I. Introduction
For an autonomous vehicle’s perception system, intention or maneuver prediction is crucial. It gives the likelihood of actions, under a limited range of options, that traffic participants could make while interacting in roads (e.g., crossing roads and lane change) [1]. This information has been incorporated into many automobile systems, including Advanced Driver Assistance System (ADAS) and driverless vehicles [2], [3]. As stated in AbuAli and Abouzeid [1], driving behavior modeling is a complex task since there are multiple attributes that influence a driver or road users, such as mood, health condition, weather, and others [4], [5]. Moreover, using a hierarchical division into strategical, tactical, and operational levels, this task can be further investigated [1], [6], [7]. The strategical level addresses decisions regarding routing and conform parameters. In turn, the tactical level concerns the choice of actions towards a short-term goal, usually defined as lateral (e.g., lane change, and left/right turn) and longitudinal (e.g., accelerate and decelerate) maneuvers. Finally, the operational level tackles safety, physical limitations, and road geometry.