Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reachability-Based Contingency Planning against Multi-Modal Predictions with Branch MPC

Mohamed-Khalil Bouzidi, Bojan Derajic, Daniel Goehring, Joerg Reichardt

TL;DR

A novel contingency planning framework that integrates learning-based multi-modal predictions of traffic participants into Branch Model Predictive Control (MPC), and addresses the computational challenges associated with Branch MPC by organizing the multitude of predictions into driving corridors.

Abstract

This paper presents a novel contingency planning framework that integrates learning-based multi-modal predictions of traffic participants into Branch Model Predictive Control (MPC). Leveraging reachability analysis, we address the computational challenges associated with Branch MPC by organizing the multitude of predictions into driving corridors. Analyzing the overlap between these corridors, their number can be reduced through pruning and clustering while ensuring safety since all prediction modes are preserved. These processed corridors directly correspond to the distinct branches of the scenario tree and provide an efficient constraint representation for the Branch MPC. We further utilize the reachability for determining maximum feasible decision postponing times, ensuring that branching decisions remain executable. Qualitative and quantitative evaluations demonstrate significantly reduced computational complexity and enhanced safety and comfort.

Reachability-Based Contingency Planning against Multi-Modal Predictions with Branch MPC

TL;DR

A novel contingency planning framework that integrates learning-based multi-modal predictions of traffic participants into Branch Model Predictive Control (MPC), and addresses the computational challenges associated with Branch MPC by organizing the multitude of predictions into driving corridors.

Abstract

This paper presents a novel contingency planning framework that integrates learning-based multi-modal predictions of traffic participants into Branch Model Predictive Control (MPC). Leveraging reachability analysis, we address the computational challenges associated with Branch MPC by organizing the multitude of predictions into driving corridors. Analyzing the overlap between these corridors, their number can be reduced through pruning and clustering while ensuring safety since all prediction modes are preserved. These processed corridors directly correspond to the distinct branches of the scenario tree and provide an efficient constraint representation for the Branch MPC. We further utilize the reachability for determining maximum feasible decision postponing times, ensuring that branching decisions remain executable. Qualitative and quantitative evaluations demonstrate significantly reduced computational complexity and enhanced safety and comfort.

Paper Structure

This paper contains 14 sections, 14 equations, 9 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Learning-based Multi-Modal Motion Predictor
  4. Nominal Model Predictive Contouring Control
  5. Branch Model Predictive Contouring Control
  6. Reachability-based Scenario Selection
  7. Driving Corridor Extraction
  8. Clustering and Pruning
  9. Collision Constraints Formulation
  10. Maximum Decision Postponing Consideration
  11. Performance Evaluation
  12. Qualitative Evaluation
  13. Quantitative Evaluation
  14. Conclusion

Figures (9)

  • Figure 1: Illustration of planning in traffic scenes with uncertain multi-modal predictions: Mode-dependent driving corridors are extracted to construct the Branch MPC scenario tree.
  • Figure 2: Our Contingency Planning Framework using a Multi-modal Predictor, Reachability-based Scenario Selection, Adaptive Decision Postponing and Branch MPCC
  • Figure 3: For each predicted mode of the TP, at least one safe driving corridor is identified. The upper plots show the respective corridors in a bird's-eye view, while the lower plots illustrate them in a path-time diagram.
  • Figure 4: Illustration of the collision constraint formulation for respective driving corridor of one branch $\mathcal{D}^s_{0:N}$
  • Figure 5: Snapshot with prediction of TPs and the selected driving corridors and respective plans of the AV.
  • ...and 4 more figures