FlowDrive: moderated flow matching with data balancing for trajectory planning

TL;DR

FlowDrive addresses the challenge of long-tailed driving data by combining data-balancing strategies with a fast, flow-matching trajectory planner. It learns a conditional velocity field $ oldsymbol{v}_{oldsymbol{ heta}}(t,oldsymbol{x},oldsymbol{c}) $ that transports a simple base distribution $ p_0 $ to the conditional data distribution $ p_{ ext{data}}(oldsymbol{x}|oldsymbol{c})$ along a rectified path, enabling few-step sampling. Data balancing via cluster-based trajectory clustering and scenario-based weighting expands coverage of rare but critical maneuvers, while moderated, in-the-loop guidance injects controlled lateral perturbations to promote multimodal, feasible trajectories. Empirically, FlowDrive and FlowDrive$^*$ achieve state-of-the-art or near state-of-the-art results on nuPlan and interPlan benchmarks with efficient inference, demonstrating strong robustness and diversity without heavy post-processing.

Abstract

Learning-based planners are sensitive to the long-tailed distribution of driving data. Common maneuvers dominate datasets, while dangerous or rare scenarios are sparse. This imbalance can bias models toward the frequent cases and degrade performance on critical scenarios. To tackle this problem, we compare balancing strategies for sampling training data and find reweighting by trajectory pattern an effective approach. We then present FlowDrive, a flow-matching trajectory planner that learns a conditional rectified flow to map noise directly to trajectory distributions with few flow-matching steps. We further introduce moderated, in-the-loop guidance that injects small perturbation between flow steps to systematically increase trajectory diversity while remaining scene-consistent. On nuPlan and the interaction-focused interPlan benchmarks, FlowDrive achieves state-of-the-art results among learning-based planners and approaches methods with rule-based refinements. After adding moderated guidance and light post-processing (FlowDrive*), it achieves overall state-of-the-art performance across nearly all benchmark splits. Our code is available at https://github.com/einsteinguang/flow_drive_planner.

Figures (10)

• Figure 1: FlowDrive architecture. Left: scene inputs (neighbor history, static objects, lanes/routes with traffic lights and speed limits). Middle: encoder with MLP-Mixer branches and multi-head attention fusion. Right: DiT-based decoder with adaptive layer norm conditioning and cross-attention to context, predicting the velocity field across flow steps.
• Figure 2: Precomputed clusters of normalized ego futures used for cluster-based sampling. Colored polylines indicate cluster centers; translucent points show a subset of sampled trajectories per cluster.
• Figure 3: Sample counts per cluster after applying different training-time sampling strategies. Cluster-based sampling produces the most balanced distributions over the clusters.
• Figure 4: Effect of injecting moderated lateral offsets at different flow steps. Black trajectories are with 0 offset, green trajectories are with offsets $[-0.5, -0.25, 0.25, 0.5]$.
• Figure 5: Moderated lateral offsets induce multi-modality. Beyond lane-following green trajectories, guided samples explore safe overtaking when context allows (black).