Interaction-aware occupancy prediction of road vehicles

Abstract

A crucial capability of autonomous road vehicles is the ability to cope with the unknown future behavior of surrounding traffic participants. This requires using nondeterministic models for prediction. While stochastic models are useful for long-term planning, we use set-valued nondeterminism capturing all possible behaviors in order to verify the safety of planned maneuvers. To reduce the set of solutions, our earlier work considers traffic rules; however, it neglects mutual influences between traffic participants. This work presents the first solution for establishing interaction within set-based prediction of traffic participants. Instead of explicitly modeling dependencies between vehicles, we trim reachable occupancy regions to consider interaction, which is computationally much more efficient. The usefulness of our approach is demonstrated by experiments from the CommonRoad benchmark repository.

Set-Based Prediction of Traffic Participants Considering Occlusions and Traffic Rules

“…This work significantly extends our previous work on setbased prediction [14], [67], [69], [70] and other previous works, especially [16], by considering 1) all safety-relevant occluded vehicles, pedestrians, and static obstacles, 2) priorities of traffic participants at intersections, 3) safe distances to the ego vehicle, 4) limited turning radii of vehicles, and 5) by validating the prediction in real-world experiments.…”

“…The work of [67] is extended in [16] by considering occlusions. Set-based prediction is also able to consider interaction between traffic participants [69] and formalized traffic rules [14], [70]. The predicted occupancy sets can also be weighted by probabilities [71], [72] d) Occlusion: The risk from occlusions is tackled either by shrinking the field of view over the prediction horizon [73]- [76] or by introducing and predicting individual, potentially present obstacles (aka phantom or virtual objects) [1], [16], [77]- [85].…”

“…Line 13 of Alg. 1 optionally considers the interaction between vehicles as described in [69], e. g., that a vehicle cannot tunnel through a stationary vehicle.…”

Membership-based Manoeuvre Negotiation in Autonomous and Safety-critical Vehicular Systems

“…)), that is, ∆t M is the first time interval after which p i or p j is guaranteed to have left the intersection. Further constraints in the predicted vehicle motions can be applied if multiple vehicles are present in a traffic scene [10]. Let {p i } i=0 N be N vehicles in a queue on a one-lane road where p 1 is the leading vehicle.…”

Goal-Directed Occupancy Prediction for Lane-Following Actors

“…In contrast, our method uses a broad set of candidate goals as input to a deep network for the purpose of occupancy prediction, rather than as a prior on strict lane-following behavior. Structured methods have also been previously applied to the problem of lane occupancy prediction, for instance predicting occupancy by modeling lane-following maneuvers and actor-actor interactions using explicit policies [14]. Our approach also performs occupancy prediction, but with less explicit structure on motion.…”