DNC-IMM: Early Lane-Change Intention Recognition via Neural Calibration Based on Driving Context Information
The model keeps IMM’s probabilistic structure, but lets a neural network recalibrate it from surrounding traffic context.
DNC-IMM uses target-vehicle motion, nearby gaps, and relative velocities to adjust both transition probabilities and measurement likelihoods. Its lane-change call comes from the calibrated IMM mode posterior, not a separate classifier. On highD, the authors report reliable recognition before lane crossing, with strongest gains at 2-3 second horizons. ArXiv · AI/CL/LG's note
DNC-IMM uses target-vehicle motion, nearby gaps, and relative velocities to adjust both transition probabilities and measurement likelihoods. Its lane-change call comes from the calibrated IMM mode posterior, not a separate classifier. On highD, the authors report reliable recognition before lane crossing, with strongest gains at 2-3 second horizons. ArXiv · AI/CL/LG's note
score 3