Efficient LiDAR-Trajectory Affinity Model for Autonomous Vehicle Orchestration

Computation and memory resource management strategies are the backbone of continuous object tracking in intelligent vehicle orchestration. Multi-object generates enormous measurements targets extended positions using light detection ranging (Lidar) sensors. Designing an adequate object-tracking system is a global challenge because dynamic data association uncertainties during scene understanding. In this regard, we develop multi-objective (IMOT) with novel measurement model, called box inflate (BDAI) to assess each target’s state trajectory without noise by Bayesian approach. The filter method filters ambiguous responses association. theoretical proof derived based on binomial expansion. Prognosticating lower-dimension than original point reduces computational complexity Two datasets (NuScenes dataset our lab dataset) considered simulations, approach measures kinematic states adequately reduced computation compared state-of-the-art methods. simulation outcomes show that proposed effective works well detect track objects. NuScenes contains 28130 samples for training, 6019 examples validation 6008 testing. IMOT achieves 58.09% accuracy 71% mAP 5 ms pre-processing time. Jetson Xavier NX consumes 49.63% GPU 9.37% average power exhibits 25.32 latency other approaches. Our trains single pair frame 169.71 affinity estimation time 12.19 ms, 0.19 mATE 0.245 approaches