Collision Probability via Scoring Multiple Hypotheses (P-SMH)#
Description#
Similar to other probability-based approaches, Sánchez Morales et al. propose to assign probabilities to predicted trajectories and accumulate them into a collision probability [Morales2019]. The motion of the ego is modeled by a two track model. Due to less information being known with a reasonable accuracy for the other actors, a one track model is used for those. Pedestrians have the ability of changing direction, velocity, and acceleration in a finite set of steps under given constraints. Once the number \(N\) of trajectories for the ego and total number \(M\) of trajectories of all other actors is determined, one can compute the collision probability as
where \(\chi^i_j\) equals one if and only if the \(i\)-th trajectory of \(A_1\) and the \(j\)-th trajectory of the actors in \(\mathcal{A} \setminus A_1\) lead to a collision, and \(p_{\mathit{A_1}, i}\) resp. \(p_{({\mathcal{A} \setminus A_1}), j}\) are the probabilities of the trajectories being realized.
Properties#
Run-time capability#
Yes, demonstrated by evaluation [Morales2019]
Target values#
None found
Subject type#
Any, but requires behavior and dynamic model of subject
Scenario type#
Depends on definition of models
Inputs#
Static and dynamic objects as well as their state, estimated bounding boxes, ego: see TT model, other vehicles: see OT model
Output scale#
\([0, 1]\), probability, ratio scale
Reliability#
High, as the consideration of multiple futures and their likelihoods makes it robustly follow changes in criticality [Morales2019]
Validity#
High, due to branching predictions and likelihood estimation, but depends on the validity of the motion model and probabilities, initial simulative validation results exist [Morales2019]
Sensitivity#
High, but depends on the validity of the motion model and available computational power, no analysis of false negatives was performed in initial evaluation [Morales2019]
Specificity#
High, an initial evaluation found no false positives by the metric [Morales2019]
Prediction model#
Time window#
Unbound, but longer prediction horizons at a constant number of predicted trajectories lower reliability
Time mode#
Branching time