Publication record · 18.cifr/2017.williams.it-mpc

Information Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving

v1.0.0

Grady Williams (Georgia Institute of Technology), Paul Drews (Georgia Institute of Technology), Brian Goldfain (Georgia Institute of Technology), James M. Rehg (Georgia Institute of Technology), Evangelos A. Theodorou (Georgia Institute of Technology)

RAI18.cifr/2017.williams.it-mpc

IEEE International Conference on Robotics and Automation (ICRA)· 2017· doi:10.1109/ICRA.2017.7989202

We present an information theoretic approach to stochastic optimal control problems that can be used to derive general sampling based optimization schemes. This new mathematical method is used to develop a sampling based model predictive control algorithm. We apply this information theoretic model predictive control (IT-MPC) scheme to the task of aggressive autonomous driving around a dirt test track, and compare its performance to a model predictive control version of the cross-entropy method.

model predictive controlstochastic optimal controlinformation theoryautonomous drivingsampling-based optimization

✦ Research context

What this agent contributes to the literature.

Problem solved

Classical MPC cannot handle non-smooth costs arising in aggressive off-road driving without expensive gradient computations. Prior sampling-based MPC methods lacked a unified theoretical foundation linking the sampling distribution update to an optimal control objective. IT-MPC addresses both gaps with GPU-parallelizable rollouts.

Novelty

IT-MPC derives its sampling update from information-theoretic first principles (free energy minimization under a KL-divergence constraint) rather than heuristically proposing a sampling scheme. This gives a provably optimal importance-sampling update rule that handles non-convex, non-differentiable cost functions without gradients.

Related research

Computing related research...

Canvas contract1-in / 1-out · unpacked into scenario legacy ports

Sample data

Loading sample data...

Total calls

This month

Citations

Last called

—

Image digest

sha256:27aa77c4f6dedb3b9e04713b1c2e77e4c89b723227ddc6d5b2a1be99ad83449d

Invoke command

python main.py

Inputs

input:application/json

Outputs

output:application/json

Citation

Loading DOI…

Invoke

CPU compute only

How to get GPU access: Your university, lab, or company can become a CIFR institutional member. Members get GPU-accelerated runs for all their researchers. Contact us

Pre-filled with the paper's canonical scenario. Click Invoke agent to reproduce the original result, or edit the JSON below to run a counterfactual.

inputapplication/jsonoptional

Unified canvas input — nested keys described in legacy_inputs

Leave empty to run the paper's canonical scenario.

Recent invocations(0)

No invocations yet — be the first to call this agent.