Dmitri Manajev

so101-ros-physical-ai. ROS 2 stack for SO-101 leader/follower: Feetech driver, bringup, teleop, MoveIt2, episode recording, Rerun, policy inference.

robokin. URDF-based forward and inverse kinematics helpers for robot arms — pluggable IK backends (Placo, PyRoki, RoboPlan), Viser 3D visualization, Rerun logging, and LeRobot/ROS 2 real-arm integration.

maddpg-zoo-torch. A modern implementation of MADDPG and MADDPG-Approx algorithms using PyTorch and PettingZoo environments. This project provides a clean, modular framework for multi-agent reinforcement learning research, featuring parallel training capabilities, comprehensive visualization tools, and support for various cooperative and competitive scenarios

marl-ppo-suite. Clean, documented implementations of PPO-based algorithms for cooperative multi-agent reinforcement learning, focusing on SMAC environments. Features MLP and RNN-based MAPPO and HAPPO with various techniques.

unity_multiagent_rl. Multi-agent reinforcement learning framework for Unity environments. Implements MAPPO, MASAC, MATD3, and MADDPG with comprehensive evaluation tools. Features sample-efficient training, competitive analysis, and pre-trained models achieving great performance in Tennis and Soccer environments.

manipulation_project. ROS 2 manipulation system for UR3E with vision-guided pick-and-place, motion planning, and gripper control. Simulation + real hardware ready.

unity-reacher-crawler-deeprl. Implementation and comparison of state-of-the-art deep reinforcement learning algorithms (DDPG, TD3, SAC) for continuous control tasks in Unity ML-Agents environments. Features multi-agent training, performance visualization, and detailed analysis across Reacher (robotic arm) and Crawler (quadruped locomotion) environments.

fastbot. Dockerized ROS2 env for Fastbot robot: Gazebo sim + real hardware (RPi), Nav2 SLAM, web UI & VPN remote access

fastbot_ros2. ROS 2 stack for the FastBot differential drive robot: real hardware control, LiDAR/camera integration, Gazebo simulation, Nav2 navigation, and RViz visualization. Inspired by The Construct.

fastbot_nav2_web. Web-based navigation interface for ROS 2 Fastbot robots, featuring real-time 3D/2D visualization, goal setting, camera streaming, and integration with Nav2, Cartographer SLAM, and rosbridge_suite.

robot_gui. A lightweight ROS1 GUI node built with cvui and OpenCV for real-time robot monitoring (info, odometry, velocities) and basic teleoperation via twist commands

warehouse_project. ROS2 Humble warehouse automation stack featuring AMCL localization, Nav2 path planning with recovery behaviors, Cartographer SLAM, and a custom orchestrator for autonomous shelf pick-and-place operations.