Reproduce “Decentralized iLQR for Cooperative Trajectory Planning of Connected Autonomous Vehicles via Dual Consensus ADMM” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhenmin Huang, Shaojie Shen, Jun Ma
A decentralized iterative Linear Quadratic Regulator (iLQR) framework, leveraging Dual Consensus Alternating Direction Method of Multipliers (ADMM), enables cooperative trajectory planning for connected autonomous vehicles. The method processes 12 vehicles in 0.250 seconds, exhibiting improved scalability and maintaining solution quality comparable to centralized approaches.
Key insights
- Decomposing the multi-agent optimal control problem into local LQR sub-problems via dual consensus ADMM provides a mechanism for efficient parallel computation across agents.
- Careful convex approximation of non-linear dynamics and non-convex collision avoidance costs is crucial for enabling the application of ADMM while preserving problem structure.
- Integrating iLQR's rapid convergence for non-linear optimal control with ADMM's distributed problem-solving capabilities offers a path to scalable, real-time cooperative multi-agent planning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.