**under development (both me and the website)

Ajit Srikanth

I am an undergraduate researcher at the Robotics Research Center (RRC), IIIT-Hyderabad, working on robotic manipulation. I am supervised by Prof K. Madhava Krishna and collaborate closely with Prof Arun Singh. I spent my summer of 2025 working with Prof David Held at Carnegie Mellon University.

Email  /  GitHub  /  Google Scholar  /  LinkedIn  /  CV

profile photo

More

I am interested in working on robot learning for manipulation (or loco-manipulation). My long-term research goal is to work towards creating robust generalist loco-manipulators that can safely work with or assist humans with any tasks. I wish to create agents which can learn from their (or some other embodiment's) prior experiences to perform tasks, while also being able to adapt to new tasks (and environments) maybe with just a few demonstrations.

Feel free to reach out to me to for any discussions, collaborations or just to say hi. Most of my social media handles are just "ajitsrikanth" "(.*?)" . I'm always down for an intresting conversation. If you dont know what to talk about, we can talk about Robotics in general, Robot Learning, Reinforcement Learning, Motion Planning, Foundation Models, Computer Vision, Deep Learning / ML, Computer architecture, Hardware acceleration, any cool math theorms / neat proofs, physics (feel free to go wild with your theories), tennis, cricket, most other sports, movies (any genre works), anime, world domination(?), or literally anything.


Publications

(* denotes equal contribution. Click on publication to view more)

project image

GPD: Guided Polynomial Diffusion for motion planning


Ajit Srikanth*, Parth Mahajan*, Kallol Saha*, Vishal Mandadi*, Pranjal Paul, Pawan Wadhwani, Brojeshwar Bhowmick, Arun Singh, and Madhava Krishna
CASE, 2025
project image

EDMP: Ensemble-of-costs-guided Diffusion for Motion Planning


Kallol Saha*, Vishal Mandadi*, Jayaram Reddy*, Ajit Srikanth, Aditya Agarwal, Bipasha Sen, Arun Singh, Madhava Krishna
ICRA, 2024

Projects

(Click on project to view more)

project image

Articubot extension


project
  • Worked on extending a zero-shot sim-to-real policy for articulated object manipulation (Articubot) by online history conditioning.
  • Solved a critical real-world kinematic limitations by implementing a whole-body control maneuver, synchronizing the mobile base and XArm7’s movements, to overcome the robot’s IK limits.
  • Improved real-world control by integrating compliant impedance control for safe interaction and optimizing the action execution pipeline (via parallel inference, selective action dropping, and other optimizations) to greatly speedup the real-world execution.
project image

Sound Source Localization on videos


project
  • Developed a system that combines semantic and temporal information to localize sound sources in videos
  • Audio-visual grounding by aligning a large Vison-Language Model’s embeddings with audio embeddings (similar to ”Can CLIP Help Sound Source Localization?”)
  • Addresses video challenges, including unidentified frames and instance disambiguation through our custom temporal module that utilizes a audio-motion transformer
project image

ASAC- Assisted Spatial Audio Construction


project
  • Automatically generating spatial audio (like Dolby Atmos) for movie scenes using video understanding.
  • Developed a system for complex multi-speaker environments with identity switching, and occlusions, using lightweight tracking, facial recognition and Active Speaker Detection techniques.
  • Currently working on improving spatial audio generation in scenes lacking proper visual cues through advanced scene understanding.
project image

Contract-NLI


project
  • Proposed a two-step BERT-based architecture, improving NLI and evidence inference for legal contracts.
  • Improved handling of long-term dependencies in evidence-spans, and context size limits , due to our modified architecture and ensemble approach, achieving higher accuracy than the baseline ContractNLI-BERT.
project image

Autonomous Omni-drive robot


project
  • Assembled a mecanum wheel robot with four Planetary Gear DC Motor driven by two Cytron Mdds30.
  • Integrated and calibrated, a Realsense T265 tracking camera and a Realsense D455 depth camera, along with voxfield for obtaining ESDFs utilized for motion planning.
  • Implemented several planning algorithms, including CEM, MPPI, and MPC, on an onboard Intel NUC, utilizing ESDFs for collision costs; designed and implemented a PID controller for path-following.
project image

8 legged Klann walker robot


project
  • Designed a walker robot with 8 klann linkage mechanims for legs, performing comprehensive kinematics and dynamics analysis with consideration for different gaits for different purposes, on Fusion360.
  • Built the mechanism using laser-cut acrylic for links and chassis, with 3D-printed joints and couplers.
  • Wrote a simple controller to control the two dynamixel XL330 motors using U2D2 for communication.
project image

Object centric reconstruction


project
  • Designed a complete pipeline for 3D reconstruction of an object using a monocular fixed camera.
  • Optimized the initial 3D reconstruction with ICP, pose graph optimization, and Bundle Adjustment for efficient and accurate results; Proposed future enhancements by utilizing NeRFs to learn appearance geometry.
project image

Single image dehazing methods


project
  • Implementation of various haze removal methods including Bayesian defogging, Single image haze removal using Dark Channel Prior, Single image haze removal using Colour Attenuation Prior.
  • Improved upon the DCP prior paper, by adding an additional preprocessing step which increases the restored image quality by utilizing CLAHE; Demonstrated their use in stream tasks like object detection.
  • Extensively benchmarked these algorithms along with other classical methods without a degradation model, and analyised their performance.
project image

Y86 ISA Processor


project
  • Designing and implementing sequential and pipelined Y86 ISA Processor in iverilog.
  • proper handling of data, structural, and control hazards.
  • A basic compiler using python for converting Y86-64 assembly code to machine code for that processor.

Teaching

  • I served as a teaching assistant for the course EC4.403 Robotics Planning and Navigation (Spring 2025) at IIIT-Hyderabad by Prof. Madhava Krishna, where I prepared and graded assignments, conducted tutorials, and taught a few lectures on topics including: MPC/MPPI/CEM, C-space planning, Time Parametrized Trajectory Generation, Collision Cone and Reactive Avoidance. In addition, I mentored students on their final projects, which involved MPC/MPPI/CEM-based control, multi-modal perception, motion planning, and reactive collision avoidance, across three embodiments: drones, manipulators, and mobile robots
  • I taught two lectures at for the course CS7.503 Mobile Robotics (Monsoon 2024) at IIIT-Hyderabad by Prof. K. Madhava Krishna, where I covered the basics of Transformations and Quaternions representations, and then ICP slam.
  • I served as a teaching assistant for the course EC2.204 Intro to Processor Architecture (Spring 2024) at IIIT-Hyderabad by Prof. Deepak Gangadharan, with around 100 students, where I prepared assignments, and conducted tutorials on topics including: Verilog HDL programming, X86 ISA, Sequential and pipelined architectures and hazard handling. Further, I helped design and mentor the students for the final project, which was to design a Y86 ISA pipelined processor, with proper branch prediction and hazard handling.


Design and source code from Jon Barron's website