About Blogs
Akash Sharma
I am an incoming Ph.D. student at Carnegie Mellon University, I graduated recently from CMU with a Masters in Degree in Robotics. As a member of Robot Perception Lab at the Robotics Institute, I am (was) advised by Prof. Michael Kaess. My research focuses on semantically assisted long term robotic mapping and 3D reconstruction through graph-based SLAM. More generally, I am interested in making robotic spatial understanding feasible.

Before joining CMU, I spent two years as a Software Developer at Infinera, where I developed infrastructure software for fault-tolerant optical communications, that manages long haul internet traffic. I received a Bachelor's degree in Engineering from Sri Jayachamarajendra College of Engineering (SJCE) in 2017, ranking amongst the top 10 students. As an undergraduate student, I have also taught and organized Robotics boot camps at SJCE for freshman students.
I'm interested in SLAM, and 3D Reconstruction in specific, and Computer Vision, Robotics and Graphics in general. My research aim is to incorporate learning in SLAM, and make SLAM robust to ambiguous posteriors.
Ongoing Projects
Compositional Scalable Object SLAM
Akash Sharma, Wei Dong, Michael Kaess
International Conference on Robotics and Automation (ICRA) 2021
Simultaneous localization and mapping of a 3D indoor environment as a posegraph of dense TSDF object volumes.
Depth Fusion for Large scale environments
Akash Sharma
Use spatial hashing techniques and submap registration methods to support large indoor scene reconstructions.
Previous Research and Projects
(Unofficial implementation) iNeRF: Inverting Neural Radiance Field for Pose estimation
Akash Sharma
Unofficial simplified implementation of the key concept in the IROS 2021 paper iNeRF
Visual SLAM for Quadrotors in Indoor environments
Akash Sharma, Shefali Vajramatti
Built a quadrotor that can map and localize itself in a static GPS denied environment.
Mobile Inverted Pendulum Robot
Akash Sharma, Adithya RH, Gururaj Kini
Built a MIP robot and implemented a PI-PD controller. The robot used stepper motors, and was controlled via an arduino. The robot also included a visual system, which allowed it to follow white lines, and detect faces.
Automated Vision Inspection System for Cylindrical Metallic Components
Krithika Govindaraj, B Vaidya, Akash Sharma, T Shreekanth
3rd International Conference on Computing and Communication (IC3 - 2018) Springer, paper
Developed a vision inspection system to sort faulty metal parts, in an industrial pipeline
Teaching and Service
Reviewer: International Conference on Robotics and Automation (ICRA) 2021
Fall 2021: Teaching assistant for 16822 - Geometry-based methods for Computer Vision
Fall 2020: Teaching assistant for 16833 - Robot Localization and Mapping (SLAM)