Giters

LeCAR-Lab / CoVO-MPC

Official implementation for the paper "CoVO-MPC: Theoretical Analysis of Sampling-based MPC and Optimal Covariance Design" accepted by L4DC 2024. CoVO-MPC is an optimal sampling-based MPC algorithm.

Home Page:https://lecar-lab.github.io/CoVO-MPC/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

📦 Transportation Project Record

jc-bao opened this issue · comments

commented

Research Question

Addressing Model Uncertainty in Quadrotor Slung Load Transportation

Contribution

This research aims to tackle the challenges associated with model uncertainty in the context of quadrotor slung load transportation. The contributions of this study are threefold:

(a) modelling(no loose state, highly underactuated), $\leftarrow$ which previous work not considered, which lead to conservative real-world policy
(b) optimization (hybrid model & entanglement with lower-level controller) $\leftarrow$ where classical model-based method fails and
(c) uncertainty/adaptation (multi-agent entanglement & lower-level higher-level controller entanglement & soft body & high-speed) $\leftarrow$ where RL fails}

By addressing these aspects, our research aims to provide a comprehensive solution to the quadrotor slung load transportation problem, particularly in the presence of model uncertainty.

commented

Baseline Implementation

In the initial phase of our research project, we aim to establish a solid foundation by focusing on the following steps:

  1. Replicating the results from previous transportation studies using actual hardware.
  2. Experimenting with the hardware to pinpoint potential research questions.

Relevant Literature

Here are some key papers that provide insights and methodologies related to our project:

  • Geometric Control and Differential Flatness of a Quadrotor UAV with a Cable-Suspended Load: This paper delves into the modeling, differential flatness, and hierarchical controller aspects.

  • Differential Flatness Based Path Planning With Direct Collocation on Hybrid Modes for a Quadrotor With a Cable-Suspended Payload: This study emphasizes path planning strategies.

  • Model-Based Meta-Reinforcement Learning for Flight with Suspended Payloads: Notably, this paper reported suboptimal performance, which could offer insights into potential pitfalls or areas of improvement.

Engineering Tasks

  • System Modeling: Understand and develop a comprehensive model of the system.

  • Algorithm Replication: Ensure that the algorithm from the referenced studies is accurately recreated for our experiments.

Areas of Uncertainty

A critical question we need to address is:

  • With our updated system model, does the system retain its differential flatness properties?

By addressing this question, we can ensure that our foundational understanding of the system is robust and accurate, setting the stage for further advancements in our research.