Efstathios Bakolas
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Efstathios Bakolas has a total of 51 co-authors. Here are the top co-authors by number of shared publications:
- Alexandros Tsolovikos
Department of Aerospace Engineering and Engineering Mechanics, The University of Texas at Austin , Austin, TX 78712
6 shared publications - Vrushabh Zinage
Dept. of Computer Science 1 Dept. of Aerospace Engineering and Engineering Mechanics The University of Texas at Austin Project hosted
5 shared publications Yoonjae Lee
The University of Texas at Austin
5 shared publications- Yoonjae Lee
Department of Aerospace Engineering and Engineering Mechanics, The University of Texas at Austin, Austin, TX, USA
5 shared publications David Goldstein
University of Texas at Austin
4 shared publications- Rohan Chandra
Dept. of Computer Science University of Virginia
4 shared publications
Latest Publications
- 1 November 2026
- Published by American Institute of Aeronautics and Astronautics (AIAA) in Journal of Aerospace Information Systems
- Vol. 23 (10) , 923-939
Goal-Oriented Multimodal Motion Prediction for Multi-Agent Systems
This paper presents a method for motion prediction and intent estimation of multi-agent systems while considering intrinsic groupings within the system. Inherent groupings of agents in real-world applications of multi-agent systems can be exploited to obtain more adaptable and realistic predictions of movement patterns in challenging scenarios, such as search and rescue operations or monitoring/supervising drone delivery systems in crowded spaces. This work aims to take advantage of these inherent groupings to provide a less complex prediction space, estimate potential agent intents, and utilize such information to further refine grouping predictions. To this end, a multiple-model unscented Kalman filter is combined with a variation of an agglomerative hierarchical clustering scheme to predict and improve agent positions depending on available data and information. The proposed method is verified in three scenarios pertaining to different applications: an evacuation scenario relevant to search and rescue missions, a collision avoidance scheme in the presence of moving obstacles that couples prediction and planning methods, and a three-dimensional drone emergency landing scenario.
- 1 November 2026
- Published by American Institute of Aeronautics and Astronautics (AIAA) in Journal of Aerospace Information Systems
- Vol. 23 (10) , 923-939
Goal-Oriented Multimodal Motion Prediction for Multi-Agent Systems
This paper presents a method for motion prediction and intent estimation of multi-agent systems while considering intrinsic groupings within the system. Inherent groupings of agents in real-world applications of multi-agent systems can be exploited to obtain more adaptable and realistic predictions of movement patterns in challenging scenarios, such as search and rescue operations or monitoring/supervising drone delivery systems in crowded spaces. This work aims to take advantage of these inherent groupings to provide a less complex prediction space, estimate potential agent intents, and utilize such information to further refine grouping predictions. To this end, a multiple-model unscented Kalman filter is combined with a variation of an agglomerative hierarchical clustering scheme to predict and improve agent positions depending on available data and information. The proposed method is verified in three scenarios pertaining to different applications: an evacuation scenario relevant to search and rescue missions, a collision avoidance scheme in the presence of moving obstacles that couples prediction and planning methods, and a three-dimensional drone emergency landing scenario.
- 27 August 2026
- Published by American Institute of Aeronautics and Astronautics (AIAA) in Journal of Aerospace Information Systems
- p. 1-11
Coupled Task Assignment for Hierarchical Multi-Agent Systems: A Nested Hungarian Approach
This work formulates a novel variant of the classical linear task assignment problem, which we refer to as the coupled linear task assignment problem. This formulation models the task assignment process involving two-echelon agents and tasks, aiming to optimally assign tasks to agents within the same echelon to maximize global utility while satisfying the coupling constraints imposed by the hierarchical structure. After discussing the problem’s inherent properties that prevent the use of standard relaxation techniques, we propose a variant of the Hungarian algorithm, the nested Hungarian algorithm, that solves the problem exactly in polynomial time. We demonstrate the superior performance of the proposed algorithm by benchmarking it against seven open-source and commercial MILP solvers. Finally, we apply our proposed approach to a fictitious disaster response scenario motivated by the January 2025 Southern California wildfires in the Greater Los Angeles area.
- 27 August 2026
- Published by American Institute of Aeronautics and Astronautics (AIAA) in Journal of Aerospace Information Systems
- p. 1-11
Coupled Task Assignment for Hierarchical Multi-Agent Systems: A Nested Hungarian Approach
This work formulates a novel variant of the classical linear task assignment problem, which we refer to as the coupled linear task assignment problem. This formulation models the task assignment process involving two-echelon agents and tasks, aiming to optimally assign tasks to agents within the same echelon to maximize global utility while satisfying the coupling constraints imposed by the hierarchical structure. After discussing the problem’s inherent properties that prevent the use of standard relaxation techniques, we propose a variant of the Hungarian algorithm, the nested Hungarian algorithm, that solves the problem exactly in polynomial time. We demonstrate the superior performance of the proposed algorithm by benchmarking it against seven open-source and commercial MILP solvers. Finally, we apply our proposed approach to a fictitious disaster response scenario motivated by the January 2025 Southern California wildfires in the Greater Los Angeles area.
- 27 June 2026
- Published by American Institute of Aeronautics and Astronautics (AIAA) in AIAA Journal
- Vol. 64 (10) , 1-14
Adaptive Feedback Flow Control for Wings
We present an adaptive flow control strategy for a turbulent separated flow using online dynamic mode decomposition. The system dynamics are identified in real time, and a linear quadratic tracking controller regulates the separation location along a predefined target. Controller gains are updated continuously based on the new system updates, allowing for an adaptive formulation of real-time feedback control that can adjust to both variations in the freestream flow due to external disturbances and actuation-induced changes of the flow dynamics. The approach is demonstrated on a NACA4412 wing at [ ] using direct numerical simulation at 0 and 10 deg angles of attack, outperforming steady blowing in moving the separation point downstream. At 10 deg, adaptive control increases lift to drag by approximately 60%. The framework is further extended to directly regulate the lift-to-drag ratio, illustrating its potential for real-time control of aerodynamic forces in unsteady, separated flows.
- 27 June 2026
- Published by American Institute of Aeronautics and Astronautics (AIAA) in AIAA Journal
- Vol. 64 (10) , 1-14
Adaptive Feedback Flow Control for Wings
We present an adaptive flow control strategy for a turbulent separated flow using online dynamic mode decomposition. The system dynamics are identified in real time, and a linear quadratic tracking controller regulates the separation location along a predefined target. Controller gains are updated continuously based on the new system updates, allowing for an adaptive formulation of real-time feedback control that can adjust to both variations in the freestream flow due to external disturbances and actuation-induced changes of the flow dynamics. The approach is demonstrated on a NACA4412 wing at [ ] using direct numerical simulation at 0 and 10 deg angles of attack, outperforming steady blowing in moving the separation point downstream. At 10 deg, adaptive control increases lift to drag by approximately 60%. The framework is further extended to directly regulate the lift-to-drag ratio, illustrating its potential for real-time control of aerodynamic forces in unsteady, separated flows.
- 7 April 2026
- Published by Elsevier in Robotics and Autonomous Systems
- Vol. 202
Motion prediction of multi-agent systems with multi-view clustering
- 7 April 2026
- Published by Elsevier in Robotics and Autonomous Systems
- Vol. 202
Motion prediction of multi-agent systems with multi-view clustering
- 1 April 2026
- Published by American Institute of Aeronautics and Astronautics (AIAA) in Journal of Guidance, Control, and Dynamics
- Vol. 49 (4) , 1023-1033
Game-Theoretic Distributed Sensor Tasking in Earth Observation Satellite Constellations
This paper presents a game-theoretic sensor selection for coordinating multiple satellites, where each satellite adjusts its roll angle to observe multiple targets. The main objective is to find a joint sensor schedule for the satellite constellation over a finite time horizon that minimizes both mean square error (MSE) and slew costs, which is a challenging combinatorial optimization problem. Standard MSE-based sensor scheduling problems typically involve optimizing non-submodular functions, which introduces computational complexity in designing algorithms that achieve provable performance guarantees. To address this issue, we propose a submodular surrogate objective function that is monotonic under specific conditions for the weights on MSE and slew cost terms. This reformulation enables the problem to be modeled as a potential game, facilitating the use of a distributed algorithm to find a Nash equilibrium that provides a bounded efficiency. The proposed solution is compared with an alternative game-theoretic solution using a different performance metric, as well as with a centralized greedy solution, to demonstrate the performance of the surrogate approach, particularly in the presence of isolated subgraphs and stochastic communication delays. Additionally, the versatility of the proposed scheme is highlighted, as it is applicable to heterogeneous satellites and sensors with disconnected communication graphs.
- 1 April 2026
- Published by American Institute of Aeronautics and Astronautics (AIAA) in Journal of Guidance, Control, and Dynamics
- Vol. 49 (4) , 1023-1033
Game-Theoretic Distributed Sensor Tasking in Earth Observation Satellite Constellations
This paper presents a game-theoretic sensor selection for coordinating multiple satellites, where each satellite adjusts its roll angle to observe multiple targets. The main objective is to find a joint sensor schedule for the satellite constellation over a finite time horizon that minimizes both mean square error (MSE) and slew costs, which is a challenging combinatorial optimization problem. Standard MSE-based sensor scheduling problems typically involve optimizing non-submodular functions, which introduces computational complexity in designing algorithms that achieve provable performance guarantees. To address this issue, we propose a submodular surrogate objective function that is monotonic under specific conditions for the weights on MSE and slew cost terms. This reformulation enables the problem to be modeled as a potential game, facilitating the use of a distributed algorithm to find a Nash equilibrium that provides a bounded efficiency. The proposed solution is compared with an alternative game-theoretic solution using a different performance metric, as well as with a centralized greedy solution, to demonstrate the performance of the surrogate approach, particularly in the presence of isolated subgraphs and stochastic communication delays. Additionally, the versatility of the proposed scheme is highlighted, as it is applicable to heterogeneous satellites and sensors with disconnected communication graphs.
- Open Access
Game-Theoretic Autonomous Driving: A Graphs of Convex Sets Approach
Multi-vehicle autonomous driving couples strategic interaction with hybrid (discrete-continuous) maneuver planning under shared safety constraints. We introduce IBR-GCS, an Iterative Best Response (IBR) planning approach based on the Graphs of Convex Sets (GCS) framework that models highway driving as a generalized noncooperative game. IBR-GCS integrates combinatorial maneuver reasoning, trajectory planning, and game-theoretic interaction within a unified framework. The key novelty is a vehicle-specific, strategy-dependent GCS construction. Specifically, at each best-response update, each vehicle builds its own graph conditioned on the current strategies of the other vehicles, with vertices representing lane-specific, time-varying, convex, collision-free regions and edges encoding dynamically feasible transitions. This yields a shortest-path problem in GCS for each best-response step, which admits an efficient convex relaxation that can be solved using convex optimization tools without exhaustive discrete tree search. We then apply an iterative best-response scheme in which vehicles update their trajectories sequentially and provide conditions under which the resulting inexact updates converge to an approximate generalized Nash equilibrium. Simulation results across multi-lane, multi-vehicle scenarios demonstrate that IBR-GCS produces safe trajectories and strategically consistent interactive behaviors.
Open AccessGame-Theoretic Autonomous Driving: A Graphs of Convex Sets Approach
Multi-vehicle autonomous driving couples strategic interaction with hybrid (discrete-continuous) maneuver planning under shared safety constraints. We introduce IBR-GCS, an Iterative Best Response (IBR) planning approach based on the Graphs of Convex Sets (GCS) framework that models highway driving as a generalized noncooperative game. IBR-GCS integrates combinatorial maneuver reasoning, trajectory planning, and game-theoretic interaction within a unified framework. The key novelty is a vehicle-specific, strategy-dependent GCS construction. Specifically, at each best-response update, each vehicle builds its own graph conditioned on the current strategies of the other vehicles, with vertices representing lane-specific, time-varying, convex, collision-free regions and edges encoding dynamically feasible transitions. This yields a shortest-path problem in GCS for each best-response step, which admits an efficient convex relaxation that can be solved using convex optimization tools without exhaustive discrete tree search. We then apply an iterative best-response scheme in which vehicles update their trajectories sequentially and provide conditions under which the resulting inexact updates converge to an approximate generalized Nash equilibrium. Simulation results across multi-lane, multi-vehicle scenarios demonstrate that IBR-GCS produces safe trajectories and strategically consistent interactive behaviors.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Control Systems Letters
- Vol. 10, 1645-1650
ADMM–EKI: Principled Refinement With Local Guarantees for Constrained Nonlinear MPC
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Control Systems Letters
- Vol. 10, 1645-1650
ADMM–EKI: Principled Refinement With Local Guarantees for Constrained Nonlinear MPC
- 5 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Control Systems Letters
- Vol. 9, 180-185
Distributed and Localized Covariance Control of Coupled Systems: A System Level Approach
This work is concerned with the finite-horizon optimal covariance steering of networked systems governed by discrete-time stochastic linear dynamics. In contrast to existing work that has only considered systems with dynamically decoupled agents, we consider a dynamically coupled system composed of interconnected subsystems. In particular, we propose a distributed algorithm to compute the localized covariance steering policy for each individual subsystem. To this end, we first formulate the localized covariance steering problem and, leveraging the system-level synthesis (SLS) framework, recast this problem as a convex optimization problem over system responses.We then examine the problems separability and introduce a problem transformation to address instances with nonseparable objective functions. Finally, we develop a distributed algorithm based on the consensus alternating direction method of multipliers (ADMM) to distribute computation across subsystems based on their local information and communication constraints. We demonstrate the effectiveness of our proposed algorithm on a power system with 36 interconnected subsystems.
- 5 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Control Systems Letters
- Vol. 9, 180-185
Distributed and Localized Covariance Control of Coupled Systems: A System Level Approach
This work is concerned with the finite-horizon optimal covariance steering of networked systems governed by discrete-time stochastic linear dynamics. In contrast to existing work that has only considered systems with dynamically decoupled agents, we consider a dynamically coupled system composed of interconnected subsystems. In particular, we propose a distributed algorithm to compute the localized covariance steering policy for each individual subsystem. To this end, we first formulate the localized covariance steering problem and, leveraging the system-level synthesis (SLS) framework, recast this problem as a convex optimization problem over system responses.We then examine the problems separability and introduce a problem transformation to address instances with nonseparable objective functions. Finally, we develop a distributed algorithm based on the consensus alternating direction method of multipliers (ADMM) to distribute computation across subsystems based on their local information and communication constraints. We demonstrate the effectiveness of our proposed algorithm on a power system with 36 interconnected subsystems.
- Top Cited PapersOpen Access
- 20 April 2025
- Published by Springer Nature in Autonomous Robots
- Vol. 49 (2) , 1-31
Deadlock-free, safe, and decentralized multi-robot navigation in social mini-games via discrete-time control barrier functions
We present an approach to ensure safe and deadlock-free navigation for decentralized multi-robot systems operating in constrained environments, including doorways and intersections. Although many solutions have been proposed that ensure safety and resolve deadlocks, optimally preventing deadlocks in a minimally invasive and decentralized fashion remains an open problem. We first formalize the objective as a non-cooperative, non-communicative, partially observable multi-robot navigation problem in constrained spaces with multiple conflicting agents, which we term as social mini-games. Formally, we solve a discrete-time optimal receding horizon control problem leveraging control barrier functions for safe long-horizon planning. Our approach to ensuring liveness rests on the insight that there exists barrier certificates that allow each robot to preemptively perturb their state in a minimally-invasive fashion onto liveness sets i.e. states where robots are deadlock-free. We evaluate our approach in simulation as well on physical robots using F1/10 robots, a Clearpath Jackal, as well as a Boston Dynamics Spot in a doorway, hallway, and corridor intersection scenario. Compared to both fully decentralized and centralized approaches with and without deadlock resolution capabilities, we demonstrate that our approach results in safer, more efficient, and smoother navigation, based on a comprehensive set of metrics including success rate, collision rate, stop time, change in velocity, path deviation, time-to-goal, and flow rate.
- 20 April 2025
- Published by Springer Nature in Autonomous Robots
- Vol. 49 (2) , 1-31
Top Cited PapersOpen AccessDeadlock-free, safe, and decentralized multi-robot navigation in social mini-games via discrete-time control barrier functions
We present an approach to ensure safe and deadlock-free navigation for decentralized multi-robot systems operating in constrained environments, including doorways and intersections. Although many solutions have been proposed that ensure safety and resolve deadlocks, optimally preventing deadlocks in a minimally invasive and decentralized fashion remains an open problem. We first formalize the objective as a non-cooperative, non-communicative, partially observable multi-robot navigation problem in constrained spaces with multiple conflicting agents, which we term as social mini-games. Formally, we solve a discrete-time optimal receding horizon control problem leveraging control barrier functions for safe long-horizon planning. Our approach to ensuring liveness rests on the insight that there exists barrier certificates that allow each robot to preemptively perturb their state in a minimally-invasive fashion onto liveness sets i.e. states where robots are deadlock-free. We evaluate our approach in simulation as well on physical robots using F1/10 robots, a Clearpath Jackal, as well as a Boston Dynamics Spot in a doorway, hallway, and corridor intersection scenario. Compared to both fully decentralized and centralized approaches with and without deadlock resolution capabilities, we demonstrate that our approach results in safer, more efficient, and smoother navigation, based on a comprehensive set of metrics including success rate, collision rate, stop time, change in velocity, path deviation, time-to-goal, and flow rate.
- Open Access
Transformer-Based Model Predictive Path Integral Control
This paper presents a novel approach to improve the Model Predictive Path Integral (MPPI) control by using a transformer to initialize the mean control sequence. Traditional MPPI methods often struggle with sample efficiency and computational costs due to suboptimal initial rollouts. We propose TransformerMPPI, which uses a transformer trained on historical control data to generate informed initial mean control sequences. TransformerMPPI combines the strengths of the attention mechanism in transformers and sampling-based control, leading to improved computational performance and sample efficiency. The ability of the transformer to capture long-horizon patterns in optimal control sequences allows TransformerMPPI to start from a more informed control sequence, reducing the number of samples required, and accelerating convergence to optimal control sequence. We evaluate our method on various control tasks, including avoidance of collisions in a 2D environment and autonomous racing in the presence of static and dynamic obstacles. Numerical simulations demonstrate that TransformerMPPI consistently outperforms traditional MPPI algorithms in terms of overall average cost, sample efficiency, and computational speed in the presence of static and dynamic obstacles.
Open AccessTransformer-Based Model Predictive Path Integral Control
This paper presents a novel approach to improve the Model Predictive Path Integral (MPPI) control by using a transformer to initialize the mean control sequence. Traditional MPPI methods often struggle with sample efficiency and computational costs due to suboptimal initial rollouts. We propose TransformerMPPI, which uses a transformer trained on historical control data to generate informed initial mean control sequences. TransformerMPPI combines the strengths of the attention mechanism in transformers and sampling-based control, leading to improved computational performance and sample efficiency. The ability of the transformer to capture long-horizon patterns in optimal control sequences allows TransformerMPPI to start from a more informed control sequence, reducing the number of samples required, and accelerating convergence to optimal control sequence. We evaluate our method on various control tasks, including avoidance of collisions in a 2D environment and autonomous racing in the presence of static and dynamic obstacles. Numerical simulations demonstrate that TransformerMPPI consistently outperforms traditional MPPI algorithms in terms of overall average cost, sample efficiency, and computational speed in the presence of static and dynamic obstacles.