibility matrices for reduced computation time, and present results for several simulations. We show that our approach can compute global and regional optima for tracking (a) 100 targets using 4 ground-based sensors over a 5-hour time horizon in under 5 minutes on a laptop computer and (b) 10,000 targets using 27 ground-based sensors over a 24-hour time horizon in about 4 hours on a high-performance computing cluster.

  PublisherOA PDFDOI

Recent grants

• CPS: Small: Recovery Algorithms for Dynamic Infrastructure Networks

  NSF · $450k · 2017–2022

• CAREER: Practical Algorithms for Next Generation Air Transportation Systems

  NSF · $400k · 2008–2014

Frequent coauthors

• Karthik Gopalakrishnan

  Amazon (United States)

  73 shared

• Max Z. Li

  36 shared

• Harshad Khadilkar

  Tata Consultancy Services (India)

  31 shared

• Sandeep Badrinath

  27 shared

• Joseph M. Sussman

  Rose–Hulman Institute of Technology

  19 shared

• Christopher Chin

  Massachusetts Institute of Technology

  19 shared

• Tom Reynolds

  Massachusetts Institute of Technology

  19 shared

• Ioannis Simaiakis

  Massachusetts Institute of Technology

  18 shared

Labs

• Dinamo GroupPI

Awards & honors

• Vickie Kerrebrock Award, MIT AeroAstro, 2021
• Undergraduate Teaching Award, American Institute of Aeronaut…
• Donald P. Eckman Award | American Automatic Control Council,…
• AIAA Undergraduate Advising Award, MIT, 2014
• Associate Fellow, AIAA, 2013