RESEARCH

Research Experience

Project Scientist, The Robotics Institute, Carnegie Mellon University

1. Led the development of a novel, Bayesian hierarchical approach for constructing task duration distributions from past data, and demonstrated its effectiveness in constructing predictive probabilistic distribution models. Unlike traditional statistical learning techniques, the proposed approach relies on minimal data, is inherently adaptive to time varying task duration distribution, and provides a rich description of confidence for decision making. This framework is expected to address the issue of uncertainty in the duration of tasks faced by planning and scheduling algorithms.
2. Led the development of scientific computing framework for fast numerical approximation of high dimensional integral. This framework is extremely useful in Bayesian predictive models for time series data
3. Led the development of motion planning algorithms under highly uncertain environments for DARPA D1 Project. In specific, the proposed metric feedback algorithm optimizes the search of multi-agent location that minimizes the metric value in a non-convex function space.
4. Led the development of an online decentralized auction-based scheduler capable of providing a scalable approach to fleet scheduling while preserving the ability to optimize under real-world constraints. The performance of the proposed algorithm is evaluated on both a set of feasibility benchmark problems from the literature and a large-scale real-world paratransit dataset. The results show that auction-based scheduler significantly improved (~50%) on the results of the current state-of-the-art algorithm
5. Co-led the detailed technology development roadmap for unmanned unwater vehicles (UUVs).
6. Led the development of distributed multi-agent scheduler for automatic test operations in Anechoic chamber
7. Led the development of real-time statistical scheduler for Army Corps of Engineers new mat-sinking unit. In specific, the main objective of the Arm scheduler is to command the available lifting arms (up to 6 lifting arms) via Arm Motion Supervisor (AMS) to minimize the build time of each launch in such a way that it adheres to safety constraints. In the process, the Arm scheduler gathers information on launch configuration, number of available arms for scheduling etc. from the Mat scheduler, obtains information from the database on approximate stack locations, and builds an adaptive look-ahead schedule for the placement of mat squares while incorporating the statistical uncertainties in the system.
8. Led the development of statistical simulator for exploring warehouse automation configurations. In specific, I led the development of statistical simulator to test various AS/RS configurations of the warehouse.
9. Co-led the development of a novel perception framework that has the ability to identify and track objects in autonomous vehicle’s field of view. The proposed algorithms do not require any training for achieving this goal. The framework makes use of ego-vehicle’s pose estimation and a KD-Tree-based segmentation algorithm to generate object clusters. In turn, using a VFH technique, the geometry of each identified object cluster is translated into a multi-modal PDF, and a motion model is initiated with every new object cluster for the purpose of robust spatio-temporal tracking. The methodology further uses statistical properties of high-dimensional probability density functions and Bayesian motion model estimates to identify and track objects from frame to frame. The effectiveness of the methodology was tested on a KITTI dataset. The results effectively showed that the median tracking accuracy is around 91% with an end-to-end computational time of 153 milliseconds.
10. Co-led the development of a distributed multi-agent sensor fusion techniques to statistically reason about identifying and correcting for false negatives in the autonomous navigation. Unlike existing autonomous navigation (AV) systems that require accurate sensor feeds for any meaningful inference, in this proposed framework, each autonomous agent simultaneously localizes and maps its local environment. This map, in turn, is encoded into a low-resolution message and shared with nearby agents. Next, the agents distributively fuse this information together to construct a world interpretation. Each agent then statistically analyzes its own interpretation with respect to the world interpretation for the common regions of interest. The proposed statistical algorithm outputs a measure of similarity between local and world interpretations and identifies false negatives (if any) for the local agent. This measure, in turn, can be used to inform the agents to update their kinematic behavior in order to account for any errors in local interpretation
11. Co-led the development of a concept called Simulated – BSM that leverages the combination of rich vehicular sensor feed (speed, object proximity, SLAM maps etc.), and emerging guaranteed low latency communication protocols (DSRC, 5G) to enhance efficiency & safety of transport facilities. The main idea is that an ego-vehicle with local sensor feed capability creates a basic safety message for any identified rogue vehicles in its field of view, and broadcasts it to infrastructure for implementing efficient and safe control actions. Collaborated with a colleague in the development new system architecture for this proposed system, and software infrastructure development to realistically simulate these ideas. The efficacy of S-BSM framework was successfully demonstrated in the context of enhancing safe & efficient operations at rural high-speed signalized intersection.
12. Lead investigator for equipping and integrating port authority buses into connected vehicle test-bed. I’m also leading the efforts of developing robust statistical models for accurate bus dwell time prediction
13. Led the development of end-to-end technological solution for automatic identification & classification of cruising for parking behavior
14. Lead investigator in developing sensor sharing models for connected and autonomous cars

Special Faculty, The Robotics Institute, Carnegie Mellon University

1. The development of next generation technologies for Intelligent Transportation Systems, including but not limited to the application of distributed coordination, market-based mechanisms and game-theoretic concepts to the design of real-time adaptive traffic signal control systems, the integration of such traffic signal capabilities with emerging connected vehicle technologies, the development of sensing techniques to enable self-monitoring of road network performance, and the development of techniques for multi-modal traffic management.
2. Led the software integration project to integrate Surtrac system (adaptive signal control system developed at CMU) with commercial traffic simulator. This software package is designed for running large scale simulations in a decentralized manner across multiple networked computers; each computer serves as a simulation client that can handle multiple local processes. Even though, the primary goal has been to simulate the behavior of Surtrac system, the modular architecture of the software makes it possible to simulate other adaptive control systems. This unique feature of the software makes it a good decision-analysis tool to conduct comparative performance analysis of different adaptive control strategies for a given urban network. As a result, we were able to successfully simulate the viability of Surtrac adaptive control system in downtown Pittsburgh. This model has been successfully demonstrated to City personnel and has led to increased discussion of the possibility of doing a pilot implementation of Surtrac in the downtown area
3. Developed and demonstrated a scheme for evaluating traffic signal network performance via ubiquitous blue-tooth sensing.

Graduate Researcher, North Carolina State University

1. Doctoral Dissertation: My dissertation treats signal control problems as a mathematical game. That is, instead of using the standard techniques of minimum greens, maximum greens, and gaps to control the signal indications, an economically based game structure is employed. The intersection’s space is viewed as a scarce commodity whose use is determined through a bidding process. 
2. Establishing and Monitoring Travel Time Reliability (SHRP-2, LO2): Worked on travel time reliability project, and here are some of my contributions to the project. 
• Developed a filtering algorithm that separates ‘travel times’ from ‘overall trip times’ in AVI data
• Developed a methodology to synthesize route-level travel time distributions from segment level observations
• Developed a planning-based reliability tools to determine operative regimes for transit-based trips
• Developed “Use Case Analysis” document that demonstrates how the travel time reliability monitoring system can be used to address questions about network reliability
3. Electronic Toll & Traffic Management Project: Developed an analysis procedure for estimating travel time from electronic toll tag reader data. The methodology presents analyses of arterial travel times based on AVI data. The data was produced by a six month experiment conducted on a small arterial network in upstate New York. Data was collected using wireless, solar powered toll tag readers. The analysis fundamentally addresses the question of what can and cannot be learned from such data. It explores the ability of such data to elucidate trends in travel times across given days, from one day to another, by day of the week, and as affected by weather and other phenomena. It presents a detailed analysis of identified trends, and points to ways in which the data can be used for incident detection, travel time reliability monitoring, travel time prediction, and overall performance monitoring. 
4. Placement of Detection Loops on High-Speed Approaches to Traffic Signals: Worked with Dr. George List on a project entitled “Placement of Detection Loops on High-Speed Approaches to Traffic Signals”, sponsored by NC Department of Transportation (NCDOT) . The objective of this research was to minimize the number of vehicles caught in a dilemma zone at the onset of yellow, without compromising the efficiency of a side street. We tested three different signal control strategies, both in the simulation and in the field, and concluded that Detection Control Strategy (DC-S) is the best control strategy for dilemma zone protection. Based on our research, the NCDOT decided to incorporate D-CS logic in OASIS software.