Spatio-temporal Keyframe Control of Traffic Simulation using Coarse-to-Fine Optimization

Yi Han,He Wang,Xiaogang Jin

DOI: https://doi.org/10.1111/cgf.14699

2022-09-26

Abstract:We present a novel traffic trajectory editing method which uses spatio-temporal keyframes to control vehicles during the simulation to generate desired traffic trajectories. By taking self-motivation, path following and collision avoidance into account, the proposed force-based traffic simulation framework updates vehicle's motions in both the Frenet coordinates and the Cartesian coordinates. With the way-points from users, lane-level navigation can be generated by reference path planning. With a given keyframe, the coarse-to-fine optimization is proposed to efficiently generate the plausible trajectory which can satisfy the spatio-temporal constraints. At first, a directed state-time graph constructed along the reference path is used to search for a coarse-grained trajectory by mapping the keyframe as the goal. Then, using the information extracted from the coarse trajectory as initialization, adjoint-based optimization is applied to generate a finer trajectory with smooth motions based on our force-based simulation. We validate our method with extensive experiments.

Graphics,Robotics

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is: How to achieve spatio - temporal key - frame control of vehicle movement in traffic simulation to generate the desired traffic trajectories that meet specific spatio - temporal constraints. Specifically, the current traffic simulation frameworks based on force models usually restrict vehicles to straight lanes, which are difficult to meet the needs of users' arbitrary settings or edits; at the same time, there is also a lack of methods for optimizing traffic trajectories with spatio - temporal key - frame control. Therefore, this paper proposes a new traffic trajectory editing method, which allows users to adjust vehicle movement through spatio - temporal key - frames and ensures the rationality of traffic behavior.

### Problems Solved by the Paper

1. **Limitations of Existing Methods**:
- The current traffic simulation frameworks based on force models restrict vehicles to straight lanes, which are difficult to meet the needs of users' arbitrary settings or edits.
- There is a lack of methods for optimizing traffic trajectories with spatio - temporal key - frame control, resulting in the inability to generate the desired traffic trajectories that meet specific spatio - temporal constraints.

2. **Proposed Solutions**:
- A new force - model - based framework is proposed, which decouples vehicle movement from the limitations of static lanes and can generate microscopic traffic simulations in complex scenarios, including diverse interactions.
- A coarse - to - fine optimization process is introduced, which combines the adjoint method and state - time - space search to effectively generate smooth trajectories and meet key - frame constraints.

### Main Contributions

- **Traffic Trajectory Editing Method**: Allows users to specify key - frames to adjust vehicle movement, generate the desired trajectories that meet spatio - temporal constraints, and at the same time maintain the rationality of traffic behavior.
- **New Force - Based Framework**: Decouples vehicle movement from the limitations of static lanes and can generate microscopic traffic simulations in complex scenarios, including diverse interactions.
- **Coarse - to - Fine Optimization Process**: Through the combination of the adjoint method and state - time - space search, effectively performs gradient descent to generate smooth trajectories that meet key - frame constraints.

### Formula Representation

All formulas involved in the paper are accurately presented in Markdown format:

For example, the self - driving force modeling formula is:
\[
\hat{f}_{o,t} = \omega_{o} m \left( \frac{2}{1 + e^{\hat{v}_{o,t} - \hat{v}_{t-1}}} \right) \hat{a}
\]

The path - keeping force formula is:
\[
\hat{f}_k =
\begin{cases}
\omega_k |d_t| u_k, & \text{if } |d_t| \geq \frac{1}{2}(w_l - w_v) \\
0, & \text{otherwise}
\end{cases}
\]

The collision - avoidance force formula is:
\[
f_j = \omega_{ca} \cdot \frac{b}{(1 + c \| p_{j,t} - p_t \|)^2} u_c
\]

These formulas ensure the computational accuracy and readability of the model.

### Summary

This paper solves the deficiencies of existing methods in spatio - temporal key - frame control by introducing a coarse - to - fine optimization process and a force - based traffic simulation framework, and provides an effective traffic trajectory editing method.