RALTPER: A Risk-Aware Local Trajectory Planner for Complex Environment with Gaussian Uncertainty

TL;DR

This paper introduces RALTPER, a risk-aware local trajectory planner for autonomous vehicles that incorporates Gaussian uncertainty to improve safety and efficiency in complex environments.

Contribution

The paper presents a novel trajectory planning method that integrates probabilistic risk assessment with nonlinear optimization for better collision avoidance.

Findings

RALTPER improves safety in narrow and complex environments.

The method effectively balances collision risk and trajectory efficiency.

Simulation results show enhanced performance over traditional planners.

Abstract

In this paper, we propose a novel Risk-Aware Local Trajectory Planner (RALTPER) for autonomous vehicles in complex environments characterized by Gaussian uncertainty. The proposed method integrates risk awareness and trajectory planning by leveraging probabilistic models to evaluate the likelihood of collisions with dynamic and static obstacles. The RALTPER focuses on collision avoidance constraints for both the ego vehicle region and the Gaussian-obstacle risk region. Additionally, this work enhances the generalization of both vehicle and obstacle models, making the planner adaptable to a wider range of scenarios. Our approach formulates the planning problem as a nonlinear optimization, solved using the IPOPT solver within the CasADi environment. The planner is evaluated through simulations of various challenging scenarios, including complex, static, mixed environment and narrow single-lane avoidance of pedestrians. Results demonstrate that RALTPER achieves safer and more efficient trajectory planning particularly in navigating narrow areas where a more accurate vehicle profile representation is critical for avoiding collisions.