KeshavAnandCode/resume
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

push

Browse Source
This commit is contained in:
Keshav Anand
2025-12-08 23:08:10 -06:00
parent c15d7aaaeb
commit bf4936568f
2 changed files with 22 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

BIN
rsi-application/main.pdf
View File

Binary file not shown.

23
rsi-application/main.tex
View File

@@ -248,7 +248,28 @@
1D CNN, resulting in state-of-the-art 99\% accuracy while limiting false positives. I also looked into signal processing 1D CNN, resulting in state-of-the-art 99\% accuracy while limiting false positives. I also looked into signal processing
within my FTC robotics team, realizing that IMU data could be used to improve odometry and localization. I implemented a within my FTC robotics team, realizing that IMU data could be used to improve odometry and localization. I implemented a
custom Kalman filter to fuse IMU data with wheel encoder reading, significantly reducing drift during autonomous navigation custom Kalman filter to fuse IMU data with wheel encoder reading, significantly reducing drift during autonomous navigation
and reducing error buildup over time. } and reducing error buildup over time.
My second major interest has been in the field of Robotics, springing from a lucky acceptance into my
Middle School's robotics team in 6th grade. Due to COVID-19, out team had to start from scratch, and as a
completely inexperienced 7th grader, it took me 7 months to simply learn to spin a motor. The same fascination
I had with computers was now being applied to physical hardware, and I have been a loyal participant in
the First Tech Challenge (FTC) robotics competition. As the software lead of my globally ranked team,
Technical Turbulence FTC, I have learned a lot about the algorithms that empower robots during the
30-second autonomous period of the competition. Today, I am intrigued by two major research questions within
the field of autonomous motion planning. First, I wonder how multiple autonomous agents can effectively
coordinate in real-time to achieve a common goal while avoiding collisions. This question fascinates me
because it combines elements of path planning, communication protocols, and decision-making under uncertainty.
Secondly, I am fascinated by the question of whether autonomous robots and vehicles can learn optimal paths from
experience rather than relying on pre-programmed maps. This idea of reinforcement learning for motion planning
excites me because it provides a pathway for devices to improve performance over time in dynamic environments.
My experience with robotics has provided me with a strong foundation to tackle these questions, as I have
designed and implemented a custom pathing algorithm for my FTC robot. The motion profiling algorithm I developed
uses cubic and quintic splines to generate smooth trajectories between points, using inverse kinematics and a PID
controller to accurate follow the path. By prioritizing endpoint accuracy over time and path accuracy, our robot's
pathing is extermely precise, resulting in a top-30 autonomous ranking globally. Outside FTC, I worked on a passion
project to allow for pathing of two vaccuum robots in a shared environment. Using A* for initial pathfinding and
a custom potential fields algorithm for real-time obstacle avoidance, I made a software system that allowed
for efficient cleaning of a dynamic space.}