Flash: Fast and Light Motion Prediction for Autonomous Driving with Bayesian Inverse Planning and Learned Motion Profiles

Antonello, Morris; Dobre, Mihai; Albrecht, Stefano V.; Redford, John; Ramamoorthy, Subramanian

doi:10.48550/arxiv.2203.08251

Cited by 2 publications

(7 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Goal-based methods [10], [11], [12], [13], [14] identify a number of potential future targets that each agent may be heading towards, determine likelihoods of each, and produce predicted trajectories towards those goals. Flash [6] uses a combination of Bayesian inverse-planning and mixturedensity networks to produce accurate predictions of trajectories in highway driving scenarios. Goal-based methods have the advantage of using the map to inform trajectory generation, and can use kinematically-sound trajectory generators.…”

Section: Existing Methodsmentioning

confidence: 99%

“…Existing models using the INTERACTION dataset have demonstrated good results on closest-mode evaluations (mi-nADE / FDE / MR) [2], [3], [4], while models using NGSIM have shown good results on probabilistic evaluations (pre-dRMS, NLL) [8], [17], [9], [6]. The joint task of producing diverse predictions at the same time as maintaining good probabilistic accuracy has generally not been addressed with interactive prediction.…”

Section: Existing Methodsmentioning

confidence: 99%

“…1) predRMS: the root-mean-square (RMS) error of the most probable predicted mode is calculated for a number of timesteps, over the set of instances of the dataset (1 ≤ n ≤ N ) as shown in (8), where µ i is the predicted position for the most probable mode i = arg max m (W m ) as used in [5], [9], [6]:…”

Section: B Evaluation Methodsmentioning

confidence: 99%

“…The spatial distribution mode weight W n is trained using two loss terms, a NLL-based loss that minimises the NLL score, and a training-weight loss that trains W n to be similar to the training mode distribution W r . The NLL-based mode loss is shown in (6), where W n is trained and µ, Σ are constant.…”

Section: A Trainingmentioning

confidence: 99%

“…Probability estimates can be evaluated using measures such as predicted-mode RMS (predRMS), which evaluates mode probability estimates along with distances between predicted trajectories and observations. Existing methods such as [5], [6] have studied prediction using probabilistic predRMS evaluations on the NGSIM highway driving dataset [7].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

DiPA: Probabilistic Multi-Modal Interactive Prediction for Autonomous Driving

Knittel¹,

Hawasly²,

Albrecht³

et al. 2022

Preprint

View full text Add to dashboard Cite

Accurate prediction is important for operating an autonomous vehicle in interactive scenarios. Previous interactive predictors have used closest-mode evaluations, which test if one of a set of predictions covers the ground-truth, but not if additional unlikely predictions are made. The presence of unlikely predictions can interfere with planning, by indicating conflict with the ego plan when it is not likely to occur. Closestmode evaluations are not sufficient for showing a predictor is useful, an effective predictor also needs to accurately estimate mode probabilities, and to be evaluated using probabilistic measures. These two evaluation approaches, eg. predicted-mode RMS and minADE/FDE, are analogous to precision and recall in binary classification, and there is a challenging trade-off between prediction strategies for each. We present DiPA, a method for producing diverse predictions while also capturing accurate probabilistic estimates. DiPA uses a flexible representation that captures interactions in widely varying road topologies, and uses a novel training regime for a Gaussian Mixture Model that supports diversity of predicted modes, along with accurate spatial distribution and mode probability estimates. DiPA achieves state-of-the-art performance on INTERACTION and NGSIM, and improves over a baseline (MFP) when both closest-mode and probabilistic evaluations are used at the same time.

show abstract