GTA for Graduate-Level Deep Learning course (CS7643)

Advisor: Prof. Zsolt KiraWorking on open-world, embodied navigation with a focus on real-world deployment.

Research Advisor: Dr. Ali-akbar Agha-mohammadi, Prof. James M RehgWorking on enhancing long-range perception capabilities for the DARPA RACER project:• Fusing information from large-scale, coarse Digital Elevation Models (DEMs) into the short-range 3D LiDAR-based elevation mapping pipeline• Performing densification of 3D LiDAR pointcloud at long range using high-resolution RGB cameras

Research Advisor: Prof. Frank Dellaert• Developing a comprehensive 4D (space-time) situational awareness pipeline based on a novel combination of well established 3D reconstruction methods.• Structure from Motion (SfM) will be performed using multiple video streams from cameras mounted on thruster-driven robots to provide real-time assistance to astronauts during spacewalks.

• Co-founded the autonomous ground vehicle team of DTU - DTUSDC - which participates in one of the world’s most competitive autonomous mobility competitions annually in Michigan, USA – the IGVC.• Successfully raised and managed USD 20000 for smooth execution of the project - acquiring sponsorships from reputable tech giants like Huawei, Nvidia, SICK and Sparton NavEx.• In only it's second attempt, DTUSDC ranked 5th at IGVC 2019 among 30+ teams from all over the world; atleast 50% of the participating teams had been attending the IGVC consecutively for more than 10 years then• Supervised sanctioning of funds (USD 60000) from the university’s administration for the development of DTUSDC’s state-of-the-art robotics laboratory.• Coordinated day-to-day proceedings and inter-departmental communication for efficient system integration.

• Designed a modular and lightweight spaceframe chassis to achieve efficient packaging, reduced assembly downtime and enhanced transportability for the IGVC 2018/19.CAD development and subsequent stress simulations were performed using SolidWorks and ANSYS respectively.• Planned and executed day-to-day manufacturing processes and assembly routines to achieve timely delivery of the vehicle for subsequent endurance testing.• Executed hardware-in-the-loop (HIL) testing procedures, including systematic tuning of various low-level PID controllers and testing of EKF-based localisation techniques using the ROS framework.

• Worked on 2D transient CFD simulations to optimize hyperloop pod design for the SpaceX HyperLoop Pod competition• Analysed various optimisation routines based on genetic algorithms to enhance pod design; in addition, various geometry paramterisation techniques such as CSTs and Bezier curves were examined

• Responsible for development of localisation and navigation modules to achieve autonomous maneuverability of the hyperloop pod using a plethora of onboard sensors• Managing implementation of various autonomous mobility algorithms using the ROS framework• Development of telemetry and data acquisition modules

• Evaluated and summarized the working of various mechanical equipment in the methanol production plant at Salalah Methanol Company.• Analysed third-party FEA (ANSYS) and HTRI technical reports to resolve longstanding issues in cold make gas-BFW heat exchanger (E1203) operation.• Engineered methodologies to enable on-site troubleshooting of the E1203 heat exchanger.• Examined the working of steam reformer and suggested possible solutions to increasing accessibility to combustion air guide vanes for quicker maintenance procedures. Understanding the steam reformer's functioning was also crucial to diagnosing the E1203 heat exchanger's irregular operation.• Executed corrective and predictive on-site maintenance activities on a day-to-day basis at the methanol production plant.

• Headed a team of 8 members responsible for the development, manufacturing and testing of the aerodynamic sub systems of the vehicle for Formula Bharat 2018/19.• Streamlined the in-house carbon fiber composite manufacturing routines, reducing costs by 20%. This was primarily done by developing innovative, cost-effective solutions to minimize the cost of expensive, intricate 3D molds required in the manufacturing process.• Focused largely on aerodynamic optimization of the front wing (using notched end plates) and diffuser (strakes and added infuser) to achieve large increases in downforce with minimal drag penalties.

• Designed and fabricated a cost-effective, in-house data acquisition system to validate mechanical design of the vehicle's sub systems accurately by logging real-time data from over 20 sensors onboard the vehicle during track testing.• Conceptualized and implemented various mechatronic control systems to enhance on-track performance of the vehicle such as:- aerodynamic drag reduction system- pneumatic gear shifting system- interactive steering wheel• Managed the electrical wiring and routing of the engine as well as associated onboard sensors.

• Designed and developed various aerodynamic sub-systems of the formula-style vehicle such as side pods, front wings and nose cone using high-fidelity CFD simulations run on Star-CCM+. CAD models were designed using SolidWorks.• Carried out in-house composite manufacturing of entire aerodynamic package (using carbon fiber) of the vehicle using techniques like vacuum bagging and resin infusion.• Performed preliminary on-track testing to validate computational simulations and efficacy of in-house composite manufacturing techniques.