ALOHOMORA is a vision-based blind stick. It aims at providing the visually impaired people with an easyand handy object identification and navigation system. The model that we will be using for objectidentification has been trained on the images of famous objects/obstacles that can be encountered by a visually impaired person in everyday life. This information will act as the basis for the results given to the user. Also, the distance to the object/obstacle will be calculated using a sensor and accordingly theinformation will be used to inform the person about the range or distance at which the object is.Furthermore, the system needs a continuous power supply to for its operation. The program on themini-computer should always be running in order to provide satisfying results to the visually impaired person. There is more more to come in project ,until wait Technologies used- 1.SOFTWARE - Operating System - Ubuntu 16.04,NOOBS 2.1.0 (Raspbian Pixel) Language - Python3 Machine Learning Libraries - Tensorflow 0.12.0,0.12.0, Inception v3 2.HARDWARE- Raspberry Pi model 3B, Ultrasonic Sensor HC SR-04, Passive Speaker Buzzer

Continuous Learning v3 for Claude Code: hook-based session recording, bash interception, and pattern detection

DQN Solution for LunarLander-v3 (Gymnasium) | Reinforcement learning agent using Deep Q-Network (DQN) to land a lunar module safely. Implements experience replay, target networks, and training visualization. This project serves as a hands-on reinforcement learning experiment for solving continuous control tasks using deep Q-learning.

A reinforcement learning project for training agents to navigate the continuous control driving environment CarRacing-v3 using a CNN-based policy with the PPO algorithm.

ACT v3.5 — Agentic Execution Framework for Claude Code. 7-phase project methodology, subagent orchestration (SDD), two-stage review, 5-level scale, 14 skills, multi-project management, Iron Laws, Thinking Models, Continuous Learning.

Automated LLM-Guided Reinforcement Learning Testbed. This project leverages the modern BipedalWalker-v3 environment from Gymnasium to orchestrate a continuous cycle of agent training and intelligent reward shaping. By combining Stable Baselines3's PPO algorithm with the reasoning capabilities of Large Language Models (LLMs)

Based on a skill in the every-claude-code repo, "Continuous learning v3"is now adds support for Windows and you can choose your model freely.

ARI3212 : Advanced Reinforcement Learning Project | Solving LunarLander-v3 (discrete) with DQN vs Double DQN (robustness under wind/turbulence), plus LunarLanderContinuous-v3 (continuous) with PPO vs SAC to compare on-policy vs off-policy learning, including hyperparameter tuning and learning-curve evaluation.

This project investigates how temporal-differences (TD) learning scales from a tabular formulation with state discretization to deep neural network function approximation on a continuous-state control task - Gymnasium LunarLander-v3 environment

A high-performance, production-grade implementation of **Soft Actor-Critic (SAC)** designed to solve the continuous control task `BipedalWalker-v3`. This project demonstrates rigorous Reinforcement Learning methodology, from modern SAC algorithm design to reproducible training pipelines.

Train a Soft Actor-Critic (SAC) agent to master the BipedalWalker-v3 environment. This end-to-end RL solution features robust actor-critic networks, a replay buffer, and real-time visual demos to showcase the agent’s continuous control performance and learning process.

"Built an autonomous driving agent that learned to master the CarRacing-v3 environment from scratch using Deep Reinforcement Learning (PPO). The agent processes raw pixel data through a Convolutional Neural Network (CNN) to make continuous control decisions (steering, gas, brake) in real-time. I engineered a robust training pipeline in Python using