Vision-based Goal Conditioned Reinforcement Learning

I worked on a project titled "Achieving Compliant Motion Control for Continuous-Contact Tasks with Reinforcement Learning (RL)" in Prof. Sebastian Trimpe's group under the supervision of MSc. Emma Cramer. We investigated if robot controllers based on reinforcement learning (RL) could outperform classical impedance controllers in a suite of surface polishing tasks.

• Machine learning on Time-series data across Lithium-ion cell production processes for quality prediction• Discrete event-based and continuous-time simulations• Dashboards for virtual production systems

The project entailed research on improving data efficiency of model-free Deep Reinforcement Learning algorithms in Prof. Sebastian Trimpe's group with MSc. Bernd Fraunknecht as my supervisor. We investigated different critic ensemble methodologies for robust Q estimation to achieve update-to-data ratios comparable to the SOTA model-based algorithms.

• Design and assembly of Autonomous Guided Vehicle (Ranger GTP M, XL)• Modelling and development of suspension and transmission systems• Integration of mechanical, electrical and control systems• Value engineering and localization• Product development lifecycle • BOM & Engineering change management

• Development of Pneumatic Actuation for Sortation System• Manufacturing processes: press operations, CNC machining, welding, 3d printing, vacuum forming

Bullethawk Racing is the official Formula Student team of the University of Delhi with a strength of 35 membersResponsibilites:Managed a budget of 8,00,000 INR accumulated over the span of 1 year. Set judicious project goals and did resource planning to produce a race ready car in a strict time bound setting. Ensured that the team functioned smoothly and its stake holders- university, tech workshops, sponsors were content with our operation in order to hand over a healthy project to the future team.Bagged 30th position in FS India 2016 among 90 participating teams. Incorporating feedbacks from judges on the same build and working on competition strategy, we placed 10th overall in SUPRA SAE India 2016 among 180 teams.

Led research, design and fabrication of suspension, chassis and brakes systems for team's 2016 build. Designing and analysis was done with Solidworks and MATLAB. Actively participated in fabrication of chassis and other crucial components whose inhouse fabrication was possible; Gained ample hands-on workshop experience.Vehicle level integration, addressing conflicting demands of ergonomics, packaging and performance. Implemented a top down design approach and supervised the over all assembly of the car. Success at this level could be assessed in terms of improved over all handling and drivability and track times of the car from its previous iteration.

Worked in the plastics operation department, studied the bumper painting and inspection process.Maintained and compiled data sheets for closing stock, issuance and consumption of the paint materials - Incorporated the data in analysis and planning of the usage of available material from current to subsequent plans.