Aditya Jha

Apr '25

Jul '24

Oct '23

May '23

Feb '23

Sep' 22

Jun '22

Apr '22

Mar '21

Working on developing softwares for drones in GNSS-limited environments.

My work involved enhanching the navigation stack by introducing multi-map capabilites wherein I developed extensive GPS-VO policies for different real-world scenarios which increased our navigation uptime by around 30% in GPS-denied conditions. I also made our stack crash-averse by introducing map reuse feature wherein we restore the SLAM state before the crash and start a fresh session with the restored data.

Worked on developing GenAI-based solution to automate IT Helpdesk operations by leveraging state-of-the-art LLM models for ticket resolution.

Worked under the supervision of Prof. Ehsan Hashemi on developing visual-based human and indoor object detection capabilites for mobile robots.

My work involved data collection using the indoor Husky platform and developing a low latency visual-odometry object tracking module for precise localisation of the dynamic obstacles around the robot. [Report]

Worked under the supervision of Prof. Ravi Kiran and Prof. Harikumar Kandath on developing a software library that helps in estimating the key structural parameters for seismic risk assessment using visual remote sensing data.

My work involved proposing and implementing a novel method for estimating the distance between two adjacent building using the sparse 3D pointclouds. I also developed a large scale image stitching algorithm which was used downstream by the segmentation module to identify the non-structrucal elements on the rooftop and the area estimates related to it. This work was accepted at the CVCIE Workshop, ECCV 2022.

Worked under the supervision of Prof. Debashish Chakravarty as a Mechatronics and SLAM team member during my undergraduate at IIT Kharagpur.

I worked on developing algorithms for Stereo-Camera based localisation in pre-mapped LiDAR environments. I also participated in various international competitions like University Rover Challenge (URC) and F1TENTH Autonomous Grand Prix and also led junior members in various research projects and competitions.

We automate the inspection of buildings through UAV-based image data collection and a post-processing module to infer and quantify the details which helps in avoiding manual inspection, reducing the time and cost. We introduced a novel method to estimate the distance between adjacent buildings and structures. We developed an architecture that can be used to segment roof tops in case of both orthogonal and non-orthogonal view using a state-of-the-art semantic segmentation model. Taking into consideration the importance of civil inspection of buildings we introduced a software library that helps in estimating the key structural parameters.

Member of the IIT Kharagpur - IUPUI, Indiana - USB Colombia collaborative team.

Designed tightly/loosely coupled high-speed localisation in for racecar localisation in pre-mapped LiDAR circuit. The localisation was using 3 static-state LiDARs. [GitHub]

Developed a ROS-based sensor testing module as part of the Base Vehicle Software team. [GitHub]

We needed to achieve high speed navigation for virtual sim racing as a part of F1TENTH at ICRA 2022. Our team developed a LiDAR-based autonomous racing stack tuned to minimize the lap times and handle corners and obstacles smoothly.

We also developed a reactive obstacle avoidance algorithm and integrated it with PID controller to achieve 3rd out of the 40 participating teams.

Co-organised by IIT Kanpur and Drona Aviation, as part of the Inter IIT Tech Meet 11.0. This challenge aimed to develop a vision based state feedback control for an indoor multi-drone system handling socket communication with the flight controller without the use of ROS. Being a core team menber, I played a pivotal role in developing state estimation algorithms using depth camera feed and markers.

We used a adaptive PID controller which received the current pose estimate on basis of precise detection of Aruco Tag using our vision module. Our team secured the first place.

We attempt to explore how we can efficiently exploit stereocameras in premapped LiDAR environments to get the best of both worlds - a low-cost relocalization solution in an HD map.

Skills: Photometric residual minimization, heterogenous relocalization, Ceres Solver

As part of my Bachelor’s thesis, I developed planning and navigation algorithms for mobile robots operating in unknown environments. I designed a custom frontier-ranking method to prioritize exploration targets and integrated it with the ROS navigation stack, leveraging gmapping for online SLAM and move_base for path planning and control. This system enables a robot to autonomously explore unfamiliar spaces by continuously alternating between mapping, frontier selection, and goal-directed navigation.