I'm a Senior Machine Learning Engineer at the Autonomous Driving team at Qualcomm…

• Planning to Explore via Self-Supervised World Models

  ICML 2020 May 2020

  Reinforcement learning allows solving complex tasks, however, the learning tends to be task-specific and the sample efficiency remains a challenge. We present Plan2Explore, a self-supervised reinforcement learning agent that tackles both these challenges through a new approach to self-supervised exploration and fast adaptation to new tasks, which need not be known during exploration. During exploration, unlike prior methods which retrospectively compute the novelty of observations after the…

  Reinforcement learning allows solving complex tasks, however, the learning tends to be task-specific and the sample efficiency remains a challenge. We present Plan2Explore, a self-supervised reinforcement learning agent that tackles both these challenges through a new approach to self-supervised exploration and fast adaptation to new tasks, which need not be known during exploration. During exploration, unlike prior methods which retrospectively compute the novelty of observations after the agent has already reached them, our agent acts efficiently by leveraging planning to seek out expected future novelty. After exploration, the agent quickly adapts to multiple downstream tasks in a zero or a few-shot manner. We evaluate on challenging control tasks from high-dimensional image inputs. Without any training supervision or task-specific interaction, Plan2Explore outperforms prior self-supervised exploration methods, and in fact, almost matches the performances oracle which has access to rewards.

  See publication

• Perception-Driven Curiosity with Bayesian Surprise

  RSS 2019 - Workshop on Combining Learning and Reasoning May 2019

  We use variational inference to learn a dynamics model of image observations, and construct an agent that maximizes Bayesian surprise of the future frames. The Bayesian agent is more robust to stochastic environments than simpler prior prediction schemes.

  See publication

• Modified Extended Kalman Filter Using Correlations Between Measurement Parameters

  Advances in Intelligent Systems and Computing, Springer August 1, 2018

  We mathematically analyze the correlations that arise between measurement parameters. This is done by understanding the geometrical transformations that a data point undergoes when correlations are determined between normally distributed measurement parameters. We use this understanding to develop a new algorithm for the discrete Kalman Filter. The analysis and methodology adopted in this work can be extended to the derivatives of Kalman Filter, resulting in similar improvements. The…

  We mathematically analyze the correlations that arise between measurement parameters. This is done by understanding the geometrical transformations that a data point undergoes when correlations are determined between normally distributed measurement parameters. We use this understanding to develop a new algorithm for the discrete Kalman Filter. The analysis and methodology adopted in this work can be extended to the derivatives of Kalman Filter, resulting in similar improvements. The effectiveness of this method is verified through simulations of mobile robot mapping problem with an Extended Kalman Filter and the results are presented.

  Other authors

  See publication

• Use of measurement noise correlations for an improved SONAR model

  IEEE December 21, 2017

  Using SONAR as the primary range finding sensor has largely been abandoned due to problems such as limited range, large bearing errors and large beam widths. However, SONAR is used conjunction with other sensors such as LIDARs, RADARs and vision sensors for ranging and obstacle avoidance in many autonomous vehicle applications. In this paper, we propose a solution to reduce the range and bearing error significantly, and thus improve the performance of the SONAR. Using the results from the…

  Using SONAR as the primary range finding sensor has largely been abandoned due to problems such as limited range, large bearing errors and large beam widths. However, SONAR is used conjunction with other sensors such as LIDARs, RADARs and vision sensors for ranging and obstacle avoidance in many autonomous vehicle applications. In this paper, we propose a solution to reduce the range and bearing error significantly, and thus improve the performance of the SONAR. Using the results from the Gaussian Correlation Inequality, we derive probabilistic transformations that can improve the range and bearing measurement of the SONAR, thus reducing the sensor error. We are also presenting simulation study, to place bounds on the types and characteristics of the SONARs within which our model's performance is optimal.

  Other authors

  See publication

• Collaboration between unmanned aerial and ground vehicles for search and rescue missions

  Oct 2017 - Apr 2018

  Other creators

• Improving Sensor models with sensor-fusion

  Mar 2017 - Dec 2017

  Other creators

• Efficient mobile-robot localization/mapping with a new Kalman Filter algorithm

  Mar 2017 - Nov 2017

  Other creators

• Merit Scholarships

  SSN College of Engineering

  Nov 2017

• Project Funding

  SSN College of Engineering

  Oct 2017

  Obtained a research grant award of INR 30000 from the Student Internal Funding Scheme of SSN College of Engineering, for the Collaboration between unmanned aerial and ground vehicles project.

• Research Fellowship Award

  Indian Academy of Sciences

  May 2017

  Obtained a research fellowship award of INR 20000 and travel award from Indian Academy of Sciences for being a summer research fellow at Raman Research Institute.

• Best all-rounded student

  DAV Matriculation Higher Secondary School, Choolaimedu

  Jun 2014

• Scholarship for pursuing Bachelor of Science, University of Hong Kong (unavailed)

  University of Hong Kong

  Mar 2014

• School Pupil Leader

  DAV Matriculation Higher Secondary School, Choolaimedu

  Jun 2013

4 others named Ramanan Sekar are on LinkedIn

See others named Ramanan Sekar