ORC Homework Repository

This repository contains homework assignments and the final project for the Optimal Robot Control (ORC) course. AUTHORS: Alejandro Enrique Barbi & Alessandro Moscatelli

The repository is structured into three main assignments, each focusing on a different aspect of robot control.

---

📂 Repository Structure

1. Assignment 1: Task Space Inverse Dynamics (TSID)

Directory: A1_TSID/

This assignment focuses on Task Space Inverse Dynamics for bipedal robots (e.g., Romeo, Talos).

• Goal: Simulate bipedal locomotion, maintain balance, and track Center of Mass (CoM) trajectories.
• Key Scripts:
  • ex_1_biped.py: Basic standing and CoM tracking.
  • ex_2_biped_walking.py: Advanced walking simulation using pre-computed trajectories.
• Dependencies: pinocchio, tsid, example-robot-data.

2. Assignment 2: Model Predictive Control (MPC)

Directory: A2_MPC/

This assignment implements Model Predictive Control (MPC) problems for a UR5 robot arm using CasADi.

• Goal: Compute optimal trajectories for the UR5 end-effector to follow geometric paths (e.g., infinity shape) while minimizing costs (velocity, torque, etc.).
• Key Scripts:
  • A2_Q1.py: Basic path following OCP.
  • A2_Q2.py: Cyclic motion constraints.
  • A2_Q3.py: Trajectory tracking with time constraints.
  • A2_Q4_*.py: Time optimization strategies.
• Dependencies: casadi, adam-robotics, pinocchio.

3. Assignment 3: Value Function Approximation (RL/OCP)

Directory: A3_RL/

This constitutes the Final Project. It implements a pipeline to approximate the Value Function of an Optimal Control Problem (OCP) using Deep Learning.

• Goal: Train a Neural Network to regress the optimal cost from an initial state, allowing it to be used as a terminal cost in an MPC formulation.
• Workflow:
  1. Data Generation: Solving OCPs for a dense grid of initial states using CasADi/IPOPT.
  2. Training: Supervised learning with a PyTorch Neural Network.
  3. Simulation: Verifying performance on Single and Double Pendulum systems.
• Key Scripts: main.py (entry point for data gen, training, and sim).
• Dependencies: pytorch, casadi.

---

🚀 Getting Started

Each directory contains its own README.md with detailed instructions on how to install dependencies and run the specific scripts.

1. Navigate directly to the assignment folder:

2. Follow the specific instructions in that folder's README.md.