Controls & Optimization in Robotics
Controls & Optimization in Robotics
Embedded Files
MEAM 5170: Controls and Optimization in Robotics at the University of Pennsylvania is an advanced graduate course focusing on model-based nonlinear control and computational techniques for dynamic robotic applications. The curriculum combines theoretical foundations of nonlinear and optimal control with practical, optimization-based algorithmic methods. Key topics include optimal control, dynamic programming, trajectory optimization, canonical underactuated systems, stability analysis, control of limit cycles, and model predictive control. The course also explores advanced applications such as walking and running robots, nonsmooth mechanics, and multi-contact dynamics. Students engage with cutting-edge research through the review of relevant academic papers, equipping them with the skills and knowledge to address complex challenges in robotics.
quadrotor_trajectory_TPV_ref_traj.mp4quadrotor_trajectory_TPV_forest3.mp4Final Simulation of Quadcopter Drone - On its way to a Forest Fire
Final Simulation of Quadcopter Drone - On its way to a Forest Fire
Paper explaining the Project
Paper explaining the Project
Final_report_MEAM_5170.pdfProject Details:
Project Details:
MPC Forest Fire Quadcopter
MPC Forest Fire Quadcopter
Project Overview:
Developed a 3D simulation for a quadcopter navigating a dense forest to extinguish a fire.
Combined RRT* for path planning, LPV-MPC for control, and real-time trajectory optimization to handle dynamic constraints.
Github Link: https://github.com/cedrichld/mpc-fire-quadrotor
Steps Taken in the project:
Steps Taken in the project:
Forest Environment Setup:
Randomly generate a 3D forest with cylindrical trees, varying in location, radius, and height, for realistic obstacle mapping.
Defined a fire zone using spline-based boundaries as the simulation’s target region.
Path Planning:
Implemented an RRT* algorithm to compute collision-free paths in cluttered environments.
Enhanced the paths with smoothing and obstacle clearance checks for practical quadcopter navigation.
Quadcopter Dynamics Modeling:
Simulated nonlinear dynamics of the quadrotor.
Incorporated system constraints like velocity, angular rates, and obstacle proximity for realistic behavior.
Control and Trajectory Optimization:
Utilized LPV-MPC for trajectory tracking, dynamically adjusting control inputs based on the quadrotor’s state.
Solved quadratic programming (QP) problems at each timestep to minimize tracking errors while ensuring feasibility.
Visualization and Results:
Created a 3D visualization of the quadcopter navigating the forest and reaching the fire zone.
Demonstrated successful path tracking with both low and high integral steps, highlighting trade-offs between precision and computation time.
Preliminary Testing:
Verified collision-free navigation using RRT* and accurate trajectory following via LPV-MPC.
Addressed small deviations caused by discretization and control tuning.
Page updated
Report abuse