IR-STP: Enhancing Autonomous Driving with Interaction Reasoning in Spatio-Temporal Planning

TL;DR

This work addresses the challenge of obtaining interactive and robust trajectories for autonomous vehicles by embedding explicit interaction modeling into spatio-temporal planning. It defines influencer and reactor relations and organizes predictions into interaction zones to guide planning decisions, updating these relations continuously during a forward search. The proposed IR-STP framework integrates these concepts into a path-speed decoupled planning pipeline and demonstrates substantial improvements in collision timing, path completeness, and failure rates across diverse prediction scenarios in CommonRoad. The approach provides interpretable, tunable mechanisms to leverage multi-modal predictions and interaction dynamics, with open-source code enabling broader adoption and extension to planning in complex traffic environments.

Abstract

Considerable research efforts have been devoted to the development of motion planning algorithms, which form a cornerstone of the autonomous driving system (ADS). Nonetheless, acquiring an interactive and secure trajectory for the ADS remains challenging due to the complex nature of interaction modeling in planning. Modern planning methods still employ a uniform treatment of prediction outcomes and solely rely on collision-avoidance strategies, leading to suboptimal planning performance. To address this limitation, this paper presents a novel prediction-based interactive planning framework for autonomous driving. Our method incorporates interaction reasoning into spatio-temporal (s-t) planning by defining interaction conditions and constraints. Specifically, it records and continually updates interaction relations for each planned state throughout the forward search. We assess the performance of our approach alongside state-of-the-art methods in the CommonRoad environment. Our experiments include a total of 232 scenarios, with variations in the accuracy of prediction outcomes, modality, and degrees of planner aggressiveness. The experimental findings demonstrate the effectiveness and robustness of our method. It leads to a reduction of collision times by approximately 17.6% in 3-modal scenarios, along with improvements of nearly 7.6% in distance completeness and 31.7% in the fail rate in single-modal scenarios. For the community's reference, our code is accessible at https://github.com/ChenYingbing/IR-STP-Planner.

Figures (11)

• Figure 1: Example scenario: The motion planner is tasked with generating trajectories that dynamically interact with the predicted future trajectories of agents $V_i$ and $V_j$. Here, $T_2$ signifies the planning horizon, and $m_{(\cdot)}$ denotes the plausible future motions of the AV at time stamps $T_1$ and $T_2$. Multiple trajectory solutions are available. $m_1$: The most cautious choice, entailing yielding to both $V_i$ and $V_j$ to ensure safety. $m_2$-$m_3$: A purely reactive trajectory which reacts to $V_i$'s trajectory at all time steps, but leads to a collision with $V_j$'s trajectory at the $T_2$ time point. $m_2$-$m_4$: An interactive and more efficient trajectory strategy. Initially, it reacts to $V_i$'s trajectory to prevent collisions within the time interval $T_1$, reaching position $m_2$ at that moment. Then, it remains nonreactive to $V_i$'s subsequent motions ($t \in (T_1, T_2]$), assuming that $V_i$ will react to the AV's motion. Eventually, the trajectory reaches $m_4$ at time $T_2$, safely accommodating $V_j$'s trajectory, while $V_i$ takes responsive action to evade collisions with the AV.
• Figure 2: Structures of the proposed planning framework (on the right) and its connections with other system components.
• Figure 3: An example of the AV and its path merging with the reference route.
• Figure 4: Pipeline of general s-t searcher for AVs.
• Figure 5: Illustrations of interactions and interaction zones ${Z}_1$ and ${Z}_2$.