zhongshun/MCTS_for_Behavior_Planning

Decision-making in Autonomous Driving using Monte Carlo Tree Search (MCTS)

This repository contains the code, datasets, and supplementary materials related to our research paper on leveraging MCTS for decision-making in autonomous vehicles.

Abstract

We present a comprehensive framework based on Monte Carlo Tree Search for decision-making in autonomous driving scenarios. Through extensive simulations in MATLAB's autonomous driving toolbox 2023a (Note that some functions may not supported for lower version.). We showcase the framework's efficacy across various driving conditions, from intricate urban intersections to highway exits. While our simulations demonstrate promising results, we highlight areas for potential improvement and suggest future research directions.

Setup & Usage

1. Clone the repository
2. Choose or set the environment. Make sure the version of MATLAB is above 2023a and the version of Automated Driving Toolbox is above 3.7.
3. Run mctsPlanning.m

Qualitative Results

Real-world Road Scenarios

We demonstrate the effectiveness of our autonomous driving algorithm in navigating through typical real-world environments, from left to right: Navigation through an Intersection, Unprotected Left Turn at Intersection, Managing Merging and Diverging, and Navigation through a Roundabout.

Roads in Shanghai, China.

Birds Eye View:

Typical Results:

Typical Highway Scenarios

Qualitative Results 1

Figure 1: Highway Exit (HE) example.

Qualitative Results 2

Figure 2: Large Curvature example.

Typical Urban Scenarios

Qualitative Results 3

Figure 3: Intersection, Straight example.

Qualitative Results 4

Figure 4: Intersection, Unprotected Left Turn example.

Qualitative Results 5

Figure 5: Intersection, Unprotected Left Turn example 2.

Qualitative Results 6

Figure 6: Intersection, Unprotected Right Turn example.

Qualitative Results 7

Figure 7: Intersection, Unprotected Straight Cross example.

Qualitative Results 8

Figure 8: Intersection, Blocked example.

Qualitative Results 9

Figure 9: Intersection, Blocked by Stationary Objects example.

Qualitative Results 10

Figure 10: Intersection, Go Straight example.

Qualitative Results 11

Figure 11: Intersection, Go Straight example 2.