This website is designed as a showcase for text-based portfolio items, using the WordPress blog structure to present each item in a separate, dated blogpost. The Recent Works page is an index to the posts, displaying brief excerpts in reverse chronological order. The homepage presents a mini-index, to just the most recent four posts.
Instructed reinforcement learning control of safe autonomous j-turn vehicle maneuvers
This paper presents a safe control policy search for autonomous J-turn maneuvers inspired by professional car drivers. These drivers have been highly trained to achieve highly aggressive J-turn maneuvers that can be used as potential vehicle safety strategies in emergency situations. We take advantage of this knowledge by integrating an expert’s instruction into reinforcement learning design. The proposed safe reinforcement learning method uses constrained Markov decision processing to find a risk-free control policy for autonomous J-turn maneuver. The proposed algorithms realize the control of continuous action as a sequence of policies with safety constraints. We demonstrate experimentally the development of policy design for the safe J-turn maneuver from simulations to real-world applications. Paper Link
Safe predictive control of four-wheel mobile robot with independent steering and drive
We propose a predictive controller that minimizes the trajectory error of a four-wheel mobile robot with independent steering and drive (4ISD) under safety constraints imposed by the presence of people. The 4ISD drivetrain is overactuated, that is, a given motion vector maps to multiple sets of actuation, not all of which are safe or optimal. We use convex optimization to satisfy Control Barrier Functions (CBF), which express the safety constraints for a robot operating in the vicinity of humans. The result is a Nonlinear Model Predictive Controller (NMPC) that produces a permissible actuation set. We develop mathematical foundations and run simulations that demonstrate the controller performance and safety around humans in the face of disturbances and uncertainties. Link to Article.
Optimal control of wheeled mobile robots: From simulation to real world
We study the problem of taking simulations to the real world (RW) for autonomous robotic systems with dynamic uncertainties and unknown disturbances while maintaining the optimal performance and stability of the designed controller designed in simulation. In general, an optimal and robust controller that is designed through simulation often does not perform similarly when deployed in the RW. We focus on using simulations to generate an optimal control policy utilizing the Memetic algorithm (MA) iteratively. The simulation-to-RW performance and stability are realized by using an adaptive fuzzy system to learn the uncertain part of the dynamic model, disturbance and noises. We demonstrate experimentally that this method permits the development of optimal control design in simulations and integrates adaptive learning rules to enable precise and repetitive trajectory tracking for the wheeled mobile robot (WMR) with disturbances and uncertainties. Link to Article
Safe and Secure Solutions for Connected Cyber-Physical Systems
Agile Safe Autonomous Systems (ASAS) was awarded to develop the leadership skills of the team using the guidance and support throughout the National Science Foundation (NSF) Innovation Corps program hosted by NE I-Corps .. at Massachusetts Institute of Technology. The mentorship and experience have been instrumental in shaping our team’s entrepreneurship and intrapreneurship skills, and our team is incredibly grateful for the opportunity to work with someone as experienced and passionate to help as you.
During the seven-week program, our team, Agile Safe Autonomy had the privilege of developing a cutting-edge business model for a technology, which is focused on security and safety for future connected autonomous systems. Through your support and encouragement, we were able to interview over 200 experts in the field of robotic, autonomous vehicles and automotive which has been instrumental in shaping our approach to emerging technologies.
Safe Iterative Feedback Linearization Control Design for Collaborative Robots
Intelligent autonomous robotic systems are not limited to industrial applications and are being used for professional service and household domains. The vast amount of these robots are designed to work alongside humans in close proximity which pushes the need for safety-guaranteed control systems. Assuring the safety of humans and the environment while maintaining the performance of the robot is still a challenging problem. We proposed a safe iterative feedback linearization control method for the safe operation of collaborative robots. This proposed approach combines feedback linearization with intelligent predictive algorithms to achieve stable and safe control of the robot’s motion. By using real-time feedback and monitoring user interactions, I demonstrate how to improve performance over time while maintaining safety. Our proposed method shows how to safely improve performance in collaborative robotics applications. Experimental results show our designed control system is a promising approach for achieving safe and efficient control of collaborative robots. Link to file