StarNet: Joint Action-Space Prediction with Star Graphs and Implicit Global Frame Self-Attention

Abstract

In this work, we present a novel multi-modal multi-agent trajectory prediction architecture, focusing on map and interaction modeling using graph representation. For the purposes of map modeling, we capture rich topological structure into vector-based star graphs, which enable an agent to directly attend to relevant regions along polylines that are used to represent the map. We denote this architecture StarNet, and integrate it in a single-agent prediction setting. As the main result, we extend this architecture to joint scene-level prediction, which produces multiple agents' predictions simultaneously.The key idea in joint-StarNet is integrating the awareness of one agent in its own reference frame with how it is perceived from the points of view of other agents. We achieve this via masked self-attention. Both proposed architectures are built on top of the action-space prediction framework introduced in our previous work, which ensures kinematically feasible trajectory predictions. We evaluate the methods on the interaction-rich inD and INTERACTION datasets, with both StarNet and joint-StarNet achieving improvements over state of the art.

DiPA: Probabilistic Multi-Modal Interactive Prediction for Autonomous Driving

“…Graph-based methods [15], [4], [2] combine map information and agent positions into a common representation, commonly processed with a graph neural-network in an encoder-decoder framework. Jia et al [16] extend a graphbased model to consider the scene from each agent's point of view rather than using a single central agent, and recurse each agent's model of other agent behaviours.…”

“…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.…”

“…Existing methods that have used the INTERACTION dataset such as [2], [3], [4] are examples of predictors that are able to capture ground-truth behaviours with one of a set of predicted trajectory modes, as evaluated using minimum average-or final-displacement error (minADE/FDE) and miss-rate (MR) (see Section IV-B for a definition). These measures compare the closest predicted mode with the ground-truth.…”