PaperlayerView exact revision on arXiv

Source-linked AI summary

DNC-IMM: Early Lane-Change Intention Recognition via Neural Calibration Based on Driving Context Information

Woong-Chan Byun, Seung-Hyun Kong

arXiv:2609.01120v1cs.ROcs.AI

TL;DR

Early lane-change intention recognition is needed to give autonomous vehicles and ADAS more time for proactive decisions, but existing approaches do not jointly provide interpretability and context-sensitive adaptation. DNC-IMM calibrates an IMM with neural driving-context signals while deriving intentions from its mode posterior; on highD, it performs particularly strongly at 2–3 s before lane crossing.

  • Problem

    Existing lane-change recognition often occurs only after lane-boundary crossing, while prior approaches do not jointly provide transparent probabilistic reasoning and learned adaptation to changing traffic context.

  • Method

    DNC-IMM uses neural driving-context features to calibrate IMM transition probabilities and measurement likelihoods, then selects LK, LCL, or LCR from the calibrated mode posterior.

  • Results

    0.9185 average macro F1 at the 2–3 s horizons, where DNC-IMM achieves the highest average among compared methods.

  • Takeaways & Limitations

    DNC-IMM recognizes lane-change intentions before lane crossing while retaining physically grounded, probabilistic reasoning and incorporating data-driven context sensitivity.

Abstract

from arXiv · show

Early recognition of lane-change intention is essential for proactive decision-making in autonomous driving and advanced driver assistance systems. This paper proposes a Dual Neural-Calibrated Interacting Multiple Model (DNC-IMM) that improves adaptability to driving context while preserving the probabilistic structure and interpretability of a conventional IMM. The proposed method encodes driving-context information, including target-vehicle motion, gaps to surrounding vehicles, and relative velocities, with a neural network that calibrates both the transition-probability matrix and measurement likelihoods. The final intention is determined from the calibrated IMM mode posterior rather than from a separate direct classifier. Experiments on the highD dataset demonstrate that the proposed method reliably recognizes lane-change intentions before lane crossing and provides particularly strong performance at the earlier 2-3 s prediction horizons.

I. INTRODUCTION

DNC-IMM addresses early lane-change intention recognition by adapting IMM inference to driving context while retaining probabilistic, interpretable decision-making. On highD, it recognizes LK, LCL, and LCR before lane crossing, with particularly strong performance at 2–3 s horizons.

  • Existing methods often detect lane changes only after lane-boundary crossing, limiting time for proactive collision avoidance, gap maintenance, and path planning.
  • Physics-based IMM methods are interpretable but fixed transition probabilities and likelihood models cannot fully reflect rapidly changing traffic context.
  • DNC-IMM calibrates IMM transition probabilities and measurement likelihoods using target-vehicle and surrounding-traffic features.
  • The neural network outputs corrections rather than direct maneuver classes, so the final LK, LCL, or LCR decision comes from the calibrated IMM mode posterior.
  • Comparative experiments show strong recognition 1–3 s before crossing, with the best average macro F1 at the more challenging 2–3 s horizons.

II. RELATED WORK

Prior lane-change intention-recognition research spans physics-based, adaptive, and data-driven models. DNC-IMM combines learned context adaptation with the probabilistic estimation structure of an IMM.

  • Physics-based approaches infer maneuver intention from explicit vehicle-motion models, probabilistic state transitions, and measurements.
  • Adaptive methods adjust model parameters or prediction thresholds using driving context, but rule-based corrections depend on designed interaction criteria.
  • Deep neural methods learn complex trajectory and interaction patterns directly, but their reasoning may be opaque or inconsistent with vehicle dynamics and roadway constraints.
  • DNC-IMM uses a Neural Calibration Branch to learn context corrections and an IMM Estimation Branch to incorporate them into Kalman-filter-based multi-model estimation.

A. Neural Calibration Branch

The Neural Calibration Branch encodes target-vehicle and surrounding-traffic context over a one-second history and produces bounded corrections for IMM transitions and likelihoods.

  • 1) Input Composition and Shared Encoder:: A shared MLP encodes target-vehicle features, surrounding context, one-second history statistics, and the prediction horizon.
  • 1) Input Composition and Shared Encoder:: The branch receives a normalized 186-dimensional input vector formed from target-vehicle and surrounding-context statistics plus a prediction-horizon indicator.
  • 1) Input Composition and Shared Encoder:: The shared encoder produces a 64-dimensional hidden representation that feeds two independently parameterized calibration heads.
  • The transition head outputs a bounded 3 × 3 transition-calibration matrix, while the likelihood head outputs a 3 × 1 vector for LK, LCL, and LCR.
  • Tanh scaling bounds corrections to [−3, 3], and mean removal retains relative likelihood scores because a shared constant cancels in softmax.
  • The correction signals are used only to calibrate IMM transition priors and measurement likelihoods, not to directly predict maneuvers.

B. IMM Estimation Branch

The IMM Estimation Branch inserts learned transition and likelihood corrections into separate stages of standard IMM estimation. Its calibrated mode posterior supplies the final state and maneuver outputs.

  • The branch receives previous state estimates, covariances, and mode probabilities for each maneuver model before performing standard IMM operations.
  • The resulting mode posterior µ_t provides the final maneuver decision and supports the final state and maneuver outputs.
  • Learned transition and likelihood corrections enter separate stages of the standard IMM procedure.

1) State Definition and Mode-Wise Kalman Estimation:

The IMM jointly represents longitudinal and lateral vehicle motion and maintains separate Kalman estimates for lane keeping and both lane-change directions. These mode-specific filters evaluate competing maneuver hypotheses in parallel from observed position and velocity information.

  • State definition: The state and observation representation jointly describes longitudinal and lateral motion using positions, velocities, and accelerations.The observation includes measurable position and velocity terms, while acceleration is inferred during Kalman estimation.
  • Mode-wise estimation: The IMM simultaneously maintains LK, LCL, and LCR motion modes.
  • Mode-wise estimation: A Kalman update is performed independently for every mode using the current observation.
  • Mode-wise estimation: The three filters retain distinct state and covariance estimates under the same observation because they use different maneuver assumptions.
  • Mode-wise estimation: The IMM evaluates lane keeping and both lane-change directions in parallel rather than relying on a single motion model.

2) Transition and Likelihood Matrix Calibration:

DNC-IMM calibrates both mode-transition probabilities and measurement likelihoods before forming the final IMM posterior. Row-wise normalization preserves a valid transition matrix, while clipping limits excessive likelihood corrections and the posterior remains the decision source.

  • Transition calibration: The neural transition-calibration matrix is combined with the base transition matrix in log space.
  • Transition calibration: Row-wise softmax normalization keeps the calibrated transition matrix valid by making every row sum to one.The coefficient βtrans controls the strength of transition calibration.
  • Likelihood calibration: A centered neural likelihood score calibrates each Kalman filter's measurement likelihood.The original log likelihood of mode j is used as the quantity being calibrated.
  • Posterior decision: The calibrated prior and likelihood are combined to obtain the final mode posterior, from which the maneuver estimate is selected.
  • Posterior decision: The network cannot bypass the IMM with a direct classification output, so each decision remains traceable to calibrated mode probabilities.

3) Posterior-Alignment Loss:

The training objective aligns the final IMM posterior with the ground-truth maneuver while constraining neural corrections and suppressing physically impossible lane changes. It therefore trains calibration within the IMM posterior framework rather than a separate classifier.

  • Posterior alignment: The final IMM posterior, rather than a separate classifier output, serves as the learning target.Training encourages transition and likelihood corrections to increase the posterior probability of the ground-truth maneuver.
  • Constraint terms: The loss combines posterior alignment with regularization that limits the magnitude of neural corrections.The coefficient λ∆ controls the contribution of the correction-magnitude regularizer.
  • Constraint terms: A lane-availability term suppresses LCL or LCR when the corresponding adjacent lane does not exist.The coefficient λinvalid determines the strength of this constraint.
  • Constraint terms: The objective combines posterior alignment with safeguards against overly large corrections and physically impossible lane-change predictions.

IV. EXPERIMENTAL RESULTS

DNC-IMM is evaluated on highD for LK, LCL, and LCR recognition before lane crossing, using class-wise and macro F1 alongside posterior trajectories. It achieves its strongest comparative performance at the earlier 2–3 s horizons and produces interpretable probability evolution.

  • Evaluation Setup: The evaluation measures LK, LCL, and LCR mode posteriors at 1, 2, and 3 s before lane crossing on highD.The dataset provides vehicle kinematics, lane information, and surrounding-vehicle gaps and relative velocities; recordings are split into training, validation, and testing groups.
  • Evaluation Metrics: Macro F1 gives equal weight to LK, LCL, and LCR, making it suitable for the class imbalance in early lane-change recognition.Performance is evaluated with both class-wise F1 and macro F1.
  • Comparative Performance: 0.9185 average macro F1 is obtained by DNC-IMM at the earlier 2–3 s horizons, the highest among the compared methods.DNC-IMM is slightly below Shi et al. at 1 s and in the all-horizon average, but exceeds the long-horizon averages of Mozaffari et al. and Liu et al.
  • Posterior Trajectories: The true lane-change posterior exceeds 0.5 before lane crossing in representative LCL and LCR scenarios, while P(LK) remains high in the LK scenario.The trajectories show probability mass moving toward the correct lane-change direction during the developing maneuver.
  • Interpretability: The calibrated posterior provides an interpretable account of probability mass moving from LK toward LCL or LCR as intention develops.This complements the final class decision with a probabilistic trajectory.

C. Component Ablation

The component ablation compares DNC-IMM with transition-only, likelihood-only, and uncalibrated IMM configurations to isolate the contributions of the two neural corrections.

  • Configurations: Four configurations are compared: DNC-IMM, Transition-only, Likelihood-only, and No Neural IMM.The first uses both corrections, the next two retain one correction, and the last removes learned calibration.
  • Ablation Design: The ablation isolates how transition calibration and likelihood calibration contribute to test-set performance.The comparison is reported as component ablation results on the test set.

D. Case Analysis of Neural Calibration Signals

The case analysis traces how driving context is converted into neural corrections and then into the IMM posterior. Using both corrections shifts the estimated maneuver from LK to LCR before lane crossing, while single-calibration variants respond later.

  • Signal and Posterior Alignment: Figure 5 aligns adjacent-lane front gap, lateral velocity, correction signals, transition probabilities, and the final IMM posterior on one time axis.The panels connect context changes to calibrated inference in an LCR case.
  • Component Effect: DNC-IMM identifies LCR from 3 s before lane crossing, whereas single-calibration variants respond later and uncalibrated IMM remains biased toward LK.The qualitative ablation highlights the joint use of transition and likelihood calibration.
  • Context-Driven Corrections: Increasing lateral motion and changing adjacent-lane front gap coincide with increases in the LCR likelihood and LK-to-LCR transition corrections.The neural branch transforms context into both correction types.
  • Posterior Shift: The calibrated LK-to-LCR transition probability rises above its base value, followed by a posterior shift from LK to LCR before lane crossing.This traces the complete calibration-to-decision chain.
  • Interpretation: The conclusion characterizes DNC-IMM as combining data-driven context sensitivity with physically grounded probabilistic reasoning.This follows from neural calibration of transition probabilities and measurement likelihoods while retaining the IMM posterior decision.
Loading 2609.01120v1…