Secure and Efficient Autonomous Systems Lab
The Secure and Efficient Autonomous Systems Lab focuses on fundamental research spanning: 1) multi-agent motion coordination; 2) security of autonomous systems and 3) scalable data-driven methods for control.
For a preview of some of the research in the lab, check out our lab's YouTube channel: https://www.youtube.com/channel/UC3Tnk9e0d3wkPWwnXQGV9_Q
Present Group Members:
Graduate Students:
1) Christopher Calle (PhD student, Topic area: Randomized methods for estimation)
2) Richard Frost (PhD student, Topic area: Perimeter guarding problems)
3) Benjamin Toaz (BS-MS Dual Degree student, Topic area: Multi-agent motion planning)
Undergraduate Students:
1) Balaji Ganeshbabu (MSU)
Lab Alumni
Graduate Students:
2) Sandeep Banik, PhD December 2023, Dissertation title: Adversarial Modeling in Game-theoretic Networks for Securing Cyber-physical Systems. Current position: Postdoctoral Associate, University of Illinois Urbana Champaign
1) Shivam Bajaj, PhD August 2023, Dissertation title: Online pursuit algorithms and optimal strategies for heterogeneous robots. Current position: Postdoctoral Associate, Purdue University
Postdoctoral Associates:
1) Dr. Bhargav Jha (Topic area: Pursuit problems and games with asymmetric information). Current position: Assistant Professor, Electrical Engineering Department, Indian Institute of Technology Kharagpur.
Undergraduates:
1) Kunj Dedhia (Visiting undergraduate researcher EnSURE 2019), now at Microsoft
2) Andrew Debaker (Visiting undergraduate researcher EnSURE 2021)
3) Ishwari Kapale (Visiting undergraduate researcher EnSURE 2022)
4) Jayden Devaull (Visiting undergraduate researcher EnSURE 2022)
5) Krishna Das Artis-Mickens (Visiting undergraduate researcher SROP 2023)
6) Ebenezer Adjah (MSU)
The following are some recent research projects that the lab has been working on. For a full list of publications, click here.
Competitive Analysis of Perimeter Guarding problems
In these works, we design and analyze novel online algorithms for autonomous vehicles for perimeter defense applications in planar environments. Numerous (possibly infinite) intruders arrive in an online manner to breach the perimeter and the goal of the vehicles is to capture the intruders before they breach. We do not impose any assumption on the arrival process of the intruders and leverage tools from online optimization to analyze our algorithms in the worst-case scenarios. The algorithms designed in these works are compared with the fundamental limits to the problem under certain parameter regimes. Recent works include multiple homogeneous and heterogeneous vehicles. Recently, we addressed an optimal control problem to determine a shortest path for a Dubins vehicle with an attached controllable laser to capture a static target.
Related publications
Shivam Bajaj, Shaunak D. Bopardikar, Alexander Von Moll, Eric Torng, and David W. Casbeer. Competitive perimeter defense with a turret and a mobile vehicle. Frontiers in Control Engineering 4 (2023): 1128597.
Shivam Bajaj, Eric Torng, Shaunak D. Bopardikar, Alexander Von Moll, Isaac Weintraub, Eloy Garcia, and David W. Casbeer. Competitive perimeter defense of conical environments. In 2022 IEEE 61st Conference on Decision and Control (CDC), pp. 6586-6593. IEEE, 2022.
Control Design using Iterative Multi-Fidelity Methods
These works examine how a low-quality approximation of a system may be used to improve a true system’s control design. For example, consider a robot for which we possess an approximate mathematical model. We aim to select the control parameters that optimize the robot’s performance. Iteratively testing these parameters on the robot would be time-consuming. Meanwhile, the mathematical approximation may be used to quickly simulate the system. However, the simulation may fail to account for differences in the true robot’s dynamics. In these works, we examine how these two sources of information may be used to more quickly arrive at an optimal control choice for the robotic system.
Related publications
1. Ethan Lau, Vaibhav Srivastava, Shaunak D. Bopardikar, "A Multi-Fidelity Bayesian Approach to Safe Controller Design", IEEE Control Systems Letters, vol.7, pp.2904-2909, 2023.
2. Ethan Lau, Vaibhav Srivastava, Shaunak D. Bopardikar, "On Multi-Fidelity Impedance Tuning for Human-Robot Cooperative Manipulation”, arXiv preprint arXiv:2310.05904, 2023.
Autonomous Path Planning under Localization Uncertainty
In these works, we revisit the classic path planning problem from a recently evolved perspective of jointly minimizing uncertainty. Our approach leverages new analysis on the evolution of the maximum eigenvalue of the covariance matrix. This approach yields near-minimum uncertainty paths in the presence of stochastically modeled sensor misdetections. This work is motivated by environmental scenarios in which ambient lighting conditions may lead to incorrect assumptions on the accuracy of the sensors. Our framework further generalizes to a multi-objective setting to yield the shortest path under localization constraints posed on the vehicle. Although the classical versions of this problem involve efficient graph search techniques, the presence of uncertainty makes these scenarios inherently complex mainly due to the sizes of the underlying graphs and the history dependence in the quantification of the uncertainty. Our new analytic treatment of the evolution of the maximum eigenvalue of the covariance matrix enables us to mitigate the history dependence to provide a sub-optimal solution in an efficient manner.
Related publications
J3. S. D. Bopardikar, B. Englot, and A. Speranzon. Multi-Objective Path Planning: Localization Constraints and Collision Probability. IEEE Transactions on Robotics, 2015.
J2. S. D. Bopardikar, B. Englot, A. Speranzon, and J. van den Berg. Robust Belief Space Planning with Intermittent Via A Maximum Eigenvalue-Based Metric, International Journal of Robotics Research. Note: To appear as of Apr. 2016..
C1. S. D. Bopardikar, B. Englot, and A. Speranzon. Multi-objective Path Planning in GPS-denied Environments under Localization Constraints. In American Control Conference, Portland, OR, USA, 2014.
Randomized approaches to Sensor Selection and Placement
In these works, we derive strong guarantees to the problem of sensor selection for the state estimation of stochastic systems. When a dynamical system is only partially observed, sensors are required to estimate the internal state of it. We seek to find a collection of good sensors that can conduct state estimation. By sampling the sensors (that we eventually deploy) in a random manner and leveraging tools, such as matrix concentration inequalities, we are able to guarantee certain properties about our state estimate. We also design an efficient algorithm for finding a sampling policy that can draw a relatively good selection of sensors. Our approach is compared against other competing approaches, such as the greedy algorithm and a convex relaxation approach. We extend our guarantees to other scenarios to address practical limitations, such as computational resources.
Related publications
J2. Christopher I. Calle, and Shaunak D. Bopardikar, "A Concentration-Based Approach for Optimizing the Estimation Performance in Stochastic Sensor Selection," in IEEE Transactions on Control of Network Systems, doi: 10.1109/TCNS.2023.3298322.
C1. Christopher I. Calle, and Shaunak D. Bopardikar. "Probabilistic performance bounds for randomized sensor selection in Kalman filtering." 2021 American Control Conference (ACC), 2021.
J1. S. D. Bopardikar, "A randomized approach to sensor placement with observability assurance", Automatica, article 123, 2021.
Randomized Methods for Large Dimensional Optimization problems
In these works, we applied randomized techniques to solve classic optimization problems such as low-rank matrix factorization especially when the data, i.e., rows and columns of the matrix, are generated sequentially in time; and matrix games, specifically two-player, zero-sum games of large-sizes. Randomized methods are leveraged to synthesize novel solutions, as well as provide rigorous analytic guarantees on the quality of the solution in probabilistic sense. For games, the procedure is applied to efficiently solve a hide-and-seek game which is known to exhibit exponential complexity. Recently, we extended the methodology and the results to dynamic games.
Related (recent) publications
C3. S. D. Bopardikar, S. S. Nair, and R. Rai. Sequential Randomized Matrix Factorization. American Control Conference, Washington DC, USA, Jun. 2013.
J2. S. D. Bopardikar, and C. Langbort. On Incremental Approximate Saddle-point Computation in Zero-sum Matrix Games. In Automatica, vol. 69, July 2016.
J1. S. D. Bopardikar, A. Borri, J. P. Hespanha, M. Prandini, and M. D. Di Benedetto. Randomized Sampling for Large Zero-Sum Games. Automatica, 2013.
Pursuit-Evasion under Physical Constraints
In this work, we consider motion planning scenarios in which one or many agents seek to achieve a certain task under various physical constraints such as sensing, motion and energy constraints. Specific examples of such a task considered in my works is pursuit of a mobile agent (evader) by a group of mobile pursuers wherein constraints such as limited range sensing and non-holonomic motion were considered, and more recently, physical constraints such as multiple pursuers required to complete capture, have been considered. In these works, we design novel motion plans that provide guarantees on successful completion of the task.
Related (recent) publications
J3. S. D. Bopardikar, and S. Suri. k-Capture in Multiagent Pursuit-Evasion or, the Lion and the Hyenas. In Theoretical Computer Science, 2014.
J2. S. D. Bopardikar, F. Bullo, and J. P. Hespanha. A cooperative Homicidal Chauffeur game. Automatica, 45(7): 1771-1777, 2009.
J1. S. D. Bopardikar, F. Bullo, and J. P. Hespanha. On Discrete-Time Pursuit-evasion Games with Sensing Limitations. IEEE Transactions on Robotics, 24(6): 1429-1439, 2008.