Browse Publications Technical Papers 2020-01-5167

Multi-target Tracking Algorithm with Adaptive Motion Model for Autonomous Urban Driving 2020-01-5167

Since situational awareness is crucial for autonomous driving in urban environments, multi-target tracking has become an increasingly popular research topic during the last several years. For autonomous driving in urban environments, cars and pedestrians are the two main types of obstacles, and their motion characteristics are not the same. While in the current related multi-target tracking research, the same motion model (such as Constant Velocity model [CV]) or motion model set (such as CV combined with Constant Acceleration model [CA]) is mostly used to track different types of obstacles simultaneously. Besides, in current research, regular motion models are mostly adopted to track pedestrians, such as CV, CA, and so on, the uncertainty in pedestrian motion is not well considered. If the motion characteristics of different types of obstacles cannot be accurately modeled, the performance of the target tracking algorithm will be degraded, which will lead to errors in the following decision-making and motion planning algorithm. To solve the above problems, this paper uses the Interacting Multiple Model (IMM) method to describe obstacles’ different motion characteristics in different motion stages. Multiple different motion models are simultaneously used to describe the motion of obstacles to make it as close as possible to the actual situation. Furthermore, due to different motion characteristics of cars and pedestrians, by using a data set, the analysis is carried out to obtain the motion model set for car tracking and that for pedestrian tracking, respectively. Then combined with the IMM algorithm, the two-step adaptation of the target’s motion model during the tracking process is realized based on object class information: motion model set adaptation and motion model weight adaptation. The number of motion models included in the motion model set for target tracking is also reduced, which reduces the calculation amount of the algorithm. For the uncertainty in pedestrian motion, Random Motion (RM) model is introduced to improve the algorithm performance in pedestrian tracking. The effectiveness of the proposed algorithm is verified through experiments on a data set, it achieves 85.99% in Multiple Object Tracking Accuracy (MOTA), 75.06% in Multiple Object Tracking Precision (MOTP), 85.03% in Mostly Tracked (MT), and 4.10% in Mostly Lost (ML), and the average time for the algorithm to process one frame of data is 12.05 ms.


Subscribers can view annotate, and download all of SAE's content. Learn More »


Members save up to 17% off list price.
Login to see discount.
Special Offer: Download multiple Technical Papers each year? TechSelect is a cost-effective subscription option to select and download 12-100 full-text Technical Papers per year. Find more information here.