Jekyll2026-09-23T15:37:24+00:00https://pranayspeed.github.io/feed.xmlblankResearch portfolio of Pranay Meshram: language-conditioned planning, terrain abstraction, 3D perception, and field robotics. Monocular Visual Odometry in Rust (KITTI)2022-07-08T04:00:00+00:002022-07-08T04:00:00+00:00https://pranayspeed.github.io/updates/2022/mono-voBuilt a monocular visual odometry (VO) pipeline in Rust on the KITTI benchmark. Implemented feature tracking, pose estimation, and trajectory reconstruction, with plotting utilities to compare estimated trajectories against ground truth.

Benchmarked VO accuracy across sequences and experimented with tuning feature detection thresholds and RANSAC parameters to handle motion blur and texture-poor regions. Added demos and visualizations to highlight drift over long trajectories and loop-closure opportunities.

Git Repository

layout: post title: Monocular Visual Odometry (Rust) —

  • Monocular visual odometry implementation based on blog post
  • The implementation is done is Rust.
  • Used Kitti dataset for experiment.
  • Git Repository

Monovo1 Monovo2

]]>DAC SDC 2022 – FPGA-Friendly CNN (UBPercept)2022-06-21T04:00:00+00:002022-06-21T04:00:00+00:00https://pranayspeed.github.io/updates/2022/dac-sdcDAC System Design Contest 2022 – 4th place (UBPercept). Built an FPGA-friendly CNN pipeline on Ultra96V2 with quantization-aware training and deployment automation, targeting tight latency/memory budgets.

Led a 7-member team through data curation, profiling, and inference optimization; automated builds and on-board evaluation to iterate rapidly on accuracy–efficiency trade-offs. Documented lessons on model pruning and kernel fusion for edge devices.

Results

]]>DAC SDC 2022 - 4th Place (UBPercept)2022-06-21T00:00:00+00:002022-06-21T00:00:00+00:00https://pranayspeed.github.io/updates/2022/DAC-SDC-2022
  • DAC System Design Contest 2022 (June 2022) - 4th place [UBPercept] Results

    Contirbution

    • Led the team of 7 and contributed towards initial investigation and implementation.
    • Worked in a team to create plan for CNN training and deployment to FPGA (Ultra96V2)
    • Major contribution in automating the deployment and result analysis for fast and efficient experiments.
    • ]]>    Cleaner Bot – Coverage Planning Demo2020-12-20T04:00:00+00:002020-12-20T04:00:00+00:00https://pranayspeed.github.io/updates/2020/Cleaner-botCleaner Bot: a simple coverage-planning demo exploring unknown 2D space. Built a fixed-size unknown-environment exploration algorithm, visualized its sweep over the map, and linked a short video demo.

    Includes an interactive JS implementation to illustrate the circular BFS-style sweep and coverage behavior.

    ]]>    EBNF Grammar Parser in Python2019-03-06T04:00:00+00:002019-03-06T04:00:00+00:00https://pranayspeed.github.io/updates/2019/ebnf-parserExtended Backus–Naur Form grammar parser built from scratch in Python. Supports defining custom grammars, validating syntax, and running example parse workflows for compiler-construction practice and teaching.

    Includes utilities for grammar inspection and error highlighting to help debug language specs and parser behavior on sample inputs.

    Git Repository

    ]]>    BNF and EBNF Grammer Parser2019-03-06T00:00:00+00:002019-03-06T00:00:00+00:00https://pranayspeed.github.io/updates/2019/EBNF-Grammer-Parser
  • Implemented Extended Backus-Naur Form grammer parser from scratch.
  • Following is a sample grammer that can be parsed using this implementation:

    S := ‘Hello’ | ‘Hi’ P | P | ‘bye’
    P := ‘World’ | ‘Everyone’ | ‘[A-Za-z0-9]+ ‘ | ‘a’

  • Developed in Python.
  • Git Repository
    • ]]>    RL Self-Driving Agent (PyTorch)2018-03-07T04:00:00+00:002018-03-07T04:00:00+00:00https://pranayspeed.github.io/updates/2018/self-driving-carReinforcement Learning driving agent that learns end-to-end navigation from start to goal on 2D road maps. Implemented in PyTorch with a Kivy UI for interactive visualization, reward diagnostics, and episode playback.

    Explored reward shaping and curriculum tweaks to improve lane following, collision avoidance, and convergence speed. Logged metrics and rendered trajectories to compare training variants and diagnose failure cases in intersections and sharp turns.

    Git Repository

    ]]>    Self Dirving Car - using Reinforcement Learning2018-03-07T00:00:00+00:002018-03-07T00:00:00+00:00https://pranayspeed.github.io/updates/2018/self-driving-car-reinforcement-learning
  • Implementation an Agent to learn to get from start to goal position, given 2D map with road.
  • The Agent is support to follow the Road towards the goal position, and getting the highest reward.
  • Developed in Python, using Pytorch and Kivy for UI.
  • Git Repository
    • ]]>    Fusion 360 - 3MF exporter plugin2016-09-02T00:00:00+00:002016-09-02T00:00:00+00:00https://pranayspeed.github.io/updates/2016/Fusion360-3MF-Exporter
  • Implemented Fusion 360 Collada exporter for generating 3MF model file from CAD Model designed in Fusion 360.
    • This plugin is helpful if you want to do 3D print the CAD design
    • 3MF support color in addition to triangulation
    • Git Repository
    • ]]>    Fusion 360 - Collada exporter plugin2016-09-01T00:00:00+00:002016-09-01T00:00:00+00:00https://pranayspeed.github.io/updates/2016/Fusion360-Collada-Exporter
  • Implemented Fusion 360 Collada exporter for generating Collada model file from CAD Model designed in Fusion 360.
    • This plugin is helpful if you are a blender user and work with animation.
    • Blender compatible model exported.
    • Git Repository
    • ]]>