Abstract:Autonomous cooperative planning (ACP) is a promising technique to improve the efficiency and safety of multi-vehicle interactions for future intelligent transportation systems. However, realizing robust ACP is a challenge due to the aggregation of perception, motion, and communication uncertainties. This paper proposes a novel multi-uncertainty aware ACP (MUACP) framework that simultaneously accounts for multiple types of uncertainties via regularized cooperative model predictive control (RC-MPC). The regularizers and constraints for perception, motion, and communication are constructed according to the confidence levels, weather conditions, and outage probabilities, respectively. The effectiveness of the proposed method is evaluated in the Car Learning to Act (CARLA) simulation platform. Results demonstrate that the proposed MUACP efficiently performs cooperative formation in real time and outperforms other benchmark approaches in various scenarios under imperfect knowledge of the environment.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: In future intelligent transportation systems, the efficiency and safety of multi - vehicle interaction can be improved through Autonomous Collaborative Planning (ACP) technology. However, achieving robust ACP faces many challenges, mainly due to the influence of perception, motion and communication uncertainties.
Specifically:
1. **Perception Uncertainty**: Learning - based perception models (such as deep neural networks) may produce errors, and these errors will propagate to the model - based planning stage, thus affecting the final trajectory planning.
2. **Motion Uncertainty**: There may be a deviation between the planned trajectory and the actual driving trajectory, and this deviation will be greater especially in bad weather conditions.
3. **Communication Uncertainty**: Shifting from single - vehicle perception to multi - vehicle perception can significantly reduce perception uncertainty, but imperfect channel state information may lead to communication interruptions, thus affecting information fusion.
To solve these problems, this paper proposes a new Multi - Uncertainty - Aware Autonomous Collaborative Planning framework (MUACP), which simultaneously considers multiple types of uncertainties through Regularized Cooperative Model Predictive Control (RC - MPC). Specifically, the paper deals with these uncertainties in the following ways:
- **Perception Uncertainty**: Construct a regularization term through the confidence level.
- **Motion Uncertainty**: Construct a regularization term according to wheel feedback and weather conditions.
- **Communication Uncertainty**: Construct constraints according to the channel distribution and outage probability.
This method has been verified on the CARLA simulation platform, and the results show that MUACP can efficiently perform collaborative formation in a real - time environment and outperforms other benchmark methods in various environmental conditions.
### Formula Summary
1. **Sideslip Angle Calculation**:
\[
\beta_{k,t}=\arctan\left(\frac{\tan(\delta_{k,t})l_r}{l_f + l_r}\right)
\]
where \(l_f\) and \(l_r\) are the distances of the front axle and the rear axle respectively.
2. **State Update Equation**:
\[
\begin{aligned}
x_{k,t + 1}&=x_{k,t}+v_{k,t}\cos(\phi_{k,t}+\beta_{k,t})\Delta t,\\
y_{k,t + 1}&=y_{k,t}+v_{k,t}\sin(\phi_{k,t}+\beta_{k,t})\Delta t,\\
\phi_{k,t + 1}&=\phi_{k,t}+\frac{v_{k,t}\cos(\beta_{k,t})\tan(\delta_{k,t})}{l_f + l_r}\Delta t,\\
v_{k,t + 1}&=v_{k,t}+a_{k,t}\Delta t.
\end{aligned}
\]
3. **Dynamic Safety Distance**:
\[
d_{kj,t}=d_{\text{min}}+(1 - \rho_{kj,t})d_{\text{max}}
\]
where \(d_{\text{min}}\) is the minimum safety distance, \(d_{\text{max}}\) is the maximum detection error, and \(\rho_{kj,t}\) is the confidence score.
4. **Motion Regularization Term**:
\[
R_{k,t}=\alpha_t\|\Delta e_{z_{k,t}}\|^2
\]
where \(\alpha_t\) is the penalty coefficient and \(\Delta e_{z_{k,t}}\) is the deviation between the target control and the actual control.
Through these formulas and methods, the paper successfully incorporates multiple uncertainties into a unified optimization framework, improving the robustness and performance of multi - vehicle collaborative planning.