Safe Control/Planning for Autonomous Ships
Fall 2025 - PresentBAIR • Prof. Murat Arcak
Developing reachability analysis algorithms and RL environments for autonomus ships in the Arctic in collaboration with NTNU. (Sponsored by NSF IRES)
i'm athul, an undergrad at UC Berkeley interested in robotics, learning, and controls.
I'm a senior at UC Berkeley pursuing dual degrees in EECS and Mechanical Engineering. I'm currently working on research in reachability analysis and creating RL environments for autonomous ships, as well as a custom 6DOF BLDC robot arm.
UC Berkeley
EECS & MechE, B.S. • May 2026
Head TA: Dynamic Systems and Feedback, Controls/Signals
BAIR • Prof. Murat Arcak
Developing reachability analysis algorithms and RL environments for autonomus ships in the Arctic in collaboration with NTNU. (Sponsored by NSF IRES)
ICON Lab • Prof. Negar Mehr
Literature review for flow models, surveying potential applications in robotics.
Hybrid Robotics Lab • Prof. Koushil Sreenath
Designed a quadrupedal robot to demonstrate effectiveness of custom cycloidal gearboxes. Used ROS2 and RViz on Linux to simulate, model, and test robot motion and kinematics.
Engineering Intern • Palo Alto, CA
Owned thermal improvement effort on Model 3/Y Chargeports, reducing thermistor thermal deviation by 80x (now in production, ~2M units/year). Developed Python nonlinear optimizer for seal geometry. Negotiated 1M+ part quotes, saving $700K/yr.
Dynamics Lead • Berkeley, CA
Led team of 12 to design, optimize, and validate suspension and steering systems. Designed novel third element (3-spring) rear pushrod suspension. Overhauled FEA simulations across entire subsystem with realistic DOF restraints.
Senior Engineer • Berkeley, CA
Created thermal simulations in MATLAB for drone charging hive temperature distribution. Led development of 8'x8' cable robot gantry system with rigorous CAD and FEA validation practices.
Creating custom 6DOF arm with BLDC motors, differential joints, and 3-phase cycloidal drives.
Used UR7e robot arm to create LLM generated 'art' with tangram pieces from text prompts. Personally worked on custom suction gripper, website, and MCP pipeline.
Trained an RL policy for a Unitree Go2 to successfully perform a backflip, via implicit curriculum learning, gait rewards, and an asymmetric actor-critic architecture.
Finetuned a transformer architecture (GPT-3) to estimate LTI system responses given frequency components of input signals.
Built robot using stepper motors, 3D-printed parts, and Arduino to solve a Rubik's cube in 3 seconds.
Built traveling salesperson approximation improving on Christofides' Algorithm. Ranked 1st out of 100+ submissions, empirically ~5/4 approximation, on average.
Implemented neural network with FC and conv blocks using only numpy. Achieved 98% accuracy on CIFAR10.
Python, Java, C/C++, MATLAB, JavaScript, SQL, RISC-V
PyTorch, ROS2, MuJoCo, Isaac Sim, JAX, OpenCV, Git, Linux
Solidworks, OnShape, Fusion 360, Ansys, CATIA V5/V6
GD&T, 3D Printing, CNC, Lasercutting, Waterjetting, Soldering
Interested in collaborating or discussing research opportunities?
athul [at] berkeley [dot] edu