Optimization tool for planning overtaking maneuvers

This tool is a part of the master's thesis of Tomáš Nagy.

It is based on a ng_trajectory[1] tool for global trajectory optimization.

To view the package documentation either run ng_help.

This tool was used to identify possible overtaking zones on a given track when the opponent's racing line is known. Moreover, we assume that the opponent cannot perform a blocking move (F1TENTH vehicles do not have a rear-facing sensor). Examples can be seen in the Section "Overtaking zones examples". The detail description of this tool and the results can be found in !NOT YET PUBLISHED!.

Requirements

• Python>=3.6
• nevergrad==0.3.0
• scipy>=0.18.0
• numpy>=1.12.0
• Pillow>=4.2.0
• tqdm

Optional requirements

• matplotlib for plotting the results (tested on 3.6.0)
• rospy for publishing the results to various ROS topics
• shapely (EXPERIMENTAL) for computing the Matryoshka transformation with a different method

Overtaking zones examples

We optimized many different maneuvers while varying initial position to determine possible overtaking zones. We present the overtaking zones in the form of a heat map where red means "most of the overtakes happened at this place".

Example 1

In this example, the opponent drives different trajectory types (purple dotted line).

Example 2

In this example, the ego vehicle has a friction coefficient increased. The opponent's trajectory is the purple dotted line.

Example 3

In this example, the ego vehicle has a maximum speed increased. The opponent's trajectory is the purple dotted line.

Citing

[1]: J. Klapálek, A. Novák, M. Sojka and Z. Hanzálek, "Car Racing Line Optimization with Genetic Algorithm using Approximate Homeomorphism," 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021, pp. 601-607, doi: 10.1109/IROS51168.2021.9636503.